Item

Abstract

Studies of protein dynamics by 4D (3D + time) confocal microscopy in vivo are hampered by global cell motion. The time between the acquisitions of the 3D images is in the order of minutes. Therefore, it is not to be expected that the cell as a whole remains fixed in the water basin on the stage. This superimposes a motion on the protein dynamics that has to be removed. We present a robust registration technqiue to align the cell images that does not require the a priori establishment of point-to-point correspondences. Instead, it uses the distribution of the labeled proteins. After correction for the translation, the 3D rotation of the cell is estimated. A robust intrinsic body coordinate system is constructed via the inertia tensor from the instensity distribution. By combining basis transformation to this intrinsic coordinate system, we can calculated [sic! Stefan] the rotation matrix in a conceptual and comutational straightforward manner. We evaluated the performance of this approach in three experiments with human osteaoarcoma cells (U-2 OS), where the nuclear proteins Histon H4 and PML were visualized. The PML is concentrated in several dozen nuclear spots. Expression of Histon H4 results in a total nuclear staining. The registration results for both channels computed independently are very similar. Practically, this means that only the labeled material needs to be observed and still registration of the cell as a whole can be achieved.