Free Amyloid-Beta Builds Synapses While Aggregated Forms Destroy Them

Alzheimer's disease research has long treated amyloid-beta (Aβ) peptides as uniformly toxic, fueling therapeutic strategies aimed at eliminating them entirely. A landmark study from Thomas Südhof's laboratory at Stanford, published in the Journal of Clinical Investigation, challenges that framework by demonstrating that the aggregation state of Aβ — not its mere presence — determines whether these peptides harm or help synapses. The work addresses a pivotal contradiction: Südhof's own prior research showed that genetically elevated Aβ42 production in human neurons via an APP-Swedish knockin was actually synaptogenic, not toxic. The new study provides a mechanistic resolution.

The researchers treated human neurons chronically (from DIV6 to DIV45) with chemically defined synthetic Aβ40, Aβ42, and Aβ42arctic peptides at concentrations ranging from 0.005 to 0.2 μM. These three variants were chosen for their distinct aggregation kinetics: Aβ40 aggregates slowly, Aβ42 aggregates more rapidly, and Aβ42arctic (carrying the E693G arctic mutation) aggregates fastest. Fresh aliquots were used at each media change to control for time-dependent aggregation, and peptide identity was confirmed by mass spectrometry. Neurotoxicity was assessed by three independent assays: MTT metabolic activity, a novel NLS-TdTomato nuclear-to-cytoplasmic leakage assay specific for neurons, and an Incucyte Cytotox Green membrane-integrity assay.

Aβ40 produced no detectable neurotoxicity at any concentration tested and generated almost no 6E10-immunoreactive aggregates by immunofluorescence or native gel analysis. In stark contrast, Aβ42 became neurotoxic at ≥0.05 μM by MTT and ≥0.15 μM by NLS-TdTomato, while Aβ42arctic was toxic at even lower thresholds (≥0.025 μM by MTT, ≥0.1 μM by NLS-TdTomato). Crucially, toxicity correlated directly with aggregate formation: 6E10-positive puncta appeared at concentrations matching the onset of toxicity, and preaggregated Aβ42 pellets caused the same viability loss as the aggregation-prone full preparation. Scrambled and reversed sequence peptides were entirely inactive, confirming sequence specificity of all effects.

The synapse data revealed the functional dichotomy most sharply. Aβ40 produced a striking, dose-dependent ~2-fold increase in dual-positive (Synapsin-1/PSD95) synapse density across almost all concentrations tested, with no change in puncta size, dendritic arborization, soma dimensions, or endosomal parameters. Aβ42 at low, non-aggregating concentrations yielded a modest but significant ~25% increase in synapse density, mirroring Aβ40. At higher, aggregating concentrations, Aβ42 reversed to cause ~35% synapse loss. Aβ42arctic, the fastest aggregator, was synaptotoxic even at the lowest concentrations studied. Mechanistically, synaptotoxicity appeared to precede frank neurotoxicity and manifested as an initial contraction of the presynaptic vesicle cluster followed by full synapse elimination — a signature implicating presynaptic vesicle dynamics as an early pathological target.

The therapeutic implications are significant. Current FDA-approved anti-amyloid antibodies (lecanemab, donanemab) preferentially target aggregated Aβ species and show greater clinical benefit than antibodies binding free Aβ — a pattern now mechanistically explained by this study. The authors propose that AD progression involves a dual pathological shift: gain of synaptotoxicity from aggregated Aβ and loss of the synaptogenic function of free Aβ. They suggest future therapies might aim not to eliminate all Aβ but to restore the balance toward free, monomeric peptides. The study used human iPSC-derived neurons co-cultured with murine glia, providing a clinically relevant cellular context, though in vivo confirmation and identification of the synaptic receptor mediating Aβ's synaptogenic effect remain open questions.