Topological classification of retinal microglia during development

Abstract

Purpose : Microglia are functionally associated with two morphological states: surveilling microglia consisting of highly branched processes, whereas activated microglia exhibit an amoeboid shape. Microglia often exist in a transitory state between the two extremes, and it is often not clear whether microglia become beneficial or detrimental to neuronal function. To determine the state of the microglia, we need an unbiased mathematical approach (the topological morphology descriptor or TMD) to create and validate a microglia morphology library.

Methods : Here, we have generated a morphological atlas of microglia in the developing retina and applied algebraic topology modeling to identify microglial morphology. Retinas from Cx3cr1-eGFP mice were fixed from postnatal day 0 to P30, immunostained for the monocyte marker Iba-1, and confocal tile-images were acquired. Then, microglia morphology was reconstructed in 3D, which results in a skeleton of connected points, each having a (X,Y,Z) coordinate. We grouped microglia based on the spatial location of their soma into microglia_OPL or microglia_IPL. The 3D-reconstructions were subject to either Sholl analysis or TMD and compared against the adult retina microglia population.

Results : We generated a library of over 10.000 3D-reconstructed microglial cells across postnatal retinal development. During the developmental trajectory, microglia morphology differs between OPL and IPL, which is confirmed with Sholl analysis and TMD. Microglia_IPL appear activated in early developmental phase and increase their spatial coverage over time, which is lacking in microglia_OPL. In addition, TMD allows us a day-by-day comparison of how microglia gradually lose primary small branches and acquire longer processes, as well as a developmental time window at which microglia reach their mature state.

Conclusions : We found that the topological morphology descriptor (TMD) is superior to the traditional Sholl analysis because it overcomes user-induced biases in cell classification introduced with the Sholl method. Our microglia library together with the TMD provides the foundation to determine microglial maturity during development. We are now planning to apply the TMD algorithm to models of retinal degeneration.

This abstract was presented at the 2019 ARVO Annual Meeting, held in Vancouver, Canada, April 28 - May 2, 2019.