Strategic subtraction can sometimes make a neural network work better
The brain has long been understood as a living architecture, constantly reshaping itself through the gain and loss of synaptic connections — yet deliberately editing that architecture has remained beyond reach. Now, researchers at South Korea's Institute for Basic Science and Korea Brain Research Institute have created a tool called SynTrogo that can selectively remove synapses from a targeted brain circuit, and in doing so, paradoxically strengthen the connections that remain. Tested in mice, the approach improved memory formation without disrupting behavior or neuronal activity, offering the first proof that the brain's physical wiring can be intentionally redrawn. The implications reach toward some of humanity's most consequential neurological disorders, where the balance of connection and disconnection has gone quietly, devastatingly wrong.
- Decades of neuroscience could modulate brain activity but could not touch the brain's physical wiring — SynTrogo breaks that barrier by engineering astrocytes to prune specific synapses with molecular precision.
- Removing 27% of excitatory synapses in the hippocampus did not weaken the circuit — it triggered a cascade of structural reinforcement in the synapses that survived, enhancing the very process believed to underlie memory.
- Mice with edited circuits showed stronger, more durable fear memories than controls, yet retained the flexibility to unlearn those memories and showed no disruption to anxiety, movement, or working memory.
- The mechanism is not fully understood — whether astrocytes directly consume synapses, crowd out their formation, or both remains unclear, and memory effects varied with conditioning strength, signaling that benefits may not be universal.
- SynTrogo is a research instrument, not a therapy, but it inaugurates the concept of 'connectome editing' — a deliberate, cell-type-based reshaping of neural architecture with potential relevance to autism, schizophrenia, Alzheimer's, and brain injury.
The brain's wiring is not fixed. Neural connections form, strengthen, and are pruned across a lifetime, and when that pruning goes wrong — too aggressive or too sparse — the consequences can include schizophrenia, autism, and Alzheimer's disease. Neuroscientists have long known how to adjust neural activity, but reshaping the brain's physical structure itself has remained out of reach. A new tool developed in South Korea may change that.
Researchers at the Institute for Basic Science and Korea Brain Research Institute created SynTrogo — short for Synthetic Trogocytosis — which selectively removes synapses from a chosen brain circuit without disturbing the electrical activity of the neurons involved. The system works by engineering astrocytes, the star-shaped support cells already involved in natural pruning, to recognize molecular tags on specific neurons. This lock-and-key mechanism allows astrocytes to dismantle targeted connections with surgical precision, leaving neighboring, untagged neurons untouched.
When the team applied SynTrogo to the hippocampal CA3-CA1 pathway in mice — a circuit central to learning and memory — excitatory synapses in the targeted area fell by roughly 27% over three weeks. What followed defied expectation. Rather than weakening the circuit, the surviving synapses grew structurally larger, accumulated more neurotransmitter, and showed enhanced long-term potentiation, the cellular hallmark of memory formation. The mice outperformed controls in fear-memory tests, retaining memories more durably — yet they could still unlearn those memories when circumstances changed, and their general behavior remained unaffected.
"The brain can adapt and strengthen its function even when the total number of synaptic connections is reduced," noted Dr. Lee Kea Joo of the Korea Brain Research Institute. The precise mechanism behind this compensation remains partially mysterious, and the memory benefits varied with conditioning intensity, suggesting the effects are not uniformly positive across all contexts.
SynTrogo is not a clinical treatment, but it establishes a new principle: the brain's structural architecture can be deliberately edited by engineering interactions between cell types, independent of neural firing. This opens the door to connectome editing — a field that could one day help researchers understand, and perhaps correct, the synaptic imbalances at the heart of some of the most complex disorders in medicine.
The brain's wiring is not fixed. Throughout a person's life, neural connections strengthen, weaken, and disappear in a constant process of pruning that keeps circuits functional. When that pruning goes wrong—becoming too aggressive or too sparse—the consequences can be severe: schizophrenia, autism, Alzheimer's disease, and other disorders of cognition and behavior. For decades, neuroscientists have known how to turn neural activity up or down. Rewiring the brain's physical architecture itself has remained elusive. Until now.
Researchers at South Korea's Institute for Basic Science and Korea Brain Research Institute have developed a tool called SynTrogo, short for Synthetic Trogocytosis, that does something previously thought impossible: it selectively removes synapses—the contact points where neurons communicate—from a chosen brain circuit without disrupting the electrical activity of the neurons themselves. The breakthrough centers on astrocytes, star-shaped support cells that wrap around synapses and already play a role in the brain's natural pruning process. By engineering these cells to recognize and respond to molecular tags on specific neurons, the researchers created a lock-and-key system that allows astrocytes to nibble away at unwanted connections with surgical precision.
In experiments with mice, the team targeted the hippocampal CA3-CA1 pathway, one of the most thoroughly studied circuits involved in learning and memory. They engineered neurons in one region to display a molecular tag and astrocytes in the adjacent region to carry a matching binding partner. Within three weeks, the number of excitatory synapses in the targeted area had dropped by approximately 27 percent. The reduction was selective: inhibitory synapses remained unchanged, and nearby neurons lacking the engineered tag showed no loss of connections. What happened next defied intuition. Rather than weakening the circuit, the remaining synapses underwent dramatic structural changes. Presynaptic terminals grew larger and accumulated more synaptic vesicles—the packets of neurotransmitter ready for release. Postsynaptic spines expanded. Mitochondria proliferated, suggesting the circuit had mobilized additional energy resources. Electrical recordings showed that long-term potentiation, the cellular process widely linked to learning and memory formation, was significantly enhanced.
When the researchers put the modified mice through fear-conditioning tests, the animals demonstrated stronger memory than controls. With mild conditioning, they froze more during both recent and remote memory tests. With stronger conditioning, they maintained their memories over time while unmodified animals showed more decay. Yet the mice retained behavioral flexibility: they could still extinguish those fear memories when conditions changed, and their general anxiety, locomotion, and working memory remained unaffected. Dr. Lee Kea Joo of the Korea Brain Research Institute noted the paradox at the heart of the finding: "The brain can adapt and strengthen its function even when the total number of synaptic connections is reduced."
The mechanism underlying this counterintuitive result remains partially mysterious. Using correlative light and electron microscopy, the team observed unusually tight interfaces between astrocytes and axons, localized membrane deformation, and partial enclosure of axonal regions by astrocytic processes. The structural changes were spatially restricted, and there was no evidence of broad axonal loss or neuron death even weeks after the intervention. Yet the authors acknowledge a significant limitation: fixed brain tissue cannot reveal the real-time sequence of synapse removal the way cultured cells can. The reduction in synapse number could result from multiple mechanisms—direct nibbling by astrocytes, impaired synapse formation from intensified astrocyte contact, or both. Additionally, the memory improvements varied depending on the strength of the conditioning stimulus, suggesting the effects may not be uniformly beneficial across all circuits or situations.
SynTrogo remains a research tool, not a clinical treatment. But it represents a conceptual breakthrough: proof that the brain's physical wiring can be directly edited by engineering interactions between different cell types, independent of neuronal activity. This opens the door to what researchers call "connectome editing," a deliberate reshaping of the brain's structural architecture. The implications extend to disorders where abnormal synapse numbers or disrupted circuit organization are thought to play a central role—autism spectrum disorder, schizophrenia, Alzheimer's disease, and traumatic brain injury. The platform may help answer a long-standing question: when synapses are lost or overproduced in disease, how much of the problem stems from the sheer number of connections, and how much from the way the remaining circuit adapts? Much work remains, particularly real-time studies in living brains and experiments across different circuits. But the findings suggest that strategic subtraction—removing the right connections in the right places—can sometimes make a neural network function better, not worse.
Notable Quotes
This is the first demonstration that brain circuits can be directly edited by engineering physical interactions between neurons and astrocytes, independent of neuronal activity.— Dr. Lee Sangkyu, IBS Center for Memory and Glioscience
The brain can adapt and strengthen its function even when the total number of synaptic connections is reduced.— Dr. Lee Kea Joo, Korea Brain Research Institute
The Hearth Conversation Another angle on the story
Why does removing synapses improve memory? Shouldn't fewer connections mean weaker signals?
That's the intuition everyone has. But the brain isn't a simple telephone network. When you remove the right synapses, the remaining ones don't just survive—they reorganize. They get bigger, hold more neurotransmitter, build more infrastructure. It's like a company that cuts staff but ends up more efficient because the remaining people are better equipped and more focused.
How do the researchers know they're removing the right synapses and not damaging the circuit?
They engineered the system to be selective. The astrocytes only respond to neurons with a specific molecular tag. Everything else is left alone. And they watched carefully—no broad cell death, no inflammation, no signs of collateral damage even weeks later.
But you said the mechanism isn't fully understood. What's still unclear?
Whether the astrocytes are actually nibbling away at synapses or whether their presence somehow prevents new synapses from forming. And whether the benefits hold across different circuits or different types of learning. The mice showed stronger fear memory, but that's one specific task.
What's the real-world application here?
Right now, it's a tool for understanding how circuits reorganize when synapse numbers change. But eventually, it could help treat disorders where synapse loss or overproduction goes wrong—Alzheimer's, autism, schizophrenia. If you could selectively prune the right connections, you might restore function.
Is this close to being used in humans?
No. This is mice, in a controlled lab setting, targeting one specific circuit. There's a long road from here to any clinical use. But it's the first time anyone has shown you can directly edit a brain's wiring without disrupting its electrical activity. That's the breakthrough.