Lose Some Brain Wiring, Gain Some Memory? Mice Suggest It’s Possible

A brain that ends up with less wiring and still remembers more sounds like a contradiction. That is roughly what a team of researchers in South Korea found after building a molecular system that reduced the density of connections between brain cells in a targeted memory circuit. Thinner connectivity did not degrade that circuit. Instead, the connections left behind grew bigger, reorganized, and the mice held onto a learned fear association longer.

Researchers call the tool SynTrogo, short for Synthetic Trogocytosis, and it borrows from a natural process in which one cell chomps off pieces of another cell’s outer membrane. By engineering a matched pair of proteins onto two different cell types, the team coaxed star-shaped support cells called astrocytes into grabbing and swallowing bits of neighboring neurons. In cultured dishes, time-lapse imaging showed astrocytes taking that material in completely. Inside the brain, what followed was a drop in the number of synapses, the junctions where brain cells pass signals, within one memory circuit.

Results appear in Nature Communications. Work was done entirely in mice, and the effect on memory was not uniform: benefits showed up under one shock intensity and held steady under another, which the authors flag as a reason the approach may not help in every behavioral situation. Still, the finding chips away at a simple assumption, that fewer synapses always means worse brain function.

Building a tool that reduces brain cell connections

Two engineered proteins do the work, one acting as bait and the other as a hook. Bait is a glowing molecule called GFP, a lab staple, anchored to the surface of the target cell. Hook is a tiny antibody fragment built to latch onto GFP, anchored to the surface of the cell doing the eating.

When cells carrying both proteins touch, the hook cell grips the bait cell’s membrane and pulls pieces of it inward, taking in membrane fragments along with whatever molecules sit nearby. Material flows in one direction only, from bait cell to hook cell. Cell type does not set that direction; which protein a cell carries does. Swallowed material mostly landed in lysosomes and the Golgi apparatus, compartments where cells break things down and recycle them.

Importantly, the process throttles itself. Each round of nibbling consumes the surface proteins that drive it, so the fuel runs down and the eating cell never finishes off its partner. In lab dishes, mixing the two cell types caused no detectable harm to survival through 72 hours, and cell division continued normally for at least 18 hours.

Testing SynTrogo inside a living mouse brain

After the concept worked in culture, researchers moved it into the brains of male C57BL/6J mice, 9 to 16 weeks old. Their target was a well-mapped memory pathway running between two regions of the hippocampus, called CA3 and CA1. Viruses delivered the genetic instructions: CA3 neurons displayed the GFP bait, while CA1 astrocytes carried the hook.

Three weeks later, the signatures were clear. GFP from neurons had piled up inside astrocyte territories. Molecules found only at the sending end of a neuron turned up inside astrocyte cell bodies, and their levels tracked with how much GFP signal had accumulated there. Confocal imaging alone cannot settle whether that material was fully swallowed, which is why the team turned to electron microscopy. Counts of excitatory connections, the kind that drive activity, fell inside treated areas. A different class of connection that dampens activity was not reduced and rose slightly across the broader region, pointing to a shift in the balance between the two.

Scale matters here. Even though more than 70 percent of the targeted cells carried the engineered proteins, synapse density dropped only about 27 percent, and the amount varied from animal to animal. Access, astrocyte appetite, and the sturdiness of individual connections all likely shape how much gets trimmed.

Neurons survived. Density of bait-carrying neurons held steady at three weeks and again at two months. Axons, the long cable-like extensions neurons use to send signals, showed no loss even two months after injection. Astrocytes and microglia, the brain’s resident immune cells, showed no signs of abnormal activation, which argues against harmful inflammation driving the result.

To see structures too small for an ordinary microscope, researchers overlaid fluorescence images onto electron microscope images of the same tissue. Astrocyte membranes made far more contact with neuronal surfaces under SynTrogo, especially at boutons, the bulb-shaped swellings along axons where connections form. At these contact sites, the combined membrane profile was thicker, about 27 nanometers versus roughly 15 nanometers in controls, alongside a narrower gap between the cells. In places, astrocyte extensions appeared to pinch and partly wrap segments of axon, and those wrapped pockets stayed continuous with the parent axon rather than being severed. About 62 percent of them held vesicles, mitochondria, or both.

Specificity was sharp. Axons displaying the bait held fewer connections than control axons, while neighboring axons without it, sitting in the very same treated tissue, kept normal numbers.

Fewer brain cell connections, stronger survivors

Connections that remained showed coordinated changes on both sides. Boutons on the sending side grew larger, packed with more signaling vesicles overall and more sitting ready at the release site. Mitochondria inside those boutons also grew, which fits the higher energy demand of heavier signaling.

On the receiving side, dendritic spines enlarged and the protein-rich zone that catches chemical messages expanded. More of those spines contained a spine apparatus, a stacked internal structure tied to calcium signaling, protein production, and the shuttling of receptors.

Electrical recordings matched the anatomy. Frequency of tiny spontaneous blips dropped, consistent with fewer active connections, while amplitude held steady. Yet the pool of vesicles ready for immediate release grew, and the odds that any given connection would fire successfully went up. After theta-burst stimulation, long-term potentiation, a lasting strengthening of signal transmission widely treated as the cellular basis of learning, stayed elevated for 70 minutes.

Behavior was the real test. Under mild foot shocks of 0.3 milliamps, SynTrogo mice froze more than controls when returned to the conditioning chamber, both two days and 23 days later, freezing being the standard readout that an animal remembers the association. Under stronger shocks of 0.5 milliamps, SynTrogo mice held their freezing levels across both tests while control animals faded by the remote test. Those stronger memories still proved erasable: after four days of extinction training, both groups ended up at similarly low freezing, and both stayed low in a spontaneous recovery test two weeks on. Locomotion, anxiety, and working memory went unchanged. Levels of GluA1, a receptor subunit involved in strengthening, started out lower in untrained SynTrogo mice and climbed back to control levels after conditioning, a pattern consistent with remodeled connections sitting primed for learning rather than already maxed out.

Reach beyond the hippocampus looks plausible but unproven. Applying SynTrogo to a pathway running from the motor cortex to the striatum produced the same redistribution of GFP into astrocyte territories. Whether the strengthening follows outside a memory circuit is still untested.

Medical hopes for editing brain cell connections

Scientifically, SynTrogo separates two things that are normally tangled. Because it works independently of electrical signaling, researchers can watch what happens when connections are physically removed without silencing or exciting the circuit at the same time. Coordinated strengthening on both the sending and receiving sides, under those conditions, points to built-in programs that fire when connections go missing.

Medically, the authors look toward disorders where synapse numbers run abnormal, naming autism spectrum disorder and schizophrenia as cases where it remains unclear how those differences drive disease. Stroke and Alzheimer’s disease come up too, since synapse loss there triggers scaling mechanisms that strengthen what survives. A tool that lowers connection density in one specific circuit could help researchers test how such changes reshape malfunctioning circuits, which is how the authors frame it: a foundation for editing synaptic architecture, with therapeutic relevance still ahead of the evidence. Experiments ran in male mice only, the tool also engages non-synaptic stretches of axon, and no one has shown whether the reduction reflects synapse elimination, disrupted synapse formation, or some combination of both. What the work does offer is a way to ask which connections exist, rather than what signals they carry.

---