( Later published in highly redacted form in 'Inference; International Review of Science, http://inference-review.com/article/the-excitable-mitochondria, and publicly reviewed on Hacker News / YCombinator; https://news.ycombinator.com/item?id=13088772 )
Plunging an electrode deep into the brain is one way to ablate hundreds of voxels of white matter. Unfortunately, there is no ‘before and after’ imaging study that could possibly tell you where in the cortex you would find the actual neuron bodies that had their axons clipped. Pyramidal cell axons project out of the cortex and branch in the white matter to contact many potential targets -- the spinal cord, deep brain stem nuclei, and contralateral hemisphere to name a few. Compromising the business end of a few thousand random neurons might be an acceptable side effect for giving a voice to the locked-in, an arm to the paraplegic, or steadiness to the tremoring hand of a Parkinson’s victim, however, for the elective augmentative procedures many future healthy subjects will eventually want and need, it may be a deal breaker.
Plunging an electrode deep into the brain is one way to ablate hundreds of voxels of white matter. Unfortunately, there is no ‘before and after’ imaging study that could possibly tell you where in the cortex you would find the actual neuron bodies that had their axons clipped. Pyramidal cell axons project out of the cortex and branch in the white matter to contact many potential targets -- the spinal cord, deep brain stem nuclei, and contralateral hemisphere to name a few. Compromising the business end of a few thousand random neurons might be an acceptable side effect for giving a voice to the locked-in, an arm to the paraplegic, or steadiness to the tremoring hand of a Parkinson’s victim, however, for the elective augmentative procedures many future healthy subjects will eventually want and need, it may be a deal breaker.
Ideally, surgeons would come to the table with a full engineering CAD model of their patient’s brain. In addition to modeling the relevant neural compartments, a thorough preoperative workup would also simulate the behavior of the local microvasculature (now known as the ‘angiome’), and the larger topology of CSF flow (the so-called ‘glymphatic’ system). Included in this package would be a dropdown menu of ‘brain physics’ modules that capture not only electrical conduction, but also parameters like thermal conductivity, bulk moduli and pressures, and the diffusion and convection of fluids flowing through various foramina, ventricles, and vessels.
While it is often noted that the brain has around 100,000 miles of glial-insulated axon tract, there is also over 100,000 miles of capillaries insulated by glial cell endfeet in the brain. When looked at from above, the brain appears to be as much a machine built to artfully distribute fluids as it does one to transport electrical signals. For comparison’s sake, if we look at how the muscles in the legs are powered, we find an informative dichotomy: delivery of a pressurized nutrient broth of glucose and oxygen, and removal of lactic acid and carbon dioxide only requires a femoral artery about a half an inch wide. But the more ethereal, ‘information-bearing’ electrical energy is supplied to the same muscle through a massive sciatic nerve some three-quarters of an inch wide. (1)
While software that keeps tabs on the fluid amenities of the nervous system would help insure a patient can get up off the table after a radical cerebral installation or deletion, the real bread and butter of the operation would be simulating what happens to their mind. Consider a patient request to eliminate some immaterial mental malady -- maybe it is a desire to squelch an intrusive internal voice, or perhaps to subdue the burning phantom of a supernumerary limb that quizzically appeared after a stroke. Traditionally, doctors would beat these bugbears over the head with mind-numbing drugs. Today however, many might presume that the ability to directly eliminate any offending neural elements using minimally invasive ablation with directed energy beams should just around the corner.
In order to do this while sparing surrounding friendly fire we would not only need pinpoint accuracy in our instruments, but also point-to-point knowledge of the brain. Without something like a high resolution ‘connectome’ on hand to trace projections through a 3D model, any outcome would be hard to predict.
While the main focus of the current and proposed brain megaprojects in Europe and the US is to understand how the brain works through simulation, connectomes, and animal brain activity maps (BAMs), I would argue that these approaches are naive and hopeless. Yes, I think we need these software tools to probe and visualize brains, but the reason for building these tools is not be that they will magically delivery any understanding. What they will do is help us build and place more powerful implants. While these implants will in turn lead to better maps, it is only through their direct and personal use -- the first and second hand human experience of implant interaction at the level consciousness -- that the understanding of the brain we seek will be gained. (2)
Brains by design
Neuroscience now supplies an open-ended stream of ever-finely scaled cellular and molecular detail. Marshalling this divergent fount of knowledge into CAD form presents several challenges. For one, nobody seems to have a plan for creating these whimsical software packages, let alone selecting what data to feed it. Electrical engineers, by comparison, have plenty of CAD-like options for creating their circuit board connectomes. They call these diagrams of the connections within or between individual chips, ‘netlists’. Once the logic performance of a circuit is verified, and its corresponding netlist generated, the designer performs the ‘layout’ or the place and route step, to position all the components in space. Simulation tools then probe worst case scenarios to see how the behavior varies from the ideal ‘perfect connections’ of the netlist.
As is the case for brains, there are many possible layouts that correspond to the same netlist. To flesh out a netlist using real world constraints the circuit engineer has a marvelous program called ’autoroute’. In addition to optimizing timing, and which paths wires should take, it can also specify how wide and how much copper should be laid to meet restrictions on capacitance and inductance imposed by high frequency signals. Similarly, for low power applications a little extra extra copper can be specified in order to decrease path resistance and save battery life. For something like a seven-layer board with hundreds of thru hole ‘vias’ linking different planes, the layout can get fairly complicated. In the layout of nervous systems, much these same considerations apply for wiring and myelination. In general, you want to minimize wire and power requirements while maximizing speed by placing high traffic links close to each other using the fattest, low resistance paths for carrying the highest rate or time sensitive signals.
For the task of safely and reversibly implanting brains with hardware want we really want is a detailed ‘parametric’ CAD model of the kind normally used by mechanical engineers. In CAD design you generally work from the ground up building your parts from completely-defined one dimensional sketches. The parts are then hierarchically mated together to form an sub-assemblies, and ultimately, the top-level assembly model. Mechanical realism is algorithmically ensured by not letting the build/rebuild step (essentially the compile step) continue in the presence of physical impossibilities. Interference fits, collisions, loose parts, or stuck parts will pop out when you grab the model with the mouse and shake it. The offending members are highlighted with ominous red ‘X’s’ in the annotations menu off to the side making it easy to supply the appropriate code equivalents of duct tape or WD-40.
Nothing in the model is considered complete until its scale and operation are fully constrained. In practice this means that every feature either needs an explicitly declared dimension, or else a relation to a declared dimension that dictates how the part model will change when those primary dimensions are tweaked. This is the whole point of the model -- to have something where the global form and behavior can be controlled by simply toggling a few critical design points. In practice, these relations are are typically scaling and transformation functions embedded at the base level into the sketches. The ‘mates’ that describe how parts interact are canned elementary descriptors like ‘mirror symmetric to’, ‘co-planar with’, ‘concentric with’, or slightly more advanced user-defined operations. This ‘easy to vary’ way of capturing the essential elements of a complex system makes it practical to explore many closely related designs, or for that matter, brains.
Where is our engineering CAD model of the brain?
At the heart of this question lies an ineluctable paradox. The field’s undisputed founding father, Santiago Ramón y Cajal, single-handedly sketched every major variety of cell and observable feature of the nervous system, in exacting detail, well over a century ago. Yet today, despite astounding advances in genetics and cell biology, there is still no practical blueprint to guide the implantation any type of hardware in the brain. Similarly, there is also nothing available to guide implantation of stem cells into the brain, let alone guide their integration into the site of a simply bone reconstructive surgery. When stem cells are used during surgery, they are not integrated as ‘structural’ materials at all. Instead, they are literally just dumped in with no more expectation that the provisioning of the one or two special chemical products they synthesize and secrete.
I’d argue that in order to go forward in building the engineering model of the brain we so sorely need, neuroscience as a whole will need to retrace its steps and address the many critical loose ends that were simply glossed over -- those basic things that the field collectively thinks it must know, but in fact does not. By that I mean all those tiny anatomical intangibles that the forefathers clearly encountered, puzzled over in the literature, and often theorized about, but were not yet equipped to answer. Many early conundrums were geometrical in nature; For example, why neurons adopt a polarized axo-dendritic form and why their axon is chiral; which direction myelin spirals going down the axon, and across adjacent axons geared together in nerves, and among the many arms of a single oligodendrocyte; why Schwann cells are single use items while oligodendrocytes are multiplexed; and how whole axon tracts warped and decussated as the the genetic body plan twisted and inverted during evolution of vertebrates.
Many of the brains signature functions have also been incorrectly or insufficiently explained -- the full multiphysical nature of the action potential(3); how receptors detect odorants(4), how receptors detect anything(5); how outer hair cells mechanically oscillate upwards of 40kh on nanometer scales, how bats could possibly detect features with nanosecond timing differences; how chemomagnetic sensation works, and why neurons use the transmitters they do -- for example, why invertebrates retained a highly enriched glutamate blend to fuel their neuromuscular junctions while vertebrates modified their transmitter mix to nearly pure acetylcholine?
In the face of all this ignorance there is hope. Imaginative answers, or at least good guesses to each of these questions can be supplied if you have the right vantage point. For example, if you were to start modeling a Schwann cell you would reach one inescapable conclusion as soon as you added the first dimension -- they are massive. Operating near its theoretical metabolic maximum during peak development a single Schwann cell 15um in diameter x 1500um long with 150 turns must rapidly crank out enough lipid-protein mortar to fill a single-unit myelinating volume of 150,000um^3. On the other hand, an oligodendrocyte only 500um-long x 5um and 50 turns per arm with similar synthesis capacity might be able to put out a whopping 50 myelinated units.
If you were to then begin adding myelinating compartments to axons in a CAD model, and bundle them in a nerve fascicle or tract, the question of how they should mate and slide together spawns its own potential answers: does the actual force driving the spiralling inner myelin tongues derive from beyond the myelin itself, perhaps from tiny chiral torques mechanically supplied by the axon spiking away inside it? ie. that axons wrap up their own myelin. Yeah I said it, neurobiologists have no idea HOW myelin wraps, and there will be no realistic CAD model till they do. The experts in the field are aware of these truths, and they have been documented.(6)
Connectomes, Brain Activity Maps, Molecular Tickertapes, and Neural Barcodes
In 2013 Jeff Lichtman hit peak connectomics with his fantastic 1-terabyte in-silico recreation of a tiny grain of mouse cortex. One of his particularly compelling reconstructions contained a cylindrical patch of tissue surrounding a single apical dendrite of a pyramidal cell. Lichtman’s team exhaustively mapped every mitochondria, every postsynaptic density, and nearly every vesicle in 774 synapses made onto the dendrite by some 680 surrounding local nerve fibers. With this approach they also managed to extract the complete membrane topology for this one-billionth volume of a mouse brain. The few advances made since have only served to pound home the obvious state of affairs in connectomics -- it would take of lot of CAD rendered terabyte ‘Lichtman cubes’ to scale up to something the size of our brains. (7)
In fact, at the height of this connectomania there was even a paper devoted solely to evaluating the economic impact of pursuing a full brain connectome, a field the authors whimsically dubbed Conneconomics. These authors had also previously produced a fantastical paper entitled ‘Physical Principles for scalable Neural Recording’ where they calculated the ultimate limits (namely space, power transmission, and heat dissipation) by neural recording, stimulation, and communication hardware that would be used in brain mapping. This was then followed by a their equally evocative paper entitled, ‘Rosetta Brains: A Strategy for Molecularly-Annotated Connectomics’. In it they detailed how molecular scale ‘barcodes’ written by tract-tracing viruses and bulk nucleic acid sequencing could be scaled up to generate whole-brain connectomes(8).
At the height of the mad scramble for BRAINI initiative funds these guys, mainly together with genetics pioneer George Church, also envisioned and patented related techniques to create ‘molecular tickertapes’ that would use a modified polymerase to write brain activity maps into DNA. Presumably, these extrachromosomal DNA nucleoids would later be amplified, read out, and decoded. Now there is no doubt that this still mostly theoretical work is completely awesome. However, there is in my mind a major oversight in all this. Namely, if anyone was actually serious about generating connectomes and maps they wouldn’t be talking about grey matter connectomes at all at this point, they would be talking about white matter connectomes. Furthermore, if the brain contains as much vascular wire as axonal wire, and both are fully vested by a composite intermeshed network of glial cells, one might wonder why nobody has called for a glial connectome(9)?
A case in point here is popular Eyewire project which seeks to map every miniscule leaf of every dendritic tree in the retina. Mind you, this is the same retina that routinely rewrites whole swaths of physical connectivity on 5 or 10 minute timescales during dark adaption. This same anatomical trick was apparently appreciated by pirates who were rumored to have kept one eye patched and ready to spot enemies at night when they came out to the deck from a well lit room below. The only reason that neuroscientists go for the grey matter is that they happen to be really good at making detailed EM sections that provide a lot of information about extremely tiny pieces of brain. Trying to map the white matter pathways this way is like bringing a magnifying glass to take see the Grand Canyon; they are simply too big to analyze whole by EM. Unfortunately, they are still too small to accurately resolve using mapping techniques like diffusion MRI tractography. That’s a bit of shame because while the grey matter is highly labile, the thickly myelinated white matter is highly stabile. If we ever hope to have something we might call a ‘living connectome’ -- a connectome that does not require destroying the brain and replacing it with a sparsified ghost -- let alone a fossil one, we are going to have to be a bit more creative.
The full scale membrane topology Einstein’s grey matter might be great attraction at a museum, but it probably wouldn’t be of much use to Einstein. Aside from the obvious data problem such a file would present, a living grey matter netlist is something that is physically impossible. You can never step into the same brain twice because neural connections change much faster than the time it would take to read them all out by any method constrained by the laws of physics. The white matter, on the other hand, would not. White matter is also highly predictable; Neighboring axons surrounding any given axon tend to signal in much the same direction, lie in much the same orientation, and are likely to be wrapped by arms from the same oligodendrocyte.
If one was writing an actual white matter connectome they might forsake a neurocentric description altogether and instead use the coordinates of each oligodendrocyte together with the fifty or so minion axons wrapped up by its arms. As far as file formats for crunching a neural netlist go, there is no mad rush to standardize because the gray matter connectomists haven’t even figured out what constitutes a ‘connection’.
Fundamental Principles and Neuroscience’s ‘Missing Link’
For several decades Eric Kandel’s ‘Principles of Neural Science’ has served as the de facto entry point to all things neuro. It is both a crash course for intrepid cross-disciplinary dilettantes, and rite of passage for its primary students. Although the bulk of this text has remained fairly solid, improvements to its foundation have been incremental. In 2015 the field got what is arguably its second great learned tome in the form of Sterling and Laughlin’s ‘Principles of Neural Design’. While descriptive, a gaping void still persists at its conclusion: from whence does all this complexity arise?
The once heretical but now seemingly endemic finding that organelles, notably mitochondria, are transferred whole or piecemeal from neuron to neuron presents a unique challenge to any ‘neuron doctrine’ touting discrete parts -- namely, if organelles don’t belong to cells, what’s a cell(10)? In cancer, several organ systems partake of a curious mitochondrial dynamic in which they initially become tumorigenic as a result of losing respirative power, and then subsequently lose the ability to repair their DNA. This ‘metabolic origins’ view of cancer throws a bit of a monkey wrench into the old nuclear DNA ‘mutations first’ oncogenic tumor-suppressor paradigm. A mind-numbing finding is that after reaching an uneasy steady-state truce with the efforts of the body and doctors to contain it, the seemingly quiescent cancer cells can suddenly go rogue and become metastatic upon donation of fresh mitochondria from the nearby healthy cells(11). One might ask, what kind of way to run a ship is that? This hither-to-now underappreciated mitochondrial plasticity is but the tip of the iceberg.
More pertinent for us here is the question of how plastic whole nervous systems can actually get? In other words, how labile is the constitution and form of neurons? Turning high resolution microscopes to the brain with a keen eye for anomaly researchers have found that the hallowed zone known as the ‘axon initial segment’ -- the place where the axon roots itself at the cell and its organizing centriole -- moves around the cell quite alarmingly. For example, in up to 10% of cells in parts of the hippocampus the axon migrates far afield of its ‘normal’ location and roots itself on a proximal dendrite(12). Needless to say, these kinds of findings are a bit awkward. Looking further afield with better imaging techniques for delineating axons via the chiral cytoskeletal hooks attached to their tubulin this kind of behavior will likely be found to be quite common. One might wager a neuron with two axons could be theoretically possible. Or maybe not.
It was not until Ramon Cajal’s famous inaugural Neuron Doctrine that it was fully accepted that nervous systems were actually made from discrete parts. Although other pioneer’s like Czech anatomist Jan Purkinje made early forward contributions, Cajal was the first to declare that the neuron, with all its dendrites acting together to feed a single polar axon, is the fundamental unit of the nervous system. Prior to Cajal, the so-called ‘continuous syncytium’ model espoused by Camillo Golgi’s reticular theory had held sway(13).
What I want to suggest is that the fundamental discrete units of nervous systems, and for that matter every organ, are not neurons or other cells, but rather the mitochondria. Furthermore, if neurons routinely compete for, harvest, or otherwise exchange synaptic and other structural wetware, possibly even whole axons, than the differences between the old synticial and the newer discrete model would seem significantly less sharp.
In a word, the primary feature we expect for an irreducible neural component -- and similarly seek of mitochondria -- is excitability. As we’ll see below, mitochondria take excitability to an extreme.
From a strictly mechanical point of view, the brain really is a kind of a syncytium. While the electrical component of a spike that invades the synapse is chemically transformed according to some only vaguely understood transfer function into a vesicle message, the mechanical portion of the spike is transformed and propagated into and across the synapse in a way dictated by the mechanical impedance of local membrane, cytoskeleton, and synaptic matrix(14).
If mitochondria are actually the fundamental units of the nervous systems, then the parts to which the most care and attention need be applied in a CAD model should be the mitochondria. The immediate question then, of “where is our brain model?”, for all intents and purposes is significantly reduced to the question of “where is our mitochondria model”? In the now beleaguered Eurobrain project, from which visionary founder Henry Markram was recently deposed, we might opine that there was not a single mitochondria to be found everywhere.
Can we construct model brains from principles?
Sterling and Laughlin’s ‘Principles‘ propose at least a partial explanation for why things look like they do in any given cube of tissue. Specifically, their cost saving analysis of different neural circuits suggest that there is some underlying logic to why neural connections have the precise diameters, numbers of vesicles, vesicle release probability, and spontaneous and maximum firing rates that they do. They also espouse ‘computing with chemistry’ rather than neural circuits whenever possible, ‘minimizing wire’ in converging and diverging circuits, and optimally locating cell bodies within the folded layers of various cortices and nuclei.
Unfortunately, their principles also suffer from narrow perspective. By that I mean that are they are derived entirely from electrical considerations; primarily how much energy it takes to generate an electrical spike in an axon, how much ATP it takes to pump the ions back out, and how noisy the channels and receptors are. If asked to sum up their musings on why brains look like they do in a single parameter that sets everything else in stone it would simply be ‘the electrical resistance of cytoplasm’. However, in contrast to the purely electrical model, the full mechanical nature of the pulse propagation in neurites, particularly in those where myelin participates in carrying some of the energy of the signal, suggests that sending spikes appreciable distances may be much more efficient than has been previously assumed.
With these forebearances in minds, where else might we look for the underlying principles for constructing neural networks that are even more general than those Sterling and Laughlin? Another famous Czech, the late neurosurgeon Karl Pribram, spent a lot of time trying to characterize exactly what it is that brains do, and how they do it. He famously came up with the four ‘F’s of evolutionary biology; feeding, fleeing, fighting, and fornicating. These top level behaviors are leagues above the crude functions, algorithms, or computations that neural modelers might seek to embed in their artificial networks.
Although simulated networks attempt to connect and integrate signals like real networks, they invariably ignore a couple important features of actual neurons that are only rarely mentioned. First, neurons connect to themselves. These ‘autapses’ may represent a minor portion of their overall synaptic budget, but they are tightly controlled by a suite of recognition molecules. The second fact, is that when neurons do connect to another neuron, they hit it in spades with a large bloom of synapses that would defy any characterization by the single ‘synaptic weights’ used for artificial neural network connections. Cajal’s drawings in his ‘Butterflies of the Soul’, show retinal cells dedicating nearly their entire axonal or dendritic arbor to just one or two vertically adjacent partners. I think these are design features rather than bugs, perhaps they are even principles.
There is another, much more nasty paradox lurking behind theoretical efforts to establish laws by which networks of cells might create complex behavior. The hideous riddle that Pribram’s four ‘F‘s’ present, is that while the network modeler seeks complex functions from the interactions of approximated neural units, each real neuron -- essentially a domesticated and docile protist -- already contains each behavior in full. In fact, if you accept the horizontal transfer of genetic material as part and parcel of the copying and reproductive behavior, then not only protists, but far simpler bacteria display all four evolutionary motivators as well.
In other words, the real conundrum we face is not figuring out how brains are wired to solve problems, but rather to first figure out how single cells solve problems — i.e. where is their brain?
When single cells reigned, there was no brain. Any computation was presumably ‘done with chemistry’ in accordance with the platonic ideals of Sterling and Laughlin. In their conception, this computation occurs at the level of receptors and the associated signalling pathways inside individual cells. Perhaps one way we might intuit how complex nervous systems later materialized from this base is to interpret their detailed network structures as spatial optimizations of the specific metabolic pathways used generate the unique transmitters completing each leg of the circuit.
Neuroscience has a lot to say about receptors and channels, but little about why neurons use the transmitters they do. At the ground floor level, I would suggest that transmitter chemistry must be ideally suited to whatever computational task it is that needs to be done at a given synapse. Some transmitter molecules might best be likened to brute force irritants used to create a little synaptic breathing room. Others transmitters might be metabolic dead ends that are simply expressed into the synaptic space as wastes. Still others might best be likened as expensive gifts or metabolic candies plied to attract.
These transmitters are generally not shuttled whole through each junction, but rather it is their parts and physical influence -- their functional groups, protons, electrons, and free energy -- that is ultimately what is transduced. In some circuits, these parts are directly rendered upon enzymatic transformation and receptor activity in the synaptic cleft. Others (like the glutamate/glutamine transmitter cycle), require uptake of the largely complete molecule by the postsynaptic cell, or other glial hosts, in order to actuate their functional groups. One thing many transmitters seem to have in common is that their metabolites are key participants in the signature transport shuttles that control the metabolism of mitochondria. These organelles are invariably concentrated and closely apposed to each other at active presynaptic and postsynaptic sites giving one the direct impression that they are there to communicate.
Clues as to how cells use mitochondria, and conversely how mitochondria might use cells to communicate beyond their borders are now most succinctly emerging from the improbable haunts of our immune and hematopoietic systems. White blood cells not only utilize mitochondria for the killing power of their oxidants, but they actually sort and translocate them to the plasma membrane according to their depolarization states, and then dangle them as immunogenic lures in the bloodstream to recruit defenses against invaders in the body(15). The reason this strategy works is because ailing mitochondria have a penchant for exposing their bacterial-style formylated peptides, cardiolipin, and DNA (mainly oxidized methylguanosine) on their surfaces.
When bloodstream is breached somewhere, thrombin and other factors transform quiescent platelets first into prickly amoeboids, and then into giant kamikaze ‘superplatelet’ balls within seconds. They do this by FM modulations to the base carrier frequency of calcium oscillations in their mitochondria, which together with a sudden reversal of proton pumping in their ATPase machinery irreversibly sets these cells upon their supernova fate(16). In the brain itself, astrocytes have co-opted these peripheral immune system actuators, like glycoprotein CD38 for example, in order to initiate the programmed transfer of mitochondria from astrocytes to neurons to repair damage after stroke(17). Conversely, the two main mitochondrial proteins implicated in controlly mitochondrial dynamics in the brain in Parkinson’s disease (Pink1 and Parkin), are actually the key regulators in the peripheral adaptive immune system which inhibit mitochondrial antigen presentation. The fact that mitochondria are so intimately involved with presenting antigens is somewhat astonishing and reflects their full integration into every cell function.(18)
The point of listing these recent findings is that the underlying activators can in each case be traced down to base level self-excitations occurring deep within the cristae of mitochondria. These nonlinear ‘mitoflashes’ are now known to be proton-triggered and sensitive to just a small number of protons persisting less than 2 ns and diffusing just 2nm in the matrix(19). Although mitochondria can not generally slew their mitoflash potentials around or repeat them at a rate anywhere near what neurons can do(), they do have a few tricks of their sleeves. When watched by parametric imaging these mitoflashes follow up their local proton and membrane voltage fluxes with a predictable sequence of calcium and redox ‘sparks’. What’s particularly interesting is that mitoflashes and spontaneous oxidative bursts have been found to be accompanied by significant mitochondrial shape changes -- namely, reversible contractions. Although the mechanical aspect of mitoflash was only initiated at a rate of 0.6 per hour in a given mitochondria, there are thousands, sometimes tens of thousands of mitochondria in each neuron(20).
If neurons are packing this kind of excitatory power then certainly a large army of these mitochondrial ‘one-shots’ (even if they take a while to recompose themselves) should influence the excitability of the neuron itself. It is not such an enormous logical leap to then suppose that mitoflash, particularly when occurring near membranes with high channel densities could initiate neural spikes, and similarly, that neural spikes could initiate mitoflash.
As mentioned, it has long been appreciated that the familiar ‘action potentials’ of neurons are multiphysical phenomena whose mechanical displacement, pressure, and counterintuitive heat absorption and release each follow their own predictable time course and spread. The electrical blimps amplified on oscilloscopes, or the narrowly prescribed wave of inrushing sodium ions and outrushing potassium ions one might see in a drawing are a bit misleading. When axons fire, there is no spatially localized pulse; the whole axon depolarizes. For example, if the spike lasts for a millisecond, and its expanding front travels at 100 meter/second, then we are talking about a physical disturbance that would extend some 10cm.
To the dismay of neurophysiologists who have been using the so-called ‘antidromic collision technique’ to trace axonal connections (this includes myself), real spikes travelling in opposite directions on axons can pass right through each other. If these observations, originally discovered in worms, hold true more generally then the neuroscience literature is likely riddled with spurious results. These persistent spikes would be more reflective of mechanical soliton-like waves then of the annihilating electrical spikes one would get with the inactivating ion channels and mandatory refractory periods of the Hodgkin Huxley model. Oops(21).
What is the Brain?
One of the most astounding revelations of modern genetics and the comparative phylogeny it enables has been tracing the origins of mitochondria in the symbiosis of two different bacteria with complementary metabolisms. The details of different theories vary a bit, but the main idea is that as the genome of one bacteria shrunk (the presumptive mitochondrial precursor), the genome of the other grew (the host) and established the nucleus of the first Eukaryote(22). After eukaryogenesis, mitochondria turned their attention to driving cell differentiation and multicellularity (and the accompanying transformation to gender binary sexual reproduction), and then ultimately to crafting the nervous system.
Sterling and Laughlin’s original cover art was a typical EM image of a section of grey matter. Like so many other neurobiology texts that offer much the same familiar images, they never get around to adequately explaining this highly recognizable structure. What we see is very clear; it’s mitochondria, as thick as weeds in an unkempt field. They’re all nestled inside a convoluted system of narrow tubes -- passageways that they themselves construct by power of their own respiration along with the strategic deployment of a cache of genes long since offloaded to centralized nuclear storehouses.
In other words, we are looking at an elaborate antfarm. If you watch ants moving about in a typical ant trail you might notice a peculiar thing. They don’t seem to be bulk migrating in either direction with any particular order or purpose. Rather than trying to avoid each other, they instead run smack into each other. After butting heads to exchange chemical status bits, they change heading, and repeat. From this simple back and forth they manage to tap out remarkably adaptive colonies. In other words, the commute itself is the computation, a dead reckoning where each step and about face is intimately composed via on board molecular Fit-bits.
When you actually look at the brain with time-lapse microscopy this is more-or-less what you see the mitochondria doing. The main difference is that mitochondria have an extra trick up their sleeve that ants do not. This additional ‘F’ behavior is fusion. If you watch a neuron for any length of time, you will invariably see two mitochondria coming together and forming one. Normally this process is constrained by eventual fission back into two new mitochondria. By fine tuning fusion and fission rates, each cell maintains a large centralized mitochondrial syncytium that dispatches and recalls punctate mitochondrial quanta to different parts of the cell as needed.
The spatial extent of this syncytial mitochondrial fluid both controls, and is in turn controlled by, the phase of the cycle cycle. As most neurons are postmitotic, their mitochondria are essentially decoupled from cell state, freeing them up to pursue other things, like driving their axon far out into parts unknown. Integral to this function is the responsibility of each mitochondria both for policing its own health, and that of the host cell. When coupled with the ability to deliberately self-destruct via a process known as mitophagy, and to subvert their host cell via autophagy, fusion/fission provides a competitive mechanism to promote desirable mtDNA, purge oxidatively damaged DNA, and ensure that the required host cell proteins are properly distributed.
Exactly why mitochondria form syncytia is not yet completely understood, but several ideas emerging from the analogous aggregation of slime molds during times of stress (and from other schooling, flocking, or swarming behaviors) have already gotten molecular backing. The kicker here, is that it isn’t just mitochondria that routinely fuse, but whole cells fuse, often quite predictably. For example, bone resorbing osteoclasts, developing muscle cells, placental cells, fly embryos, and worm germ cells all form multinucleated syncytia by one mechanism or another. Furthermore, when foreign stem cells containing heteroplasmic mitochondria (mitochondria with some mtDNA different from those of the host) are introduced into the brain they have the curious habit of fusing with the local population of neurons.
If whole neurons or their parts habitually fuse it should not be lost on anyone that we have a powerful new mechanism to help explain how the observed structure of the brain could arise. Fusion anastomoses, like those so common in the circulatory system, could dissolve borders across pre- and postsynaptic sites to yoke neurons together while de novo creation of synapses in the middle of a bare neurite could isolate them again as needed. Developmental curiosities and anatomical enigmas, like the finely interleaved stria of the basal ganglia where descending pyramidal cells of the internal capsule penetrate, or the illusive dorsal fornix collaterals that span the corpus callosum may no longer seem quite so inexplicable.
Where do mitochondria come from and where are they going?
Genetic sequence analysis has tagged mitochondria as the direct descendants of a class of bacteria known as alphaproteobacteria. In particular, the pathways involved in ATP production in certain subgroups like Rickettsia are fairly similar to those in present day mitochondria(23). Against the backdrop of this broad consensus, it should be noted that there are lots of very different, very diverse, bacteria that hail from any given bacterial subgroup -- and all of them have very different metabolisms from present day mitochondria.
One indication that some uncertainty still permeates this arena is that within certain geophysics communities (and they are often the real innovators here) the idea that the original protomitochondrion emerged from the class of magnetotactic bacteria still floats about relatively unscathed since it was proposed a few decades ago. This curious theory persists despite the fact that modern mitochondria show virtually no hint of having ever had magnetosomes or magnetotactic behavior, save perhaps for the primal and ubiquitous FeS cluster remnants ensconced away in the heart of their enzymes. (24)
Precursors to the magnetite (Fe304) and greigite (Fe3S4) assemblies used to build magnetosomes appear to have originally played critical roles as terminal electron acceptors in various electron transport circuits where oxygen now fulfils that function in the oxidative phosphorylation of mitochondria. Whether catalytic FeS clusters ‘came first’, ie. before our many other critical metal cofactors, or even before the iconic tetrapyrrole cofactors so critical in every metabolism, may seem like a remote ‘origin of life’ type question.
However, I would argue that these very questions are critical in defining the instinctual behaviors and functions of mitochondria which later guided their construction of nervous systems.
The reason for delving into the many idiosyncratic bedside manners of mitochondria above, is that the intracellular and trans-cellular mitochondrial networks provide a way to explain things about the structure of neurons and nervous systems that purely electrical considerations can not.
For example, the elaborate dendritic trees of cerebellar Purkinje cells contain tens of thousands of synaptic inputs all funneling down into the soma via single dendritic shafts. Sterling and Laughlin give a great account of the firing rates, vesicle release probabilities, material investments and placement of different parts of the cerebellar circuit. What they don’t do, is explain the huge information loss -- the blatant squandering energy and resource, that is witnessed if the fractally converging Purkinje tree is operated purely as an electrical machine(). In other words, unless things are really quiet on almost all of the dendrite’s synapses, there will simply be no bandwidth available to convey messages to the soma.
Whatever combinatorial logic it is that neural modelers imagine occurs at the dendritic bifurcation points, it appears that most of it gets throttled. Conversely, if almost all synapses are really quiet, and signals really sparse, then this design would seem to be an inefficient allocation of resource. Complementary arguments also apply at the output end of the cell, the axon. Here, a presumably identical spike signal is redundantly sent to hundreds or thousands of divergent synaptic endpoints where it is mixed in with local information resident at each synapse. By any measure this would be appear to be a strange computer architecture.
If neuroscientists still fancy the idea of brains and neurons performing computations, then where do they occur? Is it all simply ‘compute with chemistry’ using receptors and intracellular signals, or is it something more nebulous still, like ‘compute with physics’, or perhaps even ‘compute with respiration’? Could the purely energetic process of respiration, namely the electron transport chain, have been harnessed by neurons for computation in a way that parallels how electricity, originally used in bulk form for lights, motors, and solenoid switches was later refined into modern electronics? While no one expects to find the orderly transistor arrays of memory chips or logic gates chips inside neurons, one might expect that the hardware solutions brains eventually hit upon should be fairly endemic to their structure, perhaps even recognizable to us at this point in the game.
Although we cannot directly see the hardware of respiration when we look at EM images, it is now possible to infer much of what must actually be there. For example, the folds and pits which concentrate and funnel metabolites between cristae take the precise forms mandated by the many oxidation, translocation, and junctional assembly complexes embedded in their membranes. One might note in this vein that the dimers forming the ATP synthase of Complex IV are offset by 90 degrees into a ‘V’ shape not entirely unlike the pistons in an engine. They bend the local membrane geometry to orient themselves in rows at the bottom of a deep proton well(26). That Nature has full access to fine tune ATPase assembly can be seen in the unique dimerization structures found in other organisms. For example, the helical tubular arrays of cristae in paramecium mitochondria are based on a unique ‘U’ shaped dimer that offests in a precise zig-zag pattern(27).
The placement of these ATPase dimers sets the floorplan for the other three major complexes of respiration. Complex II, which includes the succinate dehydrogenase ensemble, is a critical point of convergence in the respirative logic because it is also one of the legs of the citric acid cycle around which the entire metabolism of the mitochondria is constructed. As with the other complexes, electrons in Complex II tunnel through multiple cofactor centers possessing precisely tuned redox potentials. The Complex II pipeline includes several varieties of iron-sulfur clusters, each with a geometry and stoichiometry ideally suited to function.
Curiously, one of these clusters is analogous to the greigite thiocubane Fe3S4 unit that is preferentially incorporated into magnetosomes by magnetotactic bacteria when oxygen is scarce. It is now widely held that life as we know it took root when these clusters were further metalized with nickel or molybdenum on the surfaces of certain minerals(). Electrons evolved at Complex II eventually drain to Complex III where they are gated together with those from the Complex I electron circuit. Complex III then lowers the electrons the rest of the way down to oxygen, while raising up additional protons to power ATP synthesis at Complex IV.
Laughlin and other ATP neuroaccountants tell us that both spikes and synapses are expensive in terms of energy requirements. Beyond that there is little accounting for how a given cell should apportion the two. One thing electrophysiologists seem to know with some confidence is that the probability of release of least one vesicle after a spike is received at a release zone is roughly ½, on average, throughout the brain. Does this probability merely reflect some kind of natural balance struck between sending spikes and sending vesicles for each cell? In other words, is the energy spent in structurally maintaining axons and pumping ions back out after sending spikes roughly proportional to the total energy spent in provisioning vesicles and resynthesizing transmitter at the sum total of all of an axon’s synapses?
With these details in tow we can now offer a better explanation for the elaborate branching structures of neurons, then the simplistic principles ‘electrotonic length’. While not every synapse has the prototypical bulbous spine head poised atop a restrictive spine neck, this highly recognizable structure is common throughout the brain. While fabulous things like rapid mechanical twitches and shape changes have been attributed to spines by luminaries no less than Francis Crick himself, no one has really explained their curious form. Although not every spine has their own resident mitochondria, we might admit that ‘active’ synapses energetically asserting their influence on their local environment would probably retain a captive power source in the form of mitochondria.
What I propose, is that the swollen endbulbs of postsynaptic spines found in the dendrites of Purkinje, pyramidal, and numerous other kinds of neurons serve as transient banks of incubators, up to 10,000 strong, for mitochondria.
Incubators for what?
In this conception, the blatant informational bottlenecks inherent in the presumed Cajal-esque signal flow from dendrites to axons are transformed into physical bottlenecks with a new purpose: namely, the selection and transmission of desireable mitochondria to the axon, the larger nervous system, the body itself, and perhaps beyond.
I first suggested this idea of a dendritic proving ground where mitochondria must meet certain criteria for selection a few years ago(28). Perhaps more convincingly for the reader, evidence for this type process has just been found, along with several of the molecular mechanisms that underlie it. The authors found that a quality control system dependent on Parkin ensures the passage of healthy mitochondria to the axon in order to limit impact of sub-par mitochondrial stock(29).
The real power of this simple idea is that when combined with the known penchant of neurons for sharing mitochondria, it also provides an explanation for the polarity of neurons themselves.
Rather than passively arising as by-products of electrotonically-constrained neural architecture, I would offer that these structural mitochondrial bottlenecks actually drove the transformation of the initial unpolarized neuron-like cells found in the primitive neural nets of jellyfish and hydra into the highly polarized neurons and circuits we now find in the higher mammals.
Fortunately we do not have to take this idea on faith; The evolutionary history of this rectification process is on full display in the progression from the ganglionic nervous system architectures of jellyfish up through all the other intermediary invertebrate creatures to us. This taxonomy chronicles the inexorable refinement of primitive multipurpose symmetric synapses reciprocally bombarding each other with dense-core buckets (as seen on EM) filled with nonspecific peptide soup, into the highly asymmetric synaptic diodes with small clear vesicles and discriminating transmitter profiles that we now use. These amenities are also paralleled by subtle transformations of the cytoskeleton and primitive myelinating investments to increasingly chiral forms exclusively reserved for the axon as one goes up the phyla(30).
Similarly, the inhibitory functions of the original dual purpose inhibitory-excitatory neurons were refined and off-loaded whole into small, specialized locally projecting interneurons. Traces of the primitive, more symmetric ganglia-style insect brains are all but gone in us, save perhaps for the unusual pseudounipolar dorsal root ganglion cells that still relay sensation up through our spines. In the twisted neuro-jargon of the terms ‘antidromic’ and ‘orthodromic’ the distinction in signal direction relative to the cell body seems to make little sense for this kind of ‘axon only’ cell.
Are mitochondrial bottlenecks really a thing?
We might presume that the ultimate fate of mitochondria that manage to gain access to axons would depend on what kind of circuit they are in. Some circuits may act as selective filters and transmitters of mitochondria while others would be stylized more as bingers and purgers. Evidence for the later comes from the work of Mark Ellison who uncovered a quality control circuit in the visual system where mitochondria passing into the retinal ganglion cell axons are degraded at a place called the optic nerve head, and then packaged off to specialized glial cells. Here they are auto-phagocytically absorbed in a lysosomal fusion process akin to resorption and turnover of spent photoreceptor outer segments by retinal pigment epithelial cells. A strikingly similar mechanism is also responsible for elimination of paternal sperm mitochondria at the moment the egg is fertilized. This axonal ‘transexudation’ as Ellisworth called it, is opposite and complementary to the above mentioned mitochondrial rescue circuit where astrocytes transmit mitochondria to ailing neurons after stroke.
Intriguingly, the axon is not the only place where selective mitochondrial bottlenecks are found. A series of so-called ‘maternal bottlenecks’ of this very sort can be found in the ovaries. The presumed function here is to ensure that only the purest, most desireable mitochondria get selected and transmitted by the nurse cells to charge the egg -- and therefore the next generation. In the germ cells of the hermaphroditic worm mentioned above this kind of mitochondrial scrimmage is taken to an extreme; the cells fuse into an open borders syncytium where they can duke it out unfettered by membranous barriers. There is now at least one company that offers a work-around to any woman for which the normal mitochondrial security measures have failed. One form of the treatment circumvents the now very dubious prospect of a ‘three-parent embryo’ by in vitro substitution of the misbegotten mitochondria of her eggs with alternative selections sourced from her own nurse cells. A second developmental bottleneck is found a bit later during the the early embryonic phase where the mitochondrial pool is serially diluted. Here, the replication of whole mitochondria, along with the very loosely synchronized replication the mtDNA nucleoids inside them, is completely arrested until after the 6-16 cell stage (depending on the species).
The amalgamation of polarized neural components into larger bottlenecking networks may additionally provide a mechanism to explain even more esoteric phenomena only now beginning to enter the realm of the knowable. The transgenerational transmission of acquired characteristics, particularly of habits and memories, has been widely thought impossible. A few old, but fairly well-known experiments, linger at the back of this recently fashionable topic. The protocol involved training certain habits into Planaria and then grinding them up and feeding them to other Planaria.
This sounds bizarre, but these creatures possess a peculiar facility for regeneration through a population of adult stem cells (known as neoblasts) that are dispersed through their body. Planaria have also proven their mettle at repurposing their tissues by making up for the lack of a standard issue lens by coaxing their own mitochondria into gorging on special light-refractive proteins until they are swollen and tightly packed together into an image forming device(31). While it was reported that the Planaria who were fed their ground up forebearers went on to develop similar habits despite having never been appropriately trained, these kinds of extra-genetic transfer are generally difficult to prove.
However, more recent demonstrations of Lamarckian inheritance of ancestral fears in mice(32), or of similarly conditioned olfactory responses in flies(33), are very difficult to ignore. They are also difficult to explain by appealing to traditionally established epigentic mechanisms. For starters, the ‘many-to-one’ paternal sperm bottleneck is severely bandwidth-limited: generally it is not only impervious to mitochondrial transfer to the egg, but the sperm DNA also undergoes a genetic reboot’ where the histone chromatin is replaced with DNA-compressing protamines, and the many slowly accumulated epigenetic marks on the DNA are wiped clean.
I don’t want to step into it too deep here other than to say that exceptions to rules like uniparental inheritance of mitochondria, or the blanking of the epigenetic slate in the maturation of gametes can often make the point of a larger understanding. For example, in bivalves like mussels there is an apparent violation of the axiomatic rule of uniparental inheritance because the sperm mitochondria manage to evade degradation or extrusion in the egg. This seeming contradiction to a well established evolutionary maxim is resolved by the fact that the male mitochondria are only transmitted from fathers to sons. Incredibly, the early embryo somehow distinguishes paternally inherited mitochondria and ships them exclusively to one of the blastomeres known as ‘4D’ -- which in males, goes on to differentiate into the germ cells which later make sperm(34).
Thinking and Breathing
When it comes to handling their own mitochondria, we might expect no less care than that seen in gametes or a blastula to be taken by the nervous system. While conducted at a speed considerably less than spikes, the apportioning of mitochondria is one, if not the, central preoccupation of nervous systems. Rather than electrical resistance of cytoplasm, neural structure might better be parameterized by a diffusion length of oxygen. Nervous systems then literally represent the inevitable terminus of the oxygenation of the biosphere, artisanal geophysical delicacies weathered through respiration of mitochondria.
The elementary logic typically applied to a sensory neuron receiving input bits of information from the world, integrating that information, and passing X number of bits down the line to the next neuron does not alway make good physiologic sense. For example, one can argue that a significant portion of the information in any spike train, particularly when idling away during unstimulated spontaneous activity, simply represents things a neuron is telling itself. As we know, when neurons want to speak to other neurons they generally don’t use spikes, they use vesicles, and the occasional gap junction. In this view, any external information that manages to become directionally superimposed upon a regularly spiking cell would only represent an incidental fraction of the bulk information flow throughout the neuron. Many spikes then, are just the din of the pump.
The primitive homeostatic function of a cell-wide or organelle-wide membrane potential originally served as gradient to power membrane transport, or to enable sensation of things in the local environment. Self-synchronizing spikes in the spatially extended structure of a large neuron may have arisen in part for the thousands of endosymbionts ambulating within to communicate their energetic activities to each other. Any realistic model of a nervous system would not just include a network geometry with volleys of spikes propagating from input to output -- it would also contain responsive mitochondria meandering about to power the generation of those spikes as sure as any model of a termite mound would contain models of termites.
In Nature’s palette of extremes, the 10-meter long axons of the neurons that innervate the flukes of a blue whale have a serious logistics problem on their hands. Spikes may be able to influence the entire neuron in real time, but there is no obvious way the nucleus can adapt its output -- namely, the proteins and other products it normally makes to care for this cell -- in real time. The slow phase component of axonal transport consisting of large organelles could take decades to reach their target, while even the fast pool, moving at a rate of several mm/day would still take months. During the development phase, the axons tethered to the growing whale’s tail are being pulled aft at the astounding rate of 3cm per DAY. It would seem improbable that everything the growing neuron needs could be provided for by its own nucleus. If the synapses in the tail can obtain fresh mitochondria and other supplies from a local sources, their instantaneous energetic needs could be quickly met and adapted to those of the entire organism.
At the other extreme of the animal world, are the infinitesimal fairy flies. These creatures are full-fledged invertebrate body plans packed into a footprint smaller than a paramecium. They achieve this compacted form by jettisoning huge contingents of genetic and energetic machinery during development. By offloading the nuclei and mitochondria of a large percentage of their neurons, reminiscent of what our own red blood cells do, they lighten their nervous systems down to the bare minimum. In this deprecated state they manage to run on fumes through a short inexorable while, but it is only through extreme metabolic adjustments. Although the developmental of all complex nervous systems (including fairy flies), requires the respiratory services of the ‘powerhouses of the cell’, paradoxically, it now appears that these services are not the primary essential function of mitochondria in all eukaryotes.
To discover that mitochondrial BIOS we need to look afield to those rare single cells that have shredded their own mitochondria completely out of existence. The keys to understanding the origins of all metabolisms are the metal cofactors around which they later accreted. The molybdenum ‘Moco’ clusters, iron-sulfur clusters and hemes, iodine thyroxins, cobalt cobalamins, and many others, have all gone on to spawn entire industries favored to different extents by different niches of life. Invariably, mitochondria play a significant role in constructing each and every one of the these cofactors before handing them off to the cytoplasm for final assembly and insertion into enzyme cores. By toggling the labile mitochondrial and cytoplasmic localization motifs found at start sequences of each enzyme’s genetic instructions, cells can switch the synthesis sites between the two compartments with seeming ease.
The one cofactor which has proven fairly resistant to this kind of shuffling is our FeS clusters. Only a single protist family, the monocercomonoids, has been found that manages to cobble together the essential iron-sulfur gemstones without mitochondria. They have achieved this by appropriating cytoplasmic versions of mitochondrial synthesis enzymes through horizontal gene transfer from a combination of methanoarchael sulfur mobilization systems and bacterial nitrogen fixation systems(35).
The structural forms mitochondria take are a compromise between the need for respiration and the need to provide the countless other eclectic synthesis functions for each organ system in the body. Their skill as nearly universal synthesizers comes courtesy of their powerful oxidizing provisions and protections. For example, the unique tubular phase of the cristae found in liver mitochondria are specialized to pull off the exacting enzymatic flowchart required for steroidogenesis. On the other hand, the discoid cristae of sperm mitochondria are fused to form a large ringed syncytium optimized for ATP production.
D’Arcy Wentworth Thompson’s quip that “the form of an object is a diagram of its forces,” may be particularly relevant to cristae. There aren’t many models available that attempt to depict the structure of cristae, and even less available that try to capture their function. One thing we do know is that that are not anything like the asymmetric baffle structures depicted in most popular images. There are modeling packages that can simulate parts of the metabolisms of cells or organelles, but they are generally nowhere near the complexity of the compartmental electric models that are available for simulating networks of neurons.
At the height of the neuron simulation wars in the mid-to-late nineties, there were two basic modeling packages available -- Genesis and Neuron. Judging by its later selection to be the foundation of the big brain projects like Eurobrain, Neuron was the winner. However, the Genesis program had a couple cool things going for it. For one, it was Unix-based, and therefore inspired many young researchers to tackle the difficult installation of the newly available Linux operating system on their home computers. The other thing was that it had a plug-in module for adding new dendritic growth components that responded to more than just the old Hodgkin Huxley electrical dynamics. For example, there were built-in equations and solvers for diffusion of various key ion like calcium or other molecules.
These kinds of compartmental models have already been of practical use in improving what is probably the most successful brain implant to date -- the cochlear implant. By measuring the material properties of the cochlea, and the finding field potential distribution secondary to stimulation using complex boundary and finite element electromagnetic simulation software, the response of the auditory nerve to an implant can be predicted, at least in theory. The main difficulty with all things biological is that in engineering-speak, they tend to be ill-conditioned. That means that when you try to mesh their structure and replace it with code, long thin axons and thin wispy membranes of tissue are difficult to approximate. Before any multiphysics models of the nervous system at large can be developed to constrain the operation of implants, an accurate CAD model that won’t blow up at the edge will be needed.
Concluding remarks
To try and wrap this all up, I want to make the prediction that our current course of trying to stimulate and record neurons from the grey matter, won’t be the way forward to building practical implants for the masses. From a marketing point of view, companies have already found that there is no real market for advanced implants if the only people to use them are a small pool of paralyzed or locked-in patients. As long as we continue to simply jam 100-spot pincushion arrays onto the cortex, only to irreversibly scar everything in sight a short time later, we will not arrive at a solution for everyman any time soon.
A better approach may be to develop smarter ways to access the brain, namely by putting the hardware into the ventricles and vascular system. Instrumenting tiny stents with recording electrodes -- building ‘stentrodes’ -- has already been shown to be one way to noninvasively record signals from deep inside the brain(36). Implants that noninvasively measure intracranial pressure inside the ventricles have long been sought for treating hydrocephalus. Electrode or optode arrays placed in these same interior spaces, and accurately positioned with external magnetic fields, might initially be quite similar in form and function to the implant leads now threaded through the fluid scala tympani chamber of the the cochlea.
Unfortunately, no one has figured out how to get into the ventricles without plowing straight through precious real estate. However, there might be a way to minimize collateral damage if the ventricles can be assessed via the natural vents (ie., the foramen of Lushka and Magendie) that peek out of the fourth ventricle below the cerebellum. In order to pull this off, it will be necessary to demonstrate that the fragile ciliated ependymal cells that line the ventricles and control its flow will not be disrupted. Considering that these cells are the one single part of the original primitive brain cobbled together by cells lining the feeding pores of our spongelike ancestors that have remained largely unchanged to this day, we might not be too surprised if perturbing them has some consequences that can be directly felt.
On the other hand, if it works, coursing right above and below the ependymal cells lie vast tracts of myelinated axons as far as electrode can see. Notably, 300 million corpus callosum fibers roaring right overhead might can be hit from below and also bridged topside by hardware placed into the well of the medial longitudinal fissure. Stimulating and recording from the white matter locations would have several distinct advantages over grey matter depending on whether electrical, magnetic, optical, ultrasonic thermal or direct mechanical stimulation is used. Furthermore, the genetically enhanced counterparts to each of these techniques, eg. ‘optogenetic’, ‘magnetogenetic’, etc. provide further flexibility. These critical advantages are safety, reversibility, and access. ‘Safe’ because overstimulation of an axon or an axon collateral is much more survivable for the cell than overstimulation of the soma. ‘Reversible’ because vascular or intraventricular hardware is isolated from the neuron (it won’t scar it, and heat unavoidably generated by implants can be readily exhausted to the bulk CSF), and it can be removed in much the same way that it was put in. And finally, ‘access’, because in the brain location is everything().
With each axon potentially accessible at multiple points along the ventricle one could in theory get many reads on any given cell. That would be advantageous in trying to determine the direction and connectivity of the projection you are targeting. It would also eliminate, or at least mitigate, one of the biggest headaches of any modern multiunit recording, the spike-sorting algorithm. Much of the lateral ventricles are lined by sensory projections coursing between the thalamus and cortex. For the large visual tracts in particular, there are around ten return projections back to the thalamus for each one going out to the cortex. No one has figured out what they really do. This is where you want to place those speech or image implants that you desire to be active at the level of the inner voice or the mind’s eye().
1 J. Hewitt Do glial connectomes and activity maps make any sense? http://medicalxpress.com/news/2013-09-glial-connectomes.html
2 J. Hewitt Rise of the Cyborgs http://www.extremetech.com/extreme/144579-rise-of-the-cyborgs
3 J. Hewitt The Thermodynamics of Thought: Soliton Spikes and Heimburg-Jackson pulses http://medicalxpress.com/news/2013-09-thermodynamics-thought-soliton-spikes-heimburg-jackson.html
4 Luca Turin Smells, Spanners, and Switches Inference http://inference-review.com/article/smells-spanners-and-switches
5 J. Hewitt Using the 'deuterium switch' to understand how receptors work J. Hewitt
6 J. Hewitt
How to Myelinate a nerve http://medicalxpress.com/news/2015-03-myelinate-nerve.html
Driving Myelination by Actin Disassembly http://phys.org/news/2015-07-myelination-actin-disassembly.html
The Glial Menagerie: From Simple Beginnings to Staggering Complexity http://medicalxpress.com/news/2013-11-glial-menagerie-simple-staggering-complexity.html
7 J. Hewitt Mapping the Entire Brain with New and IMproved Brainbow II Technology http://medicalxpress.com/news/2013-11-entire-brain-brainbow-ii-technology.html
8 J. Hewitt Physical Principles for Scalable Recording http://medicalxpress.com/news/2013-07-physical-principles-scalable-neural.html
9 Mo Costandi Connectome? Not So Fast https://neurophilosophy.wordpress.com/2009/08/11/connectome_not_so_fast/
10 Transcellular degradation of axonal mitochondria PNAS http://www.pnas.org/content/111/26/9633.abstract
J. Hewitt Fast spiking Neurons take mitochondria for a ride https://medicalxpress.com/news/2014-01-fast-spiking-axons-mitochondria.html
J. Hewitt Fast spiking Neurons take mitochondria for a ride https://medicalxpress.com/news/2014-01-fast-spiking-axons-mitochondria.html
11 J. Hewitt
Horizontal transfer of Mitochondria in Sickness and in Health http://medicalxpress.com/news/2015-08-horizontal-mitochondria-sickness-health.html
Mitochondria control Oncogenesis Through Metabolic Reprogramming http://medicalxpress.com/news/2015-07-mitochondria-oncogenesis-metabolic-reprogramming.html
12 J. Hewitt Axons growing out of Dendrites? Neuroscientists Hate When That Happens http://medicalxpress.com/news/2014-09-axons-dendrites-neuroscientists.html
13 Mo Costandi A New Way of Thinking About How the Brain Works https://www.theguardian.com/science/neurophilosophy/2013/aug/09/a-new-way-of-thinking-about-how-the-brain-works
14 J. Hewitt Pulse Propagation and Signal Conduction in the Hydraulic Brain http://medicalxpress.com/news/2013-09-pulse-propagation-transduction-hydraulic-brain.html
15 Connective tissue diseases: Mitochondria drive NETosis and inflammation in SLE http://www.nature.com/nrrheum/journal/v12/n4/full/nrrheum.2016.24.html
18 Parkinson’s Disease-Related Proteins PINK1 and Parkin Repress Mitochondrial Antigen Presentation
http://www.cell.com/cell/pdfExtended/S0092-8674(16)30590-6
19 Protons Trigger Mitochondrial Flashes http://www.cell.com/action/showAbstract?pii=S0006-3495%2816%2930459-3&doi=10.1016%2Fj.bpj.2016.05.052
20 J. Hewitt Fast contractions and depolarizations in mitochondria revealed with multiparametric imaging
http://medicalxpress.com/news/2014-05-fast-depolarizations-mitochondria-revealed-multiparametric.html
http://medicalxpress.com/news/2014-05-fast-depolarizations-mitochondria-revealed-multiparametric.html
21 J. Hewitt When spikes collide: Shaking the foundation of neuroscience
http://medicalxpress.com/news/2014-09-spikes-collide-foundation-neuroscience.html
http://medicalxpress.com/news/2014-09-spikes-collide-foundation-neuroscience.html
22 J. Hewitt Origin of the Eukaryotic cell: Part I - How to train your endosymbiont
http://phys.org/news/2014-12-eukaryotic-cell-endosymbiont.html
http://phys.org/news/2014-12-eukaryotic-cell-endosymbiont.html
23 J. Hewitt Review of NickLane’s The vital question: Why is life the way it is?
http://phys.org/news/2015-04-vital-life.html#ajTabs<br />
24 Joe Kirschvink http://www.diamond.ysn.ru/IGC2012/IGC2012/PDF/AUTHOR/GC125501.PDF
25 J. Hewitt How does the cerebellum work? http://medicalxpress.com/news/2014-07-cerebellum.html
26 J. Hewitt Controlling the internal structure of mitochondria http://phys.org/news/2015-05-internal-mitochondria.html?utm_source=nwletter&utm_medium=email&utm_content=splt-item&utm_campaign=daily-nwletter
27 Helical arrays of U-shaped ATP synthase dimers form tubular cristae in ciliate mitochondria
http://www.pnas.org/content/113/30/8442.long
http://www.pnas.org/content/113/30/8442.long
28 J. Hewitt Fast spiking axons take mitochondria for a ride http://medicalxpress.com/news/2014-01-fast-spiking-axons-mitochondria.html
29 Compartmentalized Regulation of Parkin-Mediated Mitochondrial Quality Control in the Drosophila Nervous System In Vivo J. Neuroscience http://www.jneurosci.org/content/36/28/7375.short
30 J. Hewitt The Origins of Polarized Nervous Systems http://phys.org/news/2015-03-polarized-nervous.html
31 J. Hewitt Fiber optic light pipes in the retina do much more than simple image transfer http://phys.org/news/2014-07-fiber-optic-pipes-retina-simple.html
32 J. Hewitt Scientists prove that fears and memories can be inherited via sperm
http://www.extremetech.com/extreme/171990-scientists-prove-that-fears-and-memories-can-be-inherited-via-sperm
http://www.extremetech.com/extreme/171990-scientists-prove-that-fears-and-memories-can-be-inherited-via-sperm
33 J. Hewitt Fly dreams and the boundaries of evolutionary science
http://phys.org/news/2014-01-boundaries-evolutionary-science.html
http://phys.org/news/2014-01-boundaries-evolutionary-science.html
34 J. Hewitt Mitochondrial DNA mutations: The good, the bad, and the ugly http://medicalxpress.com/news/2015-01-mitochondrial-dna-mutations-good-bad.html
35 A Eukaryote without a Mitochondrial Organelle http://www.cell.com/current-biology/fulltext/S0960-9822(16)30263-9
36 Chronic impedance spectroscopy of an endoascular stent-electrode array http://iopscience.iop.org/article/10.1088/1741-2560/13/4/046020/meta;jsessionid=B5FC3E6562BC9F0A3F9AEEFB5492A312.c2.iopscience.cld.iop.org
No comments:
Post a Comment