Introduction: Region-specific synapse loss is an early hallmark of Alzheimer’s disease (AD) that precedes neuronal loss and amyloid-β (Aβ) plaque deposition. Microglia mediate synapse elimination in AD, however the mechanisms that confer synapse vulnerability in specific circuits remain unclear. Astrocytes regulate synapse homeostasis through peri-synaptic processes, but their contribution to early circuit AD pathology is poorly understood.
Objective: We sought to determine whether astrocytes shape local synapse vulnerability during early amyloidosis and to define the molecular mechanisms underlying astrocyte-microglia crosstalk (1).
Methods: Using high-resolution Airyscan imaging, correlative light-electron microscopy (CLEM), in vitro microglia-synaptosome engulfment assays with astrocyte conditioned media, in vivo astrocyte-specific viral manipulations and behavioural assays, we characterised astrocyte phenotypes and synapse dynamics in the slow-progressing Aβ hAPP NL-F knock-in (KI) model and the more aggressive Aβ hAPP NL-G-F KI model. Gain- and loss-of-function approaches included viral astrocyte-specific MFG-E8 overexpression and viral CRISPR-Cas9 deletion, as well as microglia-specific integrin-β3 deletion using the tamoxifen-inducible Cx3cr1-CreERT2 model.
Results: Using tdTomato labelling and CLEM, we re-discovered a distinct region-specific population of astrocytes with terminal glycogen- and p62-enriched bulbous peri-synaptic processes that emerges at 6 months in the NL-F KI (N=5 per genotype), prior to plaque deposition and gliosis, when synapses are lost. Bulbous astrocytes accumulated in circuits vulnerable to AD pathology like the hippocampus (N=5 mice). Glycogen- and p62-enriched astrocytic structures were detected in human AD tissue within vulnerable regions (N=12 control and AD cases), supporting translational relevance. Functionally, bulbous astrocytes exhibited reduced peri-synaptic coverage (N=5-7 mice), impaired baseline and phenylephrine evoked GCaMP Ca²⁺ signalling (N=3 mice), and were associated with localised increases in microglial Homer1⁺ synapse engulfment (N=5 mice) and excitatory (Homer1+/Bassoon) synapse loss (N=6 mice). Bulbous processes were enriched for the phagocytic ligand milk fat globule-EGF factor 8 (MFG-E8) (N=5 mice), which interacts with integrin receptors to facilitate engulfment. Astrocyte-specific MFG-E8 overexpression in wild-type mice induced microglial Homer1 engulfment (N=6 mice) followed by excitatory synapse loss and increased bulbous astrocyte abundance (N=4-5 mice). Microglia-specific integrin β3 deletion abolished MFG-E8-driven Homer1 engulfment in vivo (N=4-5 mice per group), confirming integrin-dependent signalling. Conversely, CRISPR-Cas9 astrocytic Mfge8 deletion normalised microglial Homer1 engulfment and excitatory synapse density in the NL-F KI (N=4-6 mice per group) model and reduced plaque burden, dystrophic neurites, gliosis, contextual fear deficits, and bulbous astrocyte accumulation in the NL-G-F KI (N=7-12 mice per group).
Conclusions: These findings identify a previously unrecognised astrocyte-driven mechanism that establishes local synapse vulnerability in early AD through MFG-E8-integrin-dependent astrocyte-microglia signalling, positioning astrocytes as upstream regulators of circuit-specific synapse loss (2).
Statistical analysis: Analyses were performed using animal averages, with appropriate parametric or nonparametric tests based on normality testing, correction for multiple comparisons, and significance set at α = 0.05 (mean +/- S.E.M). Both male and female mice were used.
Ethics: All animal experiments were performed in accordance with the UK Animal Scientific Procedures Act 1986 and approved by the UK Home Office with institutional veterinary oversight at University College London. Human post-mortem tissue was obtained from the Newcastle Brain Tissue Resource with informed consent and appropriate ethical and material transfer approvals.