A nucleoid segmentation method robust to varying nucleoid number

Escherichia coli is an established model for the study of protein aggregation in vivo, a process that is related to its aging. These aggregates have a predisposition for locating at the cell poles, which, following cell division, results in asymmetric damage distribution by the two poles of the daughter cells. Such preference for polar localization is due the occlusion caused by the nucleoid at midcell. To study this process, it is necessary to correlate the spatial location of the protein aggregates and nucleoids during a cell life cycle. For this, and while neither structure has clear borders, it is necessary to perform their segmentation from fluorescence microscopy images. Here, we propose an adaptation of the method used in automatic Drusen detection in retinal images for detecting and segmenting DAPI-stained nucleoids at each stage of the cell cycle. We use the GPL algorithm to detect the number of nucleoids inside each cell. For the segmentation step, according to the number of nucleoids present in the cell, we use either one or two 3-dimensional, modified Gaussian functions as a point spread function model. By setting the amplitude profile parameter with a value equal to 10, the segmentation threshold is directly obtained from the value of z0. The method was applied to images of cells at different temperatures. Table 1 shows the extracted relative nucleoid major axis length for cells with one and with two nucleoids, in each temperature condition. From Table 1, the mean relative nucleoid major axis length decreases with increasing temperature for both cells with one and with two nucleoids, as expected by visual inspection of the images. Also, the positions of the borders of the nucleoid detected by our method, for cells with one nucleoid, are consistent with regions of anisotropies in aggregates motion reported in our previous studies. Additionally, the locations of the nucleoids for both cells with one and with two nucleoids are in agreement with the observed spatial distribution of the protein aggregates, which remain outside the nucleoid region according to our previous studies. These results demonstrate that the proposed method is able to differentiate between one and two nucleoids, perform their segmentation efficiently, and extract with precision the nucleoid(s) size along both the major and minor cell axis.