In Alzheimer's disease (AD), memory impairment is the most prominent feature that afflicts patients and their families. Although reactive astrocytes have been observed around amyloid plaques since the disease was first described, their role in memory impairment has been poorly understood. Here, we show that reactive astrocytes aberrantly and abundantly produce the inhibitory gliotransmitter GABA by monoamine oxidase-B (Maob) and abnormally release GABA through the bestrophin 1 channel. In the dentate gyrus of mouse models of AD, the released GABA reduces spike probability of granule cells by acting on presynaptic GABA receptors. Suppressing GABA production or release from reactive astrocytes fully restores the impaired spike probability, synaptic plasticity, and learning and memory in the mice. In the postmortem brain of individuals with AD, astrocytic GABA and MAOB are significantly upregulated. We propose that selective inhibition of astrocytic GABA synthesis or release may serve as an effective therapeutic strategy for treating memory impairment in AD.

Astrocytes release glutamate upon activation of various GPCRs to exert important roles in synaptic functions. However, the molecular mechanism of release has been controversial. Here, we report two kinetically distinct modes of nonvesicular, channel-mediated glutamate release. The fast mode requires activation of G(αi), dissociation of G(βγ), and subsequent opening of glutamate-permeable, two-pore domain potassium channel TREK-1 through direct interaction between G(βγ) and N terminus of TREK-1. The slow mode is Ca(2+) dependent and requires G(αq) activation and opening of glutamate-permeable, Ca(2+)-activated anion channel Best1. Ultrastructural analyses demonstrate that TREK-1 is preferentially localized at cell body and processes, whereas Best1 is mostly found in microdomains of astrocytes near synapses. Diffusion modeling predicts that the fast mode can target neuronal mGluR with peak glutamate concentration of 100 μM, whereas slow mode targets neuronal NMDA receptors at around 1 μM. Our results reveal two distinct sources of astrocytic glutamate that can differentially influence neighboring neurons.

The myelin sheath on vertebrate axons is critical for neural impulse transmission, but whether electrically active axons are preferentially myelinated by glial cells, and if so, whether axo-glial synapses are involved, are long-standing questions of significance to nervous system development, plasticity and disease. Here we show using an in vitro system that oligodendrocytes preferentially myelinate electrically active axons, but synapses from axons onto myelin-forming oligodendroglial cells are not required. Instead, vesicular release at nonsynaptic axo-glial junctions induces myelination. Axons releasing neurotransmitter from vesicles that accumulate in axon varicosities induces a local rise in cytoplasmic calcium in glial cell processes at these nonsynaptic functional junctions, and this signalling stimulates local translation of myelin basic protein to initiate myelination.