Automated tracking of fluorescently labeled nuclei in 4D data sets has been used for several years for cell lineage tracing (Bao et al., 2006) and mapping gene expression patterns in C. elegans (Murray et al., 2006). Fully automated tracking of nuclei during late morphogenetic movements (i.e. above 350 cells) has remained challenging owing to the increasing density of nuclear packing in the later embryo.
We have developed a new tool for tracking labeled nuclei that avoids the increasing error rates associated with fully automated lineaging. Our approach is semi-automated: starting with a curated map of all nuclei in a 4D frame at any given time point, our software conducts local searches for nuclei at the next time point. Any nucleus that appears to have moved less than a threshold distance (0.75 of its radius) is assumed to be the same nucleus as in that position at time t. Nuclei that have moved greater than the threshold distance, or new nuclei arising by division, are flagged to be curated manually.
Using this approach we can trace all nuclei in individual embryos up to the 1.5-fold stage (670 nuclei; Figure 1), after which muscle movements make tracking difficult. It takes approximately 8 hours to track nuclei up to the 350-cell stage, and 80-100 hours to the 1.5-fold stage. Our software generates visualizations of the 4D data sets that can plot nuclear velocities and trajectories over chosen intervals. Our software has been used on 4D movies of histone-GFP (zuIs178) labeled nuclei acquired on a Zeiss LSM510 confocal and saved in the Zeiss .lsm format. We have been able to track nuclei in other confocal 4D data sets, the limiting factor being blurring of nuclear GFP signal at the top of a z-stack as nuclear density increases. In principle our approach can be applied to any sample in which many objects are being tracked over time. Our code is available at http://sourceforge.net/projects/nucleitracker4d/. NucleiTracker4D v2.0 requires a computer with at least 4 Gb of RAM as well as Matlab and its image processing toolbox. We welcome feedback from the community.