Genotype/phenotype analyses of low frequency tumor cells using computerize image microscopy.

Abstract

Techniques to identify low frequency (i.e., 10(-4)-10(-5)) tumor cells in bone marrow and peripheral blood of cancer patients provide opportunities for early detection of disseminated disease and characterization of the properties of cells released from the tumor. Low frequency epithelial cells in marrow and blood can be detected using immunophenotypic markers directed against intracellular and/or cell surface antigens. However, nonspecific phenotypic labeling may compromise the ability to discriminate tumor from nontumor cells. We describe optimization of slide-based approaches that facilitate identification and subsequent molecular cytogenetic characterization of rare tumor cell populations in hematopoietic tissues. Colon tumor cells seeded in a hematopoietic background provided a model system to optimize methodologies that are applicable to detection and quantification of micrometastases in clinical specimens. Mixtures of cytogenetically aberrant epithelial cells and hematopoietic cells on slides were labeled with an anti-cytokeratin 20 (anti-CK20) antibody that recognizes > 90% of colon adenocarcinomas. Computerized image analysis was used to record the location of the immunofluorescent cells on microscopic slides. Cells on slides were then hybridized using fluorescence in situ hybridization (FISH) with repeat-sequence DNA probes to detect aneusomies. Previously discriminated epithelial cells were relocated for molecular cytogenetic characterization. Tumor cells present at frequencies approximating 5 x 10(-5) were discriminated on the basis of immunofluorescence and cell size. A low frequency (4 x 10(-4)) population of normal hematopoietic cells also labeled with anti-CK20 (data not shown). This strategy of sequential immunophenotyping and molecular cytogenetic analyses may be useful to discriminate tumor from nontumor cells in cancer patients with micrometastatic disease.