Skip to main content

Data from: Comparisons of quantitative approaches for assessing microglial morphology reveal inconsistencies, ecological fallacy, and a need for standardization

Cite this dataset

Rowe, Rachel; Green, Tabitha; Murphy, Sean (2022). Data from: Comparisons of quantitative approaches for assessing microglial morphology reveal inconsistencies, ecological fallacy, and a need for standardization [Dataset]. Dryad.


Microglia morphology is used as a measure of neuroinflammation and pathology, but different methods to quantify microglia morphology are frequently employed across neuroscience. For reliable inference, it is critical that microglial morphology is accurately quantified and that results can be easily interpreted and compared across studies. We applied five of the most commonly used ImageJ-based methods for quantifying the microglial morphological response to a stimulus to identical photomicrographs and isolated microglial cells, which allowed for direct comparisons of the specificity and reliability of each method.


Tissue samples analyzed in this study were generated from animals used in a previous study that examined the role of microglia in sleep after an immune challenge. All data used in that study are publicly available in the Dryad digital repository. Briefly, male C57BL/6J mice were randomly assigned to PLX (Plexxikon 5622 1200 ppm; formulated in AIN-76A rodent chow) diet to deplete microglia or control diet (AIN-76A rodent chow) for 21 days. On day 21, mice were given an intraperitoneal lipopolysaccharide (LPS; E. coli 0111:B4, Sigma-Aldrich in sterile saline) injection at 0.4 mg/kg in a volume of 0.05 ml to induce an inflammatory challenge. Four days post-LPS injection, all mice were returned to standard diet and were given 10 days for microglia to repopulate. After 10 days of repopulation, all mice were given a second LPS injection and maintained on the standard rodent chow. Brains were collected at 7 days following the second LPS administration.

Perfusion and tissue processing

Seven days following the second LPS injection, a lethal dose of Euthasol® was administered. Mice were transcardially perfused with phosphate-buffered saline (PBS). Brains were drop-fixed in 4% PFA for 24 hours and then cryoprotected in successive concentrations of sucrose (15%, 30%). Using the Megabrain technique, brains were frozen and cryosectioned at 40 µm in the coronal plane and were immediately mounted onto slides.


Prior to staining, the slides were baked at 56ºC for 3 hours. Slides underwent antigen retrieval (sodium citrate buffer PH 6.0 for 100 minutes). After washing with PBS, PAP pen was applied to slides. Slides were incubated in blocking solution (4% Normal horse serum [NHS], 0.1% Triton-100 in PBS) for 60 minutes, followed by incubation in primary antibody solution (rabbit anti-Iba1; WAKO cat #019919741 at 1:1000 concentration in 1% NHS, 0.1% triton-100 in PBS) overnight at 4°C. Slides were then washed in PBS and 0.1% tween, and were incubated in secondary antibody solution (biotinylated horse anti-rabbit IgG (H+L); vector BA-1100 at 1:250 concentration in 4% NHS and 0.4% triton-100 in PBS) 60 minutes at room temperature. Slides were washed in PBS and endogenous peroxidases were blocked by incubation in hydrogen peroxide for 30 minutes. After washing in PBS, ABC solution (Vectastain ABC kit PK-6100) was applied for 30 minutes, followed by a PBS wash. 3,3′-Diaminobenzidine [from Vector DAB peroxidase substrate kit SK-4100] was applied to the slides for 10 minutes. Slides were then placed in tap water and ethanol of increasing concentrations (70%, 90%,100%). After treating the tissue with Citrosolve, coverslips were applied using dibutyl phthalate polystyrene xylene mounting medium.


Z-stacked photomicrographs were taken using a 40 × objective lens on a Zeiss Imager A2 microscope via AxioCam MRc5 digital camera and Neurolucida 360 software, with consistent microscope settings and Z-stack parameters. Three slices per animal were taken from between bregma and lambda and were imaged in the retrosplenial, somatosensory, and entorhinal cortices. A total of 669 microglia (randomly selected using coordinates and a random number generator) from 225 photomicrographs (345 microglia from 13 control mice, 324 microglia from 12 treatment mice) were analyzed. All analyses used the same photomicrographs and isolated microglia for direct comparison.

Percent coverage photomicrograph analysis

The steps followed for percent coverage calculations were based on previous studies that used percent or pixel coverage techniques to indicate microglial reactivity. Percent coverage analysis is often referred to as optical/pixel density or intensity of staining/fluorescence. Using ImageJ, raw photomicrographs were converted to 8-bit and the ‘subtract background’ function was applied. Then the photomicrograph was converted to binary and minimal adjustments to the threshold were made so that the binary image best represented the raw data, and any processing artifact was filtered out. The percentage of the image covered by dark pixels was calculated. Data from each mouse included 9 cortical photomicrographs (3 brain slices per mouse, 3 photomicrographs per slice). We employed two approaches for generating percent coverage data: 1) The percent coverage values for the 9 photomicrographs were averaged to obtain a single percent coverage value per mouse, and 2) each of the 9 separate percent coverage values per mouse was retained as individual data points without averaging the values.

Full photomicrograph skeletal analysis

Iba1 staining was analyzed using the skeletal analysis plugin following the protocol previously published. In brief, photomicrographs were pre-processed by converting to 8-bit and applying the FFT bandpass filter in ImageJ. The brightness/contrast of the photomicrograph was then adjusted to best visualize the branches of the microglia. The unsharp mask was then applied to further increase the contrast of the photomicrograph, and the despeckle function was applied to remove pixels/noise. The threshold was then adjusted, and the despeckle, close, and remove outliers functions were applied. The binarized image was then skeletonized. Microglial cell somas were counted manually to obtain a total microglial count per photomicrograph. The total microglial branch length, branch endpoints, and number of branches were calculated across the entire image and then averaged by the number of microglial cells per frame.

Fractal analysis

According to the previously published protocol, randomly selected microglia were isolated from the photomicrographs and underwent fractal analysis. In brief, this involved converting the photomicrograph to binary, creating a region of interest (ROI) that was sized to fit around all microglia in the study to ensure the scale of the isolated microglia was the same. Using a random number generator, coordinates were used to randomly select 3 microglia per image. Using the ROI, each selected cell was removed from the binary image. The paintbrush tool was used to remove any fragments that were not attached to the cell and connect any branches that became fragmented due to image processing, using the original photomicrograph as a reference. Once microglia were isolated, the binary image was converted to an outline. The FracLac plug-in was used to analyze the cells, with ‘box counting’ applied and the ‘grid design Num G’ set to 4. The convex hull and bounding circle of the cell (image 2D) were measured. The fractal dimension (a statistical measure of pattern complexity), lacunarity (a geometric measure of how a pattern fills space), circularity (how circular the microglial cell is), span ratio (longest length/longest width), and density (number of pixels/area) were measured.

Single cell skeletal analysis

The same cells that were isolated for fractal analysis were skeletonized and analyzed with the skeletal analysis plugin in ImageJ. The number of branches, the total branch length, and the endpoints per microglia were calculated per isolated cell using the ‘analyze skeleton function’. The cell body area and perimeter were calculated using the multipoint area selection tool.

Sholl analysis

Sholl analysis was performed in ImageJ on the same isolated cells that were used for fractal and skeletal analysis. A radius was drawn from the center of the cell body to the end of the longest branch to set the upper and lower limit for concentric circle placement. The first circle was set as close to the edge of the cell body as possible to ensure the cell body was not counted as an intercept on the circle. The distance between each circle was set at 5 µm for all cells. The number of times that the microglial branches intercepted each of the circles was calculated.