2.3. Microscopy and Blinded Colocalization Analysis

Control and stained specimens were completely screened, and cells were documented using an upright microscope (Axioskop 2, Zeiss, Oberkochen, Germany) equipped with halogen illumination (HA 100, Zeiss) and narrow-band fluorescence filter sets (472/30/495 DC mirror/513/17 (green channel), 560/25/585/620/60 (red channel), and 406/15/425/460/50 (blue channel)). Images were taken at 10×, 20× (Plan-apo), or 100× (N.A. 1.3, oil) using a 2048 × 2048 pixel monochrome camera (Spot Insight 4, Diagnostic Instruments, Sterling Heights, MI, USA) with 7.45 µm square pixels and 14 bit sampling depth (CCD chip type KAI 4021). Channel overlay was performed at the step of image acquisition using the software Spot 5.0 Advanced (SPOT Imaging Solutions, Diagnostic Instruments), at the same time correcting for chromatic aberration [26].

Uneven image illumination was flattened applying the “rolling ball” algorithm (at 500 pixel radius) as implemented in ImageJ [27]. Background was adjusted on the basis of the control staining, and identically for all frames from a given staining, using Adobe Photoshop. For presentation, brightness and color were adapted to represent the microscope image. Only global, non-selective, and linear image operations were carried out. In some frames of Rab6A/NeuN-stained specimens (see Figure S9), deconvolution was applied to further increase image clarity and resolution, using a motorized microscope (Zeiss Cell Observer) and a CCD (Sony IXC285AL) with 6.45 µm square pixels. Stacks of 60 images were acquired at 50 nm spacing and 100× final magnification (63× 1.4N.A. × 1.6×). The image stacks were deconvolved applying calculated point spread functions (PSFs) and iterative deconvolution (Volocity Software (Quorum Technologies Inc., Puslinch, ON, Canada)). Quantitation of cellular colocalization was blinded to reduce bias for cell morphology and cell type. In double and triple labellings, a reference population was determined corresponding logically to the issue, e.g., to investigate the proportion of Rab6A⁺ microglial cells, the reference population was Iba1⁺, not vice versa. Of the 2 or 3 color channels, all except that of the reference population were rendered invisible in Photoshop. A large number of cells appearing “complete” (with obvious soma, nucleus (presumed, if not stained), and stem processes in the section plane) were first dot-marked on a transparent “overlay” (in Photoshop, see Figure S1). In a second step, the channel to be examined was re-opened for colocalization. Only the dot-marked cells were checked for cellular anti-gen colocalization to finally obtain the percentage of Rab6A⁺/Iba1⁺ microglial cells. In some cases, double and triple labellings were repeatedly examined with different reference populations, to study different questions.

Assessing colocalization at the level of cross-sectioned soma and nucleus largely precludes false positive colocalization by superimposition in 3D. Firstly, absence of tissue permeabilization, as in our protocol, results in antibody penetration depth of less than 1 µm. Also, cells below the section surface can only be labelled intracellularly if continuous with a process physically cut by tissue sectioning; the plethora of small astrocytic processes exposed at the section surface are too narrow to provide antibody access to deeper structures. Second, with a depth of focus of approximately 500 nm (100× objective), and an established astrocyte nucleus size of 5–6 µm, spurious superimpostion of the Rab6A⁺ section surface and a deeper cell soma is highly improbable. In this context, we did not observe Rab6A⁺ cell nuclei. In summary, the images examined for colocalization represent labelling at the cell surface, in a focal plane approximately 500 nm thick.

Distribution of Rab6A⁺ in astrocytes and other cells was assessed qualitatively by structural analysis of “complete cells” in the cortex, corpus callosum, hippocampus, and thalamus, imaged at 100×, altogether 14.568 astrocytes in 1774 frames, from 3 mice. For quantitative assessment, morphometric object-oriented image analysis was performed on regions-of-interest (ROIs) of 10 representative cells each from the 4 types found and classified qualitatively. Using particle analysis and planimetry commands in ImageJ [27], we determined individual area and number of Rab6A⁺ objects.