Abstract
Dexamethasone (DEX) promotes hepatocyte differentiation, the formation of bile canaliculus-like intercellular spaces, and the appearance of connexin32 (Cx32) between adjacent fetal hepatocytes. Although the microscopic images were qualitatively and subjectively examined by pathologist, we quantitatively analyzed the intensity of Cx32 in cultured rat hepatocytes using the image processing software, ImageJ, and compared with our macro-routine. Livers were extracted from fetal rats on gestational day 17, and hepatocytes were separated by collagenase digestion and low centrifugation. Cells were cultured in base or DEX-supplemented medium and fixed at 1, 2, or 3 days after the start of the culture. After fixation, cells were stained using the fluorescein-labeled antibody method for Cx32 and DAPI (diaminophenylindole) staining for nuclei. The quantitative index of Cx32 (QI Cx32) was defined as the area (number of pixels) of Cx32s/cell. For each image, it was automatically processed to measure the area (pixels) of Cx32 and to count the nuclei. Images of Cx32 and nuclei are captured from microscopic fields with color filters, green for Cx32 and red for nuclei, for cell counting using the image processing software, ImageJ. Although particle (Cx32) counting can be done automatically using the ImageJ command of particle analysis, the process requires the image to be a “binary” one. A histogram equalization might be unnecessary, however, it made it easier to find and compare threshold points. The threshold values we used were 200 in a range of 0 to 255 for a Cx32 image and 30 for a nucleus image. A command can be used to segment the image into features of interest and the background. The particle analysis function is commonly used to count and measure area of particles and shows numerical results about particles: count, area (pixels), etc. For nucleus images, particles that were larger than 300 pixels were excluded with optional functions. QICx32 were calculated with Excel using the area of Cx32 and the number of cells for each image. The Cx32s per cell for each sample were tallied with mean and SE. The values were compared among different conditions by one-way ANOVA and Tukey's HSD post hoc test. Twenty pairs of original images for Cx32 and nuclei were obtained from 20 microscopic fields of each experimental condition (control and days 1-3). A total of 160 original images were studied for which the image processing time was decreased from 4 h by manual operations to 1 min with ImageJ and our macro-routine. The results were subjectively compatible with the pathology. A quantitative index of the Cx32 pixels/cell was easily obtained by using ImageJ with our macro-routine. Changes of the Cx32 area were compatible with the pathology.
