2.6.2. Fluorescent Immunohistochemistry

Triple immunofluorescent staining was used to visualize neurons, astrocytes, and microglia. Brain sections were transferred into a 24-well tissue culture plate (TPP, Trasadingen, Switzerland) and were immunostained in a free-floating manner in a 500 μL volume on an orbital shaker (Heidolph Instruments, Schwabach, Germany). After three 10 min washes in 0.1 M PB, sections were washed three times for 10 min in tris-buffered saline (TBS), then blocked in TBS containing 10% normal horse serum (NHS; Vector Laboratories,

Burlingame, CA, USA) and 0.1% Triton-X (Sigma-Aldrich, St. Louis, MO, USA) for 45 min, in order to block nonspecific binding sites and to enhance antibody penetration.

Sections were then incubated with the primary antibodies against NeuN for neurons (NeuN; guinea pig raised-polyclonal, dilution 1:500; product no: 266004, Synaptics Systems GmbH, Goettingen, Germany), GFAP for astrocytes (GFAP; mouse raised-monoclonal dilution 1:500; product no: 173211, Synaptics Systems GmbH, Goettingen, Germany), and IBA1 for microglia (IBA1; rabbit raised-polyclonal, dilution 1:500; HistoSure: HS234013, Synaptics Systems GmbH, Goettingen, Germany) in 0.1% TBS-T overnight at room temperature.

The next day, sections were washed thoroughly in TBS, then fluorescently labeled secondary antibodies made up in TBS were applied at room temperature for 4 h to label the NeuN immunostaining with Alexa488-conjugated donkey anti-guineapig (1:500, Jackson ImmunoResearch Laboratories, West Grove, PA, USA), to label the GFAP immunostaining with Alexa647-conjugated donkey anti-mouse (1:500, Jackson ImmunoResearch Laboratories, West Grove, PA, USA) and to label the IBA1 immunostaining with Alexa594- conjugated donkey anti-rabbit (1:500, Jackson ImmunoResearch Laboratories, West Grove, PA, USA).

Afterwards, stained sections were washed 3 times for 20 min in TBS and 2 times for 10 min in 0.1 M PB, mounted on slides and coverslipped with a mounting medium (Vectashield, Vector Laboratories, Burlingame, CA, USA) and sealed with nail polish.

#### 2.6.3. Confocal Image Acquisition and Analysis of Fluorescent Immunostaining

The digitalization of sections was conducted using confocal microscopy. To obtain high-resolution z-stacks, all images were acquired using a Leica TCS SP8 confocal laser scanning microscope (Leica Microsystems GmbH, Wetzlar, Germany) using HC PL APO CS2 20X/0.75 and HC PL APO CS40X/0.85 dry objectives and unidirectional scanning at 200 Hz. Images were processed and quantified using the Leica Application Suite X (Leica Microsystems GmbH, Wetzlar, Germany) software. High-magnification images of 1024 × 1024 pixels were collected, and regions of interest (ROIs) in individual sections were selected (400 μm × 600 μm). Z-stack deepness was defined as 5 μm, each image comprising of 3 subsequent z-stack layers, resulting in 10 μm-deep recordings. The image of the hippocampal sample was acquired in three different channels. The NeuN staining was used to distinguish hippocampal regions and layers based on the density and relative location of the cells. Light signals from photon scatter around the edges of tissue, tears in the tissue, and vasculature were excluded from analysis.

Fluorescence intensity was measured over ROIs, then corrected for autofluorescence and non-specific signals using a background subtraction. Astrocytes, microglia, and neuronal debris in CA1 and CA3 were consistently counted in the same area in all slices and were expressed as cells/mm2. Image J software was used for manual cell-counting. Three sections from each slide, four slides per animal, and five to eight animals per group were used for histological assessment.

#### Custom Cell Counter Algorithm

For the validation of cell-counting procedures, all ROIs previously investigated were cropped and split to separate RGB (red–green–blue) channels using the Image J software. Each immunostaining channel (green for NeuN, cyan for GFAP, and red for IBA1, respectively) were further analyzed by our custom cell-counter algorithm written in the Python programming language (version 3.9.9), implementing the OpenCV library (version 4.5.5). ROIs were preprocessed by Gaussian blurring, in order to reduce background noise. Desirable foreground image objects were evidentiated using adaptive thresholding, which was followed by morphological opening and closing functions, based on the extent of cell clustering and overall image quality. Next, the area and longest diagonal of all cell-like structures were calculated. If any surface detected had an overlap of at least 60% with another in the subsequent layers and its longest diagonal exceeded the value corresponding to 6.5 μm, the element could be considered a cell. Cellular debris was defined as NeuN- positive fragments with dimensions between 2.5 and 6.5 μm, encircled by GFAP-positive astrocyte signals as described previously [41]. Finally, all cell contours were projected on the original ROI, thus enabling a visual inspection of the detected structures (Figure 2).

**Figure 2.** Analysis of different cell types by a cell-counter algorithm. (**a**) NeuN+ cells (green) with an explicitly delimited pyramidal layer (yellow), (**b**) GFAP+ (cyan), and (**c**) IBA1+ (red) cells evidentiated and contoured by our custom cell-counter algorithm. All three immunostainings are superposed in their z-stack maximum-intensity projections.
