(Wetzlar, Germany) was used to obtain Z-plane optical stacks at 0.8
µm thickness using a 63× oil immersion lens (Morris et al., 2010a (link)). Forty cells were
sampled per animal, with both individual cells and clusters pseudo-randomly
selected from multiple positions along both the superior blade and inferior
blade from at least five separate sections per brain. Image-Pro Plus
(Version 6.3, Media Cybernetics, Silver Spring, MD, USA) was used to render
a 3D model from each section of tissue but only included three channels per
stack at a time due to technological limitations of the software. Initially,
models using Ki67, Sox2, and NeuroD1 were uploaded in order to investigate
presumed Type 2a, 2b, and 3 progenitor cells (as defined in Figure 4A). A subsequent model using
Ki67, GFAP, and Sox2 combined with the data from the first model allowed for
the differentiation of Type 1 cells from the Type 2a pool. Surface values
for each channel were optimized to reduce background signal without loss of
positively stained cells. To confirm the accuracy of the 3D reconstruction,
the models were compared to the raw images to ensure inclusion of all
positively stained cells.