Using multi-channel level sets to measure the cytoplasmic localization of HCMV pUL97 in GFP-B-gal fusion constructs

- Quantification of intracellular protein distribution of the HCMV kinase pUL97.
- Automated image segmentation by the use of watershed transformation and fast marching level sets.
- Strong correlation of automated measurements with manual observations.
- Different efficiencies of nuclear translocation between two pUL97 isoforms.

Human cytomegalovirus UL97-encoded protein kinase (pUL97) phosphorylates cellular and viral proteins and is critical for viral replication. To quantify the efficiency of nuclear translocation and to elucidate the role of putative nuclear localization signal (NLS) elements, immunofluorescence analysis of different pUL97 expression constructs was performed. Since manual quantitation of respective expression levels lacks objectivity and reproducibility, and is time-consuming as well, a computer-based model is established. This model enables objective quantitation of the degree of cytoplasmic localization λ. To determine the degree of cytoplasmic localization of different pUL97-GFP-β-gal fusion proteins automatically, a multi-channel segmentation of the nucleus and cytoplasm of transfected HeLa cells is performed in DAPI and GFP micrographs. A watershed transform-based segmentation scheme is used for the segmentation of the cell nuclei. Subsequently, the cytoplasm is segmented using a fast marching level set method. Based on the segmentation of cell nuclei and cytoplasm, λ can be determined for each HeLa cell by quantitation of the ratio of average signal intensity outside and inside the nucleus. The degree of cytoplasmic localization of an individual construct is then determined by evaluating the average and standard deviation of λ for the corresponding HeLa cells. Evaluation demonstrates that nuclear transport of pUL97 is a multilayered mechanism resulting in different efficiencies of nuclear translocation between a small and a large isoform and objective quantitation of the cytoplasmic localization is possible with a high accuracy (96.7% and 94.3%).