The blood-brain barrier is the most effective security system in medicine: it keeps toxins and pathogens out of the brain, and it keeps almost every drug out with it. Biologics such as antibodies rarely reach the brain at therapeutic concentrations, a bottleneck that has frustrated attempts to treat brain disease with gene-silencing drugs. Researchers have tried viruses, antibodies, focused ultrasound, and synthetic nanoparticles to smuggle therapies across, with mixed results. A team from Southern Medical University and the Shenzhen Bay Laboratory has now built the delivery vehicle out of the brain’s own immune cells. The approach fuses membranes from microglia, the resident immune cells of the brain, with synthetic liposomes carrying a piece of RNA that silences an enzyme fueling inflammation. In two mouse models, the hybrid nanovesicles crossed the barrier, quieted the brain’s inflammatory response, and restored cognitive function. The work appears in Nature Communications.
The target is an epigenetic enzyme with an unwieldy name: METTL3, the core writer of N6-methyladenosine, or m6A, the most abundant internal modification of messenger RNA. m6A marks are deposited by writer enzymes, stripped by erasers, and interpreted by reader proteins, a cycle that is dynamic and reversible. In inflamed microglia, METTL3-driven m6A marks stabilize inflammatory gene programs. Silencing METTL3 has the opposite, restorative effect: it stabilizes SOCS3, a natural brake on the JAK2-STAT3 signaling pathway, pushing microglia from the inflammation-amplifying M1 state toward the repair-oriented M2 state. The approach is a form of epigenetic editing, which changes gene expression without altering the DNA sequence. That reversibility is a genuine advantage over CRISPR-based editing, which cuts DNA and carries a small risk of unintended rearrangements.
The delivery engineering is the heart of the paper. The researchers polarized cultured microglia into the anti-inflammatory M2 state, stripped their plasma membranes, and processed them into nanovesicles roughly 150 nanometers across. Separately, they loaded cationic liposomes with small interfering RNA against METTL3. The two were fused by mixing, sonication, and extrusion into a single hybrid particle, siMETTL3-hNVs, about 190 nanometers in diameter. The fusion matters because the nanovesicle half brings biological intelligence that no synthetic particle can imitate. It carries the microglial surface protein CCR2, which follows the chemokine CCL2, a signal that inflamed brain tissue emits like a flare. It also carries cytokine receptors, IL-6 receptor, IL-1-beta receptor, and TNF-alpha receptor, which act as decoys that soak up inflammatory cytokines in the neighborhood. The result is a particle with a homing system, a decoy system, and a gene-silencing payload.
Delivery uses both tricks at once. CCR2-CCL2 chemotaxis pulls the nanovesicles toward the inflamed brain, and caveolae-mediated transcytosis carries them across the blood-brain barrier, the same route some viruses use to sneak in. Once inside, the particle works on two fronts: its receptors neutralize cytokines extracellularly, while the siRNA silences METTL3 inside microglia. The two effects are synergistic. In vitro, 40 micrograms of the hybrid removed measurable amounts of TNF-alpha (42.01 picograms), IL-6 (42.62 pg), and IL-1-beta (198.47 pg), and the membrane-based scavenging contributed about 2.8 times more to cytokine reduction than the liposome payload alone.
The in vivo results come from two female mouse models. In acute neuroinflammation induced by lipopolysaccharide, treated mice recovered weight faster, showed a novel-object recognition preference matching healthy controls, and navigated the Morris water maze significantly better, indicating restored spatial memory. In radiation-induced brain injury, produced by 30 gray whole-brain irradiation in six-week-old mice, a model of the progressive cognitive decline seen in patients after whole-brain radiotherapy, the treatment attenuated weight loss, improved behavioral and memory performance, and reduced the activation markers GFAP and IBA-1 in the hippocampus, with pro-inflammatory genes Il1b, Il6, and Tnf suppressed in hippocampal tissue. The authors report a favorable safety profile across both models.
The limitations are stated plainly in the discussion. Both models are acute or injury-driven; the utility of the platform in chronic neurodegenerative conditions such as Alzheimer’s or Parkinson’s disease remains untested. Species differences in nanomedicine pharmacokinetics mean mouse results may not translate directly. Dosing regimens, long-term safety, and scalable manufacturing all still need work. Three authors hold patent applications on the biomimetic delivery system, a fact disclosed in the paper. And the manuscript is an early-access version, published before final copyediting.
The conceptual elegance of the approach is that it treats inflammation as a delivery problem with a biological answer. The inflamed brain already summons immune cells by emitting CCL2; the nanovesicle simply answers the call in the immune system’s own language, carrying a message of its own. If the strategy survives translation, it would give neurobiologists a platform for epigenetic reprogramming of brain inflammation that is reversible, targetable, and built from materials the body already makes.
Sources: Xu, L., Pan, Y., Li, G. et al. Targeting m6A writer METTL3 with engineered nanovesicles reduces neuroinflammation in vitro and in vivo. Nature Communications (2026). DOI: 10.1038/s41467-026-75862-4.