The keratin-filament cycle of assembly and disassembly
Final Report Abstract
Towards an extensive spatiotemporal analysis of the keratin cytoskeleton, methods for the analysis of its structural and dynamic properties were developed. First, a novel model for signal and noise transfer in CLSM has been developed which permits the simulation of the imaging process and thus the generation of realistic synthetic image data that enable the quantitative evaluation of the developed image analysis methods. The quantitative evaluation based on simulated data is complemented throughout the project with a qualitative assessment based on visual inspection and with the verification of consistency with results obtained from other experiments. For image enhancement, a novel denoising method was developed, because analyzing the keratin dynamics was found to be very sensitive to noise. The method was tailored towards curvilinear structures in CLSM by employing a curvelet transform-based denoising scheme so that the strong image noise can effectively be reduced while not blurring the filaments. Analysis of the structural properties of keratin filament networks was realized via segmentation. Segmentation was done following the methodology of line enhancement filtering using flux and centerline extraction. The result is a graph representation of the extracted centerline. The analysis of keratin dynamics was addressed by motion and turnover analysis. Concerning motion analysis, different kinds of motion caused by different phenomena had to be separated to not confound them. Therefore, a framework was introduced that allows to analyze cell locomotion, cell contraction and local filament motion separately. Global cell movement during recording was compensated for by rigid preregistration. Shape changes over time were optionally compensated for by means of cell shape normalization. Shape normalization transforms the cell's shape in each frame of a time-series into an idealized circular unit cell. Therefore shape normalization also allows for averaging of analysis results over multiple cells yielding model-like cell behavior. In view of the high variability of cells, this is a huge advantage. For the analysis of local filament motion, a novel motion analysis method was formulated as registration method and solved using a graph-cut based optimization algorithm. Application to keratin filament time-series showed a consistent pattern of steady filament motion directed from the cell periphery towards the interior. For complete characterization of the keratin dynamics, a novel approach to turnover analysis was proposed. The goal was to identify keratin turnover in terms of locations and amount of filament assembly and disassembly. In contrast to the state-of-the-art method for turnover quantification (fluorescence recovery after photobleaching (FRAP)), turnover can now be localized and quantified in an entire cell, and, in one and the same cell before and after drug treatment. The novel method is based on the computation of the keratin flux density from the filament motion vector field. A mass balance computation reveals the sources and sinks of the keratin flux density that correspond to the locations of keratin filament polymerization and depolymerization. Application of the novel method to keratin filament time-series revealed that keratin filament polymerization primarily takes place in the cell periphery while depolymerization mainly occurs in the central part of the cell where the network surrounds the nucleus. In two large-scale experiments, the proposed image analysis methods were applied to CLSM time-series of vulva carcinoma-derived cells. The results confirmed the keratin cycling model and demonstrate the use for cell biology research in practice.
Publications
- 3D Segmentation of Keratin Intermediate Filaments in Confocal Laser Scanning Microscopy. Proceedings of the 33rd Int. Conf. of IEEE Eng Med Biol Soc., pp. 7751-7754, 2011
Gerlind Herberich, Reinhard Windoffer, Rudolf Leube and Til Aach
(See online at https://doi.org/10.1109/iembs.2011.6091910) - Flux-Based 3D Segmentation of Keratin Intermediate Filaments in Confocal Laser Scanning Microscopy. Proceedings of the 9th IEEE Int. Symposium on Biomedical Imaging, pp. 166-169, 2012
Gerlind Herberich, Andreas Friedrich, Til Aach, Reinhard Windoffer and Rudolf E. Leube
(See online at https://doi.org/10.1109/ISBI.2012.6235510) - Signal and Noise Modeling in Confocal Laser Scanning Fluorescence Microscopy. MICCAI 2012, LNCS 7510, Part I, pp. 381-388, Springer, 2012
Gerlind Herberich, Reinhard Windoffer, Rudolf E. Leube and Til Aach
(See online at https://doi.org/10.1007/978-3-642-33415-3_47) - Measuring the regulation of keratin filament network dynamics. Proceedings of the National Academy of Sciences of the United States of America, 110(26):10664-9, Jun 25; 2013
Marcin Moch, Gerlind Herberich, Til Aach, Rudolf E. Leube and Reinhard Windoffer
(See online at https://dx.doi.org/10.1073%2Fpnas.1306020110) - Quantification of keratin network dynamics from confocal microscopy images. The 10th Conference Focus On Microscopy, pp. 159, 2013
Marcin Moch, Gerlind Herberich, Til Aach, Reinhard Windoffer and Rudolf E. Leube
