#Neuroscience #NIH #YourTaxDollarsAtWork #BrainComputerInterface #WhyBCIsFail #SciencePolicy

🚡 We know that oligodendrocytes matter for BCI performance. We don’t yet know whether restoring oligodendrocyte health after injury is possible, or which intervention window is most critical. We showed that Clemastine helps in mice when started before implantation. We don’t know whether it helps when started after, or by how much, or whether the mechanism holds in human tissue.

🪫 We know that metabolic stress silences neurons near chronically implanted electrodes. We don’t know the full recovery potential, whether those silenced neurons can be re-engaged with the right intervention, and for how long. We suspect lysosomal dysfunction plays a role in the transition from metabolic suppression to irreversible degeneration. We don’t yet know where that threshold is.

📔 We know that temporal patterning of electrical stimulation can inhibit neural networks rather than simply exciting them. We don’t yet know how to harness that inhibition clinically, which conditions would benefit most, what the optimal parameters are, or how long the effect can be sustained before the network adapts.

🩸 We know that focal brain injury can accelerate Alzheimer’s-like pathology in genetically vulnerable rodents. We don’t know the clinical significance of that finding for human BCI participants, or how to identify which participants carry that vulnerability, or whether it changes the risk-benefit calculation for enrollment.

🔬 These are the questions that drive my lab’s work. They are also questions where the answers would matter not just for the relatively small number of people currently using BCIs experimentally, but for understanding neurodegeneration more broadly. The brain’s response to an implanted electrode turns out to be a remarkably useful model system for studying how the brain responds to injury and disease in general.

🧠 That is why I think this research deserves more attention than it gets. Not because of what it says about devices, but because of what it says about the brain.

Thanks for reading. I’m happy to answer questions, engage with critiques, and hear from anyone working on related problems, whether in research, engineering, clinical practice, or lived experience.

The work continues. https://www.bioniclab.org/

#Neuroscience #Neurotechnology #BrainComputerInterface #BasicScience #OpenScience #WhyBCIsFail

His answer: “It has nothing to do directly with defending our country except to help make it worth defending.”

🧬 I think about that line often when people ask me why I study glial cells and blood vessels instead of building brain-computer interfaces.

🦠 The answer is the same. The basic biology is what makes the devices worth building.

📅 Over the course of this series, I’ve walked through what actually happens when you put an electrode in the brain. Insertion trauma that severs capillaries and tears through the neuropil. An immune response that never resolves. Glial cells that wall off the device. A neurovascular unit that reorganizes around a wound it cannot close. Neurons that die, migrate, or simply stop talking to the electrode.

⚙️ None of these problems can be engineered around. They must be understood first.

🔬 That understanding comes from basic science. The kind funded by NIH, pursued in academic labs, published in journals most people will never read. It is slow work. It does not generate headlines or Series B valuations. It generates knowledge, the kind that determines whether a neural interface lasts five weeks or fifty years.

🦅 Wilson was asked if his accelerator would keep the country competitive. He said it had to do with whether we were good painters, good sculptors, great poets. “All the things that we really venerate and honor in our country and are patriotic about.”

🚸 Then he was asked about priorities and hungry children.

🚚 He agreed that feeding children comes first. But he added that it is also important to get on with the things that make life worth living. And he warned that if we falter, “I can see the whole effort dispersed and lost.”

🛣️ That warning applies to neural engineering today. The trainees who would characterize chronic glial dynamics or map neurovascular remodeling at implant sites need funded labs to work in. If those grants don’t come, those people retrain. The knowledge they would have generated simply will not exist. And every BCI company building on a biological foundation they assume someone else will maintain will eventually hit a wall that no amount of engineering capital can move.

🧪 The question was never whether basic science matters. The question is whether we will still be funding it when the runway ends.

⚠️ Fund the science that explains why brain-computer interfaces fail. That is how you build ones that don’t.

Review: https://www.nature.com/articles/s41551-017-0154-1Review: https://advanced.onlinelibrary.wiley.com/doi/abs/10.1002/adfm.201701269
Review: https://www.nature.com/articles/s41551-025-01583-6
Review: https://pubs.acs.org/doi/full/10.1021/cn500256e
Perspective: https://iopscience.iop.org/article/10.1088/1741-2552/ae54cf

#neurotech #BCI #NIH #basicscience #neuroscience #braincomputerinterface #WhyBCIsFail

The answer is not simply a better electrode. It is a different relationship between the device and the biology around it.

🏗️ A durable BCI would need to be implanted without disrupting the vascular architecture that supplies oxygen and nutrients to the neurons it’s recording from. It would need to be small enough to avoid triggering a substantial foreign body response, or coated with materials that redirect rather than suppress that response. It would need to exist in a tissue environment where oligodendrocytes remain healthy enough to support sustained neural activity, and where lysosomes remain functional enough to clear cellular waste. It would need to deliver stimulation in patterns that engage the brain’s natural circuit dynamics rather than overriding them in ways that accelerate fatigue.

📋 None of these requirements are individually beyond reach. Several are actively being addressed in research labs around the world. The challenge is that they have mostly been addressed in isolation, better materials science here, better surgical technique there, better stimulation parameters somewhere else, without a unifying framework that treats the brain as the sophisticated biological system it is.

🔬 My lab’s contribution to that framework is the recognition that non-neuronal cells are not obstacles to BCI function, they are participants in it. Oligodendrocytes, microglia, astrocytes, and pericytes each play roles in determining whether the neural tissue around an electrode remains healthy, active, and recordable over time. Designing devices and interventions that work with those cell types rather than against them is, I believe, the most underinvested area in the entire field.

🔎 ⚙️ 🩺 This is not a criticism of the engineers and clinicians pushing the technology forward. It is an observation about where the scientific gaps are, and where I think the next decade of research needs to focus. The devices are getting better. The biology hasn’t changed.

Engineering (Adv Func Mat): https://lnkd.in/eqtrZAcB
Biology (Nature BME): https://lnkd.in/ebB5H4K
Stimulation (Nature BME): https://lnkd.in/enHUqypN
Materials (Nature Materials): https://lnkd.in/epWswUdi
Modulation (Nature Communication): https://lnkd.in/g9Zz-TNi
Lab: https://www.bioniclab.org/

#Neurotechnology #BrainComputerInterface #Bioengineering #NeuralEngineering #Innovation

🔬 Here is the trap. Biocompatibility sounds like a property a material has, the way gold is shiny or steel is strong. It is not. The accepted definition, set by David Williams in 1987 and still used by the FDA, says a material is biocompatible only when it does its job in one specific place in the body without the body shutting it down. Move the same material somewhere else and the answer can flip. So titanium is not biocompatible. Titanium works beautifully fused into bone, and can fail in other tissue. The material did not change. The location did.

⚠️ This is not word games. If you believe biocompatibility lives in the material, you test it once, early, with a standard safety panel that checks one thing, does the material poison nearby cells. A brain electrode passes that panel easily. Nothing toxic leaks out. On paper it is safe. Then you put it in cortex.

🧩 The brain does not ask whether the electrode is toxic. It asks whether the electrode belongs. It decides no. Immune cells crawl to the electrode and wall it off the way your body walls off a splinter, scar tissue builds, and the neurons the electrode was supposed to listen to either die back or drift out of range. Over weeks to months the signal fades. The metal is still sitting there, perfectly safe, recording almost nothing.

📊 Both things are true at once. The material is safe. The device failed. That only sounds like a contradiction if you think biocompatibility is a sticker you put on a material. It is not a sticker. It is a verdict the body issues about a specific device, in a specific place, over time, and the body keeps voting for as long as the implant is in.

That is why "is it biocompatible" is the wrong question. The right question is "does it still work in this tissue two years from now," and almost no standard test asks that, because we underfund the biology that would answer it.

More in the Why BCIs Fail series: https://bioniclab.substack.com/s/why-bcis-fail

#WhyBCIsFail #BrainComputerInterface #Neuroscience #NeuralEngineering #Biomaterials Biocompatibility

✋ BCI research would not exist without participants. The findings that make headlines, the demonstrations that inspire funding, the data that drives the next generation of devices, all of it depends on people who agree to have electrodes implanted in their brains for the benefit of science and for the hope of improving their own lives and the lives of others with similar conditions. That is an extraordinary act. It deserves to be named clearly and often.

🔬 I study what happens to brain tissue around implanted electrodes. My work is largely in animal models, but it is motivated entirely by what happens in humans. When I find that neurons near an electrode can be silenced by metabolic stress rather than cellular death, I think about what that means for someone enrolled in a trial whose device has stopped performing as well as it once did. When I find that oligodendrocyte health affects recording stability, I think about what might be done differently in future trials to protect those cells... and I’m aware that finding could only have been made in a living brain, where myelin, vasculature, immune cells, and electrophysiology interact over months in ways no cell culture or organoid can yet reproduce. When I find that focal brain injury can accelerate Alzheimer’s-like pathology in genetically vulnerable animals, I think about the tension between the obligation to share what science finds and the harm that incomplete knowledge, communicated carelessly, can do to people making irreversible decisions.

⚠️ These are not abstract concerns. They are obligations.

📣 I also want to say something about the asymmetry in this relationship. Participants give researchers something irreplaceable: access to the human brain under carefully controlled conditions that cannot be replicated in any animal model. Researchers give participants something more modest: the best science they can do, and the commitment to share what they learn honestly, including the uncomfortable findings.

🏥 If you or someone you know is considering enrollment in a BCI trial, ask questions. Ask about what happens if the device stops working. Ask about long-term monitoring. Ask about what the research team knows about failure modes. A good research team, like at UPMC, will welcome those questions. They should be able to answer them.

🙏 And if you are already enrolled, or have completed a trial: thank you. The field owes you more than it typically acknowledges.

https://www.rnel.pitt.edu/research/neuroprosthetics/intracortical-brain-computer-interfaces

#PatientAdvocacy #ClinicalTrials #Neuroethics #Disability #BrainComputerInterface

🔬 Our paper in Nature Biomedical Engineering built the first unified mechanistic framework for how intracortical microstimulation creates artificial sensation, tracing the full chain from pulse to percept across species.

🧠 Several findings from this series converge here. The inhibitory recruitment described in Post 7, where stimulation at certain frequencies drives inhibitory networks harder than excitatory ones, appears in primate and human data too. The metabolic constraints from Post 5, where neurons near the electrode lack energy to sustain activity, shape how long a sensation can last. The neurovascular disruption from Posts 13 and 14 determines whether tissue can metabolically support repeated stimulation over months.

📊 The biology is conserved across species. Inhibitory network recruitment, metabolic fatigue during sustained stimulation, frequency-dependent microglial responses (Post 15), all appear in primate and human recordings. The mouse wasn’t lying. The fundamental biology of how brain tissue responds to electrical stimulation is consistent from rodent cortex to human cortex.

🤔 This matters practically. BCI users report that stimulation-evoked sensations fade during sustained use and vary unpredictably across sessions. The framework explains why. It’s not just neural adaptation. It’s metabolic depletion, vascular compromise, and immune remodeling operating on different timescales around the same electrode. Each mechanism described in this series contributes to the clinical experience of sensation that works, then fades, then sometimes returns.

🔧 The engineering and the biology aren’t separate disciplines working on the same device. They’re the same problem viewed from different sides. Designing stimulation that creates stable, lasting sensation requires solving them together.

ICMS mechanisms (Nat BME): https://www.nature.com/articles/s41551-025-01583-6
Inception framework: https://iopscience.iop.org/article/10.1088/1741-2552/ae54cf

#Neuroscience #Neurotechnology #BrainComputerInterface #BrainStimulation #Bioengineering

🔬 Using finite element analysis, we simulated how brain tissue strains and deforms around implanted electrode arrays during normal daily life. Not mouse electrodes. The same Utah arrays currently used in human clinical trials at Pittsburgh, Caltech, and Stanford. The brain moves constantly, every heartbeat, every breath, every postural shift creates micromotion (Post 2 on the stiffness problem). A rigid array anchored to the skull doesn’t move with it. The tissue stretches, compresses, and shears against the device surface, over and over, thousands of times per day.

📊 Our model predicted which channels would experience the most mechanical strain based on their position in the array, their depth in cortex, and the tissue properties surrounding them. Then we compared those predictions against actual recording performance data from human BCI participants and non-human primates using the same arrays.

🎯 The predictions matched. High-strain channels showed the worst recording performance in living humans. Low-strain channels maintained better signal quality over time. The mechanical failure mode described throughout this series, chronic micromotion disrupting the biological environment around the electrode, maps directly onto real clinical BCI outcomes. The mouse data weren’t misleading. The physics and biology are the same.

🔗 This connects the tissue-level biology from earlier posts to clinical results. Pericyte constriction (Post 13), network fragmentation (Post 6), metabolic silencing (Post 5), all of these processes are driven or worsened by chronic mechanical strain. The model shows that strain isn’t uniform across an array. It varies by position, which means the biological damage varies by position too. That’s why some channels fail in the first weeks while others last years in the same participant.

🔧 If we can predict which channels will experience the most strain before implantation, we can optimize array placement, weight decoder algorithms toward reliable channels, or design arrays that distribute strain more evenly. The biology of failure has a geometry, and now we can map it.

Mechanical modeling: https://iopscience.iop.org/article/10.1088/1741-2552/ae1bda
Stiffness problem: https://bioniclab.substack.com/p/the-stiffness-problem

#Neuroscience #Neurotechnology #BrainComputerInterface #TranslationalResearch #Bioengineering

❓ So we asked a different question: what if we could redirect the brain’s immune response using nothing but sound?

💆 Low-intensity pulsed ultrasound, or LIPUS, delivers mechanical energy through acoustic waves at intensities far below those used in therapeutic ultrasound. At these low intensities, LIPUS doesn’t damage tissue. What it does is alter cell behavior with a gentle massage, and in the context of brain implants, what it alters is microglia.

🛠️ Following electrode implantation, microglia migrate to the injury site and eventually encapsulate the electrode in a dense cellular sheath that impairs recording. Our question was whether LIPUS could shift microglia from encapsulation toward repair.

The answer, published in Nature Communications in 2024, was yes. In rats receiving LIPUS following implantation, microglial activation was substantially reduced, glial scarring decreased, and chronic recording performance improved measurably. LIPUS accelerated the transition of microglia from a reactive, encapsulating state toward a surveillant, tissue-remodeling state, nudging the brain’s immune response toward healing rather than walling off.

This connects to something from an earlier post: the frequency and pattern of electrical stimulation also shapes microglial behavior. Our 2026 work showed that 10 Hz stimulation causes microglial processes to extend preferentially toward metabolically stressed neurons without triggering classical inflammatory activation. Both LIPUS and appropriately tuned electrical stimulation converge on the same biological target. Two different physical mechanisms, one shared lever.

🩺 What makes LIPUS particularly significant is delivery. It can be applied non-invasively through the skull, administered to a person with an implanted BCI without additional surgery, as a post-implantation treatment or periodic maintenance intervention. This work is now being developed toward clinical application by Actuated Medical. The path from rodent to human is long, but the biology is sound and the delivery approach is clinically tractable in a way that electrode-level interventions often are not.

Ultrasound: https://lnkd.in/g9Zz-TNi
Electrical: https://lnkd.in/e38qd6hN

#Neuroscience #Neurotechnology #Ultrasound #BrainComputerInterface #Bioengineering

🔬 We delivered sustained electrical stimulation through implanted microelectrodes while imaging microglia in living mouse brain. During stimulation, microglial processes extended preferentially toward the neurons that were most active. Not toward the electrode. Toward the activated cells. The immune system was tracking which neurons were working hardest.

🤔 This is fundamentally different from the foreign body response described in Post 3. Scar-forming microglia react to the device as a physical object. These surveillant microglia were responding to the biological consequences of device operation. Two different immune behaviors, triggered by two different signals, operating simultaneously around the same implant.

⚡ The stimulation frequency mattered. At 10 Hz, microglia directed surveillance toward metabolically stressed neurons without triggering classical inflammation. Different frequencies produced different immune responses, a variable no current clinical stimulation protocol accounts for. The same device delivering different pulse patterns could create very different immune environments in surrounding tissue.

🔗 This connects the metabolic story running through the series. Neurons near the electrode are energy-deprived (Post 5). Blood supply is compromised (Post 13). Stimulation itself disrupts neurovascular coupling (Post 14). Now we’re seeing immune cells detect that metabolic stress and respond in real time. The immune, vascular, and metabolic systems aren’t separate problems. They’re one integrated biological response to a device that is both present and active.

🔧 This collapses the boundary between “biocompatibility” and “device function.” Designing stimulation patterns requires considering immune consequences, not just neural ones.

Microglia and stimulation: https://iopscience.iop.org/article/10.1088/1741-2552/ae4652
Immune response review: https://www.nature.com/articles/s41551-017-0154-1

#Neuroscience #Immunology #BrainStimulation #Neurotechnology #BrainComputerInterface

🔬 Using calcium imaging and hemodynamic monitoring in awake mice, we tracked how repeated electrical stimulation changes both neural activity and local blood flow over weeks of chronic stimulation in the visual cortex. This let us watch two things simultaneously: what neurons were doing, and whether the blood supply was keeping up.

📊 The results challenged a core assumption. Chronic stimulation didn’t just activate neurons in the moment. It altered the relationship between neural activity and blood flow over time. The hemodynamic signals, the brain’s way of directing fuel to active regions, shifted in ways that persisted beyond each stimulation session. The coupling between “neurons need energy” and “blood delivers energy” became less reliable with repeated use.

🚨 Think of it this way. Your brain has a thermostat that matches energy delivery to energy demand in real time. When neurons fire more, blood flow increases to feed them. Our data show that repeated stimulation disrupts that matching. The thermostat starts misreading the room. Neurons are working hard, but the fuel delivery system isn’t responding proportionally.

😶 This creates a problem layered on top of everything this series has described. The implant already disrupts blood flow through vessel damage during insertion (Post 1). Pericytes constrict and stay constricted for weeks (Post 13). The foreign body response creates a metabolically hostile environment (Post 3). Now we’re showing that the therapeutic stimulation itself, the very thing the device is designed to do, may further alter the metabolic environment it depends on. The failures compound.

🧩 This also connects to the finding that stimulation can recruit inhibitory networks in unexpected ways (Post 7). If the blood supply can’t keep pace with what stimulation demands, neurons may fatigue not because the stimulation pattern is wrong, but because the metabolic infrastructure can’t support it.

🔧 Designing stimulation that works long-term may require monitoring not just whether neurons respond, but whether the energy supply can sustain what we’re asking those neurons to do. The brain isn’t just a circuit. It’s a circuit with a power bill.

Chronic ICMS hemodynamics: https://www.sciencedirect.com/science/article/pii/S0142961226003005
Energy crisis: https://bioniclab.substack.com/p/the-energy-crisis-nobody-talks-about
Review: https://iopscience.iop.org/article/10.1088/1741-2552/ae54cf/

#Neuroscience #BrainStimulation #Neurotechnology #BrainComputerInterface #Bioengineering

A common pattern across engineering-and-biology conversations in BCI translation runs like this. Investigator demonstrates that mechanism A is sufficient to drive outcome B, with an intervention experiment that perturbs A and watches B change. Critic responds, “you have not shown A is the only cause, B also depends on C, D, E.” Investigator agrees that C, D, E matter. Critic concludes the experiment proves nothing.

⚡ The confusion is between three distinct logical operators. Sufficiency, if A, then B. Necessity, B requires A. Exclusivity, only A produces B. Most intervention experiments demonstrate sufficiency. Almost no one claims exclusivity, because biology rarely runs through a single pathway. But engineers trained on system identification often treat sufficiency claims as exclusivity claims, because in a well-characterized linear system, finding one input-output transfer function is supposed to be the answer.

🔬 Biological systems do not work that way. Multiple independent pathways can produce the same phenotype. Sir Austin Bradford Hill made this explicit in the 1965 Proceedings of the Royal Society of Medicine paper that founded modern causal inference in medicine (link below), arguing that intervention evidence should be weighed against the reality of multiple causes rather than held to a standard of demonstrating sole causation.

📊 The cost of confusing sufficiency with exclusivity is measurable. For chronic intracortical recording, the field has spent over two decades targeting neuron protection (anti-inflammatory coatings, softer materials, smaller electrodes) on the assumption that neuron death drives signal loss. Recent genetic-perturbation work shows that oligodendrocyte metabolic support gates recording quality at chronic-implant inflammatory baseline, independent of neuron loss. Inflammation matters. Neuron death matters. Oligodendrocyte function matters. They are not the same variable, and intervention on any of them can change the recording outcome.

🔧 The translational question is not “what is the unique cause” but “what is a tractable intervention point.” Sufficient cause is not the cause. It is a cause. Demonstrating that A is sufficient says A is an axis you can move to change the output, which is what therapeutic strategy actually requires.

🇺🇸 The decades-long mistargeting of BCI intervention strategy comes not from bad science but from a logical category error that Hill warned against sixty years ago. Sufficient cause is not the cause. It is a cause. That distinction is worth recovering.

Hill 1965 (free): https://www.ncbi.nlm.nih.gov/pmc/articles/PMC1898525/

#WhyBCIsFail #Neuroscience #BrainComputerInterface #CausalInference #NeuralEngineering

🔬 Your brain has its own plumbing system. Capillaries deliver oxygen and fuel to neurons, and wrapped around them are mural cells, pericytes and smooth muscle cells that act as valves, opening and closing to direct blood where it’s needed. Without them, the brain has no way to route fuel to active regions.

🚨 When an electrode is implanted, pericytes near the device constrict, squeezing capillaries shut. Blood flow to nearby neurons drops. Not because vessels were destroyed during insertion, but because the biological valves controlling flow clamped down in response to injury and stayed shut.

📊 We tracked individual pericytes on individual capillaries for weeks after implantation using two-photon microscopy. The mural cells didn’t just react once and recover. They changed structurally over time, shifting their shape, position, and grip on the vessels they regulate. Some that initially constricted never fully relaxed. Others migrated away from their normal positions along the capillary wall. The blood supply restriction wasn’t a one-time injury. It was an evolving problem that worsened even as the initial insertion wound healed.

🧠 This matters because neurons are energy hogs that can’t stockpile fuel. When blood flow drops even modestly, they lose the ability to sustain activity. They don’t die. They go quiet, rationing energy the way any system does when supply is constrained (Post 5 on the energy crisis).

🔗 Connect this to the timeline from earlier posts. The electrode severs vessels on insertion (Post 1). Immune cells swarm within hours (Post 3). Oligodendrocytes that keep neurons metabolically efficient are damaged within days (Post 4). Now we’re showing that the vascular cells responsible for restoring blood flow are themselves compromised for weeks. Each layer of the response compounds the last.

🔧 The device might be fine. The wiring might be fine. But the power grid went down. If we want implants that last decades, we need strategies that protect the brain’s plumbing, not just its circuitry.

Mural cells: https://www.sciencedirect.com/science/article/pii/S0142961224004988
Energy crisis: https://bioniclab.substack.com/p/the-energy-crisis-nobody-talks-about
Review: https://iopscience.iop.org/article/10.1088/1741-2552/ae54cf/

#Neuroscience #Neurotechnology #BrainComputerInterface #Bioengineering #BrainHealth

🔬 We found it in mouse neurons within weeks of electrode implantation.

🔋 The mechanism is metabolic. Lysosomes, the cell’s recycling organelles, digest damaged proteins using a highly acidic internal environment maintained by an ATP-consuming proton pump. When ATP runs short, the pump slows, acidity falls, digestive enzymes stop working, and waste accumulates as lipofuscin. We confirmed this directly: lysosomal enzyme activity and acidification were disrupted in cells near chronically implanted electrodes, tracking with proximity and metabolic stress.

🪫 Around a chronically implanted electrode, the conditions driving that energetic shortage are everywhere. Reduced tissue oxygenation from vascular disruption (Post 5). Pericytes clamping capillaries shut for weeks (Post 13). Demyelinated axons burning ATP faster than it can be replenished (Posts 10, 12). The neurons closest to the electrode are running an energy deficit they cannot dig out of. The lipofuscin is the cellular receipt for that deficit, frozen in place.

⚠️ This is not aging in the chronological sense. The mice are young. But at the cellular level, neurons near the electrode carry the waste burden of a cell decades older than they are. They look alive. They are structurally intact. They are also functionally compromised in ways that appear in recordings as silenced or unreliable units.

🚨 Lysosomal dysfunction is a central feature of Alzheimer’s, Parkinson’s, and ALS. The cellular pathway near an implanted electrode overlaps directly with pathways that drive neurodegeneration. That is why the APP/PS1 mice from Post 8 develop plaques months ahead of schedule when implanted. The electrode didn’t cause Alzheimer’s. It accelerated a degenerative cascade that was already primed, and lysosomal failure is the mechanism connecting metabolic stress to disease progression.

🔧 The intervention target is metabolic. Restore the energy supply, and lysosomes can resume function before waste accumulates past recovery. That is the window we are trying to characterize, the line between a silenced neuron that can come back and one that can’t.

🕰️ The trash piles up quietly. By the time you see it in your recordings, the cell has already aged decades.

Lysosomal dysfunction: https://www.sciencedirect.com/science/article/pii/S0142961223003241

Metabolic stress: https://pubs.acs.org/doi/full/10.1021/cn500256e

Alzheimer’s connection: https://iopscience.iop.org/article/10.1088/1741-2552/aceca5/

Why BCIs Fail Series: https://bioniclab.substack.com

#Neuroscience #Neurodegeneration #BrainHealth #Neurotechnology #BrainComputerInterface #BrainAging

🔬 We engineered mice in which we could selectively alter a single non-neuronal cell type, oligodendrocytes, while leaving every neuron completely untouched. Same electrode. Same surgery. Same brain region. Same neurons. The only variable was the health of the cells that wrap and insulate nerve fibers, the cells most BCI engineers have never considered a target.

📊 The results were unambiguous. Mice with healthier oligodendrocytes maintained sharper single-neuron signals over the chronic recording period. The fidelity of individual units, how cleanly we could isolate one neuron’s voice from the crowd, was better preserved. Network-level dynamics were more stable. The neurons were identical across both groups. The oligodendrocytes made the difference.

💡 This is causal evidence, not correlation. Not “oligodendrocytes happen to be damaged near failing electrodes.” Not “a drug that helps oligodendrocytes also helps recordings” (Post 11). This is: change one cell type genetically, hold everything else constant, and recording quality changes with it. The link runs directly through cells the field has treated as background infrastructure.

🔗 The causal chain connects the last several posts. The energy crisis (Post 5) showed neurons going silent from metabolic stress. Post 10 explained that oligodendrocytes dramatically reduce the energy cost of neural firing. Post 11 showed a drug targeting oligodendrocytes improved recording stability. Now the genetic proof closes the loop: oligodendrocyte health directly determines whether neurons can sustain the firing patterns that carry information for a BCI decoder.

🔧 The field has spent decades engineering better ways to protect neurons from implants. Our data say the neurons were never the weakest link. The support cells were. The target for next-generation BCI longevity isn’t the electrode surface. It’s the cellular neighborhood.

Genetic proof: https://iopscience.iop.org/article/10.1088/1741-2552/ae4d8b
Oligodendrocyte review: https://www.sciencedirect.com/science/article/pii/S0142961218306069

#Neuroscience #Neurotechnology #BrainComputerInterface #GlialCells #Bioengineering

We chose Clemastine, an antihistamine that has been shown in other contexts to promote oligodendrocyte differentiation and myelination. It is FDA-approved, orally administered, crosses the blood-brain barrier, and has a known safety profile. It targets signaling pathways specific to oligodendrocytes through muscarinic receptors. If you wanted to test the oligodendrocyte hypothesis in a living brain, it was a reasonable tool.

📈 The results were striking. In mice receiving Clemastine starting one week before electrode implantation, myelin levels near the electrode were roughly 180% of controls. Chronic recording stability improved by approximately 140%. Critically, inhibitory neuronal activity, the first thing to degrade in our recordings without intervention, was substantially better preserved.

🧫 Demonstrating that a drug targeting oligodendrocytes improves chronic BCI recording requires a living brain, a functioning immune system, intact vasculature, and months of electrophysiology, conditions no current non-animal model (NAMs) can provide.

⚠️ I want to be careful about how I frame this. Clemastine is not a BCI treatment. These are rodent experiments, and translation from rodent to human is never guaranteed. The appropriate response to these findings is not to start taking Clemastine before a BCI surgery. It is to take the oligodendrocyte hypothesis seriously as a target, fund research to understand the mechanism more deeply, and explore whether this or related compounds have a role in improving device longevity in humans.

🔎 But the experiment taught us something beyond the drug effect. The fact that a pharmacological intervention targeting a non-neuronal cell type, not neurons, not the electrode surface, not the biomaterial coating, could meaningfully change recording performance is itself important. It tells us that the failure we observe in chronic BCI recordings is not purely a materials problem or a surgical problem. It is partly a cellular biology problem. And cellular biology problems have biological solutions.

👨‍👩‍👧‍👦 The field has spent enormous resources making better electrodes. We may also need to make better biological neighborhoods for those electrodes to live in.

Clemastine: https://lnkd.in/eyNPriA9
Fus in Oligo: https://lnkd.in/eFErQ2i9

#Neuroscience #DrugDiscovery #Neurotechnology #BrainComputerInterface #Bioengineering

🧩 That explanation is correct. It is also, we believe, incomplete in a way that has misled the BCI field for years.

🔋 If speed were the primary function of myelination, you would expect that depleting myelin would slow neural signals. And it does, but only modestly, and only for the third+ action potential in a sequence. What changes dramatically when myelin is lost is not the speed of individual spikes but the ability of neurons to sustain high firing rates over time.

🪫 We showed this by depleting oligodendrocytes using a well-established pharmacological model. Neurons in the demyelinated cortex showed normal or near-normal latency for single evoked responses. But ask those same neurons to sustain elevated activity for more than 400 milliseconds, and they failed. They fatigued in a way that myelinated neurons did not.

⚡ The mechanism is metabolic. Maintaining myelin is energetically expensive. Evolution does not make that investment without a return. The return is not primarily speed. It is the dramatic reduction in the metabolic cost of sustained firing. By restricting ion exchange to the nodes of Ranvier, myelin reduces the total sodium and potassium flux during an action potential, which reduces the ATP required to restore the ion gradients afterward.

💡 For BCIs, the implication is direct. A device implanted in a region where oligodendrocytes have been injured or depleted is not just recording from neurons in a hostile chemical environment. It is recording from neurons that have lost their metabolic efficiency, neurons that can respond to a brief stimulus but cannot sustain the patterns of activity that make their signals informationally rich. In contrast, reinforcing myelin with genetic manipulation or drugs can help preserve energy efficiency and function.

🖨️ Fixing the recording problem may require fixing the oligodendrocyte problem first.

Review: https://lnkd.in/edpJ2zzy
Oligo: https://lnkd.in/ePXYRr2C
Fatigue: https://lnkd.in/egAwPGv
Network Stability: https://lnkd.in/eFErQ2i9
Drugs: https://lnkd.in/eyNPriA9

#Neuroscience #Myelin #MultipleSclerosis #BrainHealth #NeuralEngineering

🔋 Lysosomes work by maintaining a highly acidic internal environment. That acidity requires constant energy, specifically an ATP-consuming proton pump that continuously pushes hydrogen ions into the lysosome to keep it acidic enough to activate its digestive enzymes (red lysosomes).

Here’s what happens near a chronically implanted electrode.

🪫 The metabolic stress described in previous posts, reduced tissue oxygenation, disrupted blood flow, energetically depleted neurons, compromises the cell’s ability to run that proton pump. When the pump slows, the lysosome becomes less acidic. When it becomes less acidic (yellow and green), the digestive enzymes stop working properly. Cellular waste that should be broken down and recycled instead accumulates as a granular material called lipofuscin, sometimes called the age pigment because it accumulates naturally in very old cells. (This is not something you can study in NAMS, because it requires a fully integrated and self-contained metabolic system.)

🔬 We found lipofuscin accumulation in neurons and glial cells near chronically implanted electrodes in patterns that correlated with metabolic stress, not simply with time or neuronal death. In other words, cells that appeared alive and structurally intact were carrying a waste burden consistent with severe metabolic dysfunction.

🔎 This matters for two reasons. First, lysosomal dysfunction is a central feature of Alzheimer’s disease, Parkinson’s disease, ALS, and several other neurodegenerative conditions. The cellular pathway we’re observing near implanted electrodes overlaps significantly with disease pathways. Second, it suggests a potential intervention target: if we can restore metabolic supply and lysosomal function near the electrode, we may be able to prevent or reverse the cellular deterioration that silences neurons over time.

🚧 The trash piles up quietly. By the time it’s visible in recordings, the problem has been building for weeks.

Lysosome: https://lnkd.in/eEraH5Fv
Lipofuscin: https://lnkd.in/g6vNw8NP
Metabolic: https://lnkd.in/eZ_7ZsTN
Review: https://lnkd.in/e7MrVCc7

#Neuroscience #Neurodegeneration #CellBiology #BrainHealth #Bioengineering #Lysosomes

Thanks to Bin He for organizing.

BCI failure is not primarily a materials problem. Signal loss begins before glial scar maturation, and two shanks on the same Utah array, 400 µm apart, identical in geometry, stiffness, surface chemistry, and insertion velocity, show completely different tissue responses. That variability alone should have ended the materials-first narrative years ago.

The tissue-electrode interface is an actively remodeling microenvironment. Three biological axes drive that remodeling, none addressed by smaller, softer, or more biocompatible materials.

🧠 Axis 1: Neurovascular disruption Every insertion severs capillaries and ruptures the BBB. Pericyte loss follows within days, capillaries constrict from ~5-6 µm to ~3.5 µm, and the resulting metabolic deficit silences neurons within 100 µm of the implant without killing them.

⚡ Axis 2: Glial reactivity beyond the scar The morphological scar extends ~50 µm. Functional changes in excitability and synaptic function extend beyond 300 µm. Your LFP is partly glial. Decoding models trained on week 1 assume a stable neural population that no longer exists by week 4.

🔬 Axis 3: Oligodendrocyte metabolic failure Demyelination begins within days. Clemastine improves chronic recording without reducing glial scar. A FUS conditional knockout increasing oligodendrocyte cholesterol biosynthesis improved recording again, but myelin sheath thickness did not predict it. Cholesterol biosynthesis activity, not wrapping, maps to performance.

🔥 On inflammation We study the inflammatory response extensively. This talk focused on what it drives downstream: how glial activation alters neurocomputation at the tissue-electrode interface. The inflammatory cascade is the upstream input; circuit-level dysfunction is the output that determines device performance.

💰 On funding Proposed NIH cuts, NOFO collapse, and multi-year funding compression are squeezing pay lines from ~15-20% to ~3-5%. Industry will not fund pre-competitive biological questions. Two actions if you are a U.S. taxpayer:

NIH RFI (deadline May 16) https://www.bioniclab.org/news/request-for-information-nih

Contact representatives on FY2027 appropriations: NOFO collapse and multi-year funding mechanisms

If you work in clinical neurotechnology with unexplained failure patterns, or are a basic neuroscientist with no engineering translation path, reach out.

#NeuralEngineering #BrainComputerInterface #Neuroscience #NIHFunding #BasicScience

📋 Brain-computer interface clinical trials are enrolling participants without routinely screening for genetic risk factors (which adds significant cost). At the same time, we have data showing that implanting an electrode into the brain of a mouse that carries Alzheimer’s risk genes triggers amyloid plaque accumulation and disease pathology months earlier than it would have developed naturally.

🔬 Let me be precise about what that means and what it doesn’t mean.

In our experiments, we implanted microelectrodes into two-month-old mice carrying genetic mutations (APP/PS1) associated with familial Alzheimer’s disease (This can’t be done in NAMS, because NAMS lack the full metabolic system that Alzheimer’s disease exploits). These mice don’t normally develop MX04-labeled Alzheimer’s plaques until four to five months of age. Following electrode implantation, we observed plaque accumulation and markers of Alzheimer’s pathology at two months, two to three months ahead of the typical timeline. The focal brain injury from implantation appeared to accelerate disease progression in animals already on a genetic trajectory toward neurodegeneration.

🚫 This does NOT mean that BCI implants cause Alzheimer’s disease in healthy humans. It means that in a brain already moving toward neurodegeneration, a focal injury event, including but not limited to electrode implantation, may accelerate that progression.

🩺 The implication for clinical trials is not that we should stop enrolling participants. It is that we should know something about the metabolic and genetic landscape of the brain we are implanting into. A 35-year-old with a family history of aggressive early-onset Alzheimer’s and a 35-year-old without that history may respond to the same implant procedure very differently. We don’t currently have the data to know the magnitude of that difference in humans, which is precisely why we need to start asking the question systematically.

🦺 I raise this not to alarm anyone currently enrolled in a BCI trial 🌟the devices have strong safety records for the populations studied🌟 but because science done in mice has a way of becoming relevant in humans, and the time to think about it is before, not after, problems emerge.

AD: https://lnkd.in/g6vNw8NP

#Alzheimers #Neurodegeneration #BrainHealth #ClinicalTrials #Neuroscience #Neurotechnology #BCI