Office of Biological and Environmental Research
DOE Low Dose Radiation Research Program Workshop III

Abstract

____________________________________________________________

Quantitative Analysis of Connexin Expression in Cultured Colonies

Authors: B. Parvin, Q. Yang, R. L. Henshall-Powell and M.H. Barcellos Hoff

We are studying the effects of ionizing radiation on the signaling between human mammary epithelial cells and the extracellular microenvironment. To do so we use an assay based on the ability of the cells to organize into three-dimensional acini when embedded into an extracellular matrix. Although tumorigenic and non-tumorigenic mammary epithelial cells are nearly indistinguishable when cultured as monolayers, their biological character readily diverge when tissue-specific morphogenesis is analyzed. Non-malignant human mammary epithelial cells (HMEC) cultured within a reconstituted basement membrane organize into acinar-like structures with polarity; in contrast, breast cancer cells form disorganized aggregates similar to tumors in vivo. These are studied using immunofluorescence and confocal microscopy, which permits the reconstruction of 3-dimensional organization.

Automatic detection of cell structures and localization of protein expression from volumetric dataset is an important step in large scale analysis of cultured colonies and their intercellular interactions. Detection of an individual nucleus reveals morphological features like size and shape, can be used to map the multicellular organization of each colony, and enables localization of intercellular signaling components as a function of treatment. The focus of this initial study is to determine the frequency of gap junction protein complexes. Connexins are a family of proteins associated with gap junctions that modulate the transfer of molecules between cells. Connexins-43 and -32 localized as distinct aggregates between cells of HMEC acini.

We found that automated analysis and counting of connexin aggreates was hampered by abundant speckle noise, which has signature similar to connexin in the volumetric dataset. The detection of individual nuclei provides the necessary “context” to filter speckle noise and enables automatic counting and characterization of connexin expression. In general the nucleus of a cell is ellipsoidal, but neighboring nuclei may overlap, and thus making delineation a necessary component. Our experience indicates that detection of nuclei in 2D is more complex due to inherent lack of 3D information, however, a more efficient techniques is needed to detect blobs in 3D. In general, analysis of these images is complex due to the fact that nuclei of interests (1) do not respond uniformly to the fluorescent compounds, (2) may have many internal substructures, and (3) overlap each other as a result of cell division. The first step of our algorithm is extraction of elliptic features. A multiscale representation of the image is generated and the Hessian is computed to detect and classify each point in the image. If the Hessian is negative (positive) definite then the point is classified as bright (dark) elliptic feature. This classification is then used to group similar features using 3D connected component algorithm. Although false labeling is unavoidable, sufficient

information can be gathered so that a higher level technique can group partial information into a whole. This higher level constraint is expressed in terms of convexity, implemented as a convex hull for improved efficiency, and applied to 3D connected components. The convex hull of a set of points is the smallest convex set that contains the points.

In summary, we have developed a layered computational protocol to segment cultured colonies for simultaneous segmentation of nuclei, characterizing their organization, and mapping inter cellular communication. Preliminary experimental data of connexin expression in human mammary epithelial cells surviving low doses of ionizing radiation and the impact of chronic exposure to the cytokine transforming growth factor $1 will be presented.