Credit: lineartestpilot/iStock/Getty Images Plus

Microglia are proposed to remodel neuronal synapses by pruning them via phagocytosis. However, Weinhard et al. now report that, rather than removing whole synapses, microglia prune presynaptic structures through a selective partial phagocytic process termed trogocytosis, or ‘nibbling’.

The authors analysed fixed hippocampal tissue from postnatal day 15 (P15) mice using quantitative confocal microscopy as well as correlative light and electron microscopy, and found no evidence of phagocytosis of dendritic spines by microglia. Instead, microglia only pruned presynaptic structures — of 250 nm on average — by trogocytosis. Interestingly, trogocytosis seemed to involve the ‘sinking’ of presynaptic structures into the microglial cytoplasm prior to membrane closure, whereas pseudopodia, a hallmark of phagocytosis, were not observed.

microglia only pruned presynaptic structures … by trogocytosis

Next, the researchers used light-sheet fluorescence microscopy of hippocampal slice cultures (derived from P4 mice and maintained for up to 3 weeks) to study the dynamics of microglia–synapse interactions. These cultures recapitulated in vivo microglial physiology, as the dynamics of extension and retraction of microglial processes were consistent with those in previous in vivo studies. Importantly, the authors detected only partial engulfment of presynaptic (and not postsynaptic) structures; these trogocytic events were rapid, often completing in <3 min. Interestingly, unlike the phagocytic clearance of apoptotic cells by microglia in the developing hippocampus, trogocytosis of synaptic structures by microglia does not involve the complement system, as the authors found no defects in trogocytosis in mice deficient in the CR3 complement receptor.

Further analysis of the light-sheet fluorescence microscopy data revealed that microglia contacted dendritic spines often and briefly, and multiple contacts with the same spine were observed. Although 10% of spines contacted by microglia disappeared during imaging, these spines were not in contact with microglia when they disappeared. Moreover, there was no difference in the frequency of disappearance of microglia-contacted versus non-contacted spines. Together, these data suggest that microglia do not induce spine elimination in this preparation.

Intriguingly, transient contacts between spines and microglial processes often resulted in formation of filopodia from the spine head, which extended towards the microglial process, whereas no spine-head filopodia formed from non-contacted spines. Additionally, contact with the microglial process sometimes resulted in dramatic stretching of spines, suggesting that microglia could influence the morphology of spines.

These data suggest that, in the mouse postnatal hippocampus, microglia seem to have a more nuanced role in synaptic remodelling, including the trogocytosis of axonal structures and the restructuring of postsynaptic sites.