The Antibody Is Not the Broom: How Lecanemab Rewires Microglia

A Nature Neuroscience study in mice carrying transplanted human brain immune cells found that lecanemab only clears plaque when it can grip microglia and switch on a clearing program. Osteopontin sits near the centre of it. This is laboratory work, not a patient result.

For three years, the standard explanation of how lecanemab works has been the kind of sentence that sounds complete enough to stop a conversation: the antibody sticks to amyloid, and the brain's immune cells come and eat the marked plaque. Flag, then sweep. A team led by Giulia Albertini and Magdalena Zielonka in Leuven decided to test the second half of that sentence, and found it is the part that does the real work. Their paper, published open access in Nature Neuroscience on 24 November 2025, reports that flagging a plaque achieves nothing at all on its own. The broom has to be switched on, and lecanemab is what switches it.

The experiment rests on an unusual animal. The researchers used amyloid-depositing AppNL-G-F mice crossed onto a Csf1r FIRE background, which means the mice grow up without their own microglia, and transplanted human stem-cell-derived microglial progenitors into their brains shortly after birth. The result is a mouse brain policed by human immune cells, which is as close to human microglial biology as a living model currently gets. From four months of age, groups of these mice received weekly injections of 10 mg/kg for eight weeks: real lecanemab, a control human IgG1, or lecanemab LALA-PG, an engineered twin that binds amyloid exactly as the real drug does but has had its Fc tail disabled so it cannot be grabbed by an immune cell. Lecanemab reduced both beta-sheeted plaque area and total amyloid peptide burden. The Fc-silenced twin did not, and the telling detail is visual: the disabled antibody piled up on the plaques and stayed there, decorating them, while in the lecanemab animals the antibody was found inside the human microglia instead. Mice that never received human microglia showed no benefit from lecanemab at all. In a separate cohort of immunocompetent AppNL-G-F mice with an intact mouse immune system, a murine version of the antibody, mAb158, lowered amyloid levels, and its LALA-PG variant again did nothing.

To see what the engaged cells were actually doing, the team ran single-cell RNA sequencing on the sorted human microglia and paired it with Nova-ST spatial transcriptomics, which let them read gene expression in small tissue domains sitting right next to plaques; in the preprint version of the work they report 32,568 such domains across four cortical samples, with two mice per condition. What came back was not generic inflammation. It was a focused program: phagocytosis, lysosomal degradation, metabolic reprogramming, interferon gamma genes and antigen presentation, enriched specifically in the microglia near plaques in lecanemab-treated animals and absent in the Fc-silenced group. The single most strongly upregulated gene in both the single-cell and spatial analyses was SPP1, which encodes osteopontin, sitting as a hub in a gene module alongside LGALS1, CTSD and ITGAX.

That gene list is where the finding gets interesting, because those are markers of the disease-associated microglia that pathologists already see clustered around plaques in Alzheimer's brains, crowding the problem without solving it. The authors' reading is that lecanemab pushes those same cells further than disease alone ever pushes them, past huddling and into clearing. They then tested osteopontin directly, plating human microglia onto brain cryosections from six-month-old AppNL-G-F mice and adding the protein at rising concentrations from 17 up to 1,350 ng/ml. More osteopontin meant less amyloid coverage, a dose-dependent effect across three independent experiments. Sections without human microglia showed nothing, which is the control that matters: osteopontin does not dissolve plaque, it changes what a cell does to plaque.

For a family reading this, the honest translation is narrow but not trivial. Nothing here measured memory, thinking or daily function; nothing here involved a person taking a drug. What it offers is a mechanism, and mechanisms are how dead ends get explained and new routes get drawn. If clearance depends on Fc engagement rather than on binding alone, that is a design principle for the antibodies still in development, and it raises the possibility, explicitly floated by the authors, that the microglial program itself could one day be a target without an antibody at all. The paper also reports a smaller point worth holding onto: peri-plaque neuritic damage, measured by LAMP1 staining, was significantly reduced in lecanemab-treated animals, while synaptic puncta around plaques were not significantly changed, which is at least a hint that the activated cells were not indiscriminate about what they removed.

The field spent three years arguing about what anti-amyloid antibodies are worth. This study argues, in mice with human microglia, about what they actually do, and the answer is that the drug's value may lie less in what it binds than in which cell it manages to wake up.