Detection of chromosome aneuploidy in breast lesions with fluorescence in situ hybridization: comparison of whole nuclei to thin tissue sections and correlation with flow cytometric DNA analysis.

We compared flow-cytometric DNA histogram pattern to counts of 4 fluorescent-labelled centromeric probes (chromosomes 1, 7, 8, and 17) in whole nuclei (WN) and in nuclei from the corresponding formalin-fixed deparaffinized thin tissue section (TS) in 25 breast lesions (9 invasive carcinomas, 1 duct carcinoma-in-situ, 5 fibroadenomas, 10 fibrocystic change). In benign lesions, signal gains (i.e., trisomic nuclei) were never observed in greater than 10% of nuclei from either WN or TS preparations. Loss of signal in benign breast lesions, however, varied considerably (0-43%) between individual case and between chromosome probes. The mean incidence of signal loss in WN of benign lesions ranged from 8.9% (chromosome 7) to 14.4% (chromosome 1) of nuclei. These signal loss frequencies exceeded those of benign lymphoid control cells. In three benign lesions, signal loss in WN (with one probe) was observed in at least 25% of nuclei. Signal losses in benign TS, on average, were 50-150% greater than in matched WN preparations (chromosome 1-21.7%, chromosome 7-21.5%). Malignant lesions generally, but not always, displayed fewer monosomic nuclei and more trisomic nuclei in WN compared to TS, compatible with a slicing (i.e., nuclear truncation) artifact. Signal counts in carcinomas correlated well with flow cytometric DNA index; however, they were also characterized by evidence of genetic instability, manifest as signal gains in a subset of nuclei (10-25%) with individual probes in diploid range cases, as well as intratumoral heterogeneity, reflected as discrepancies in probe counts between WN and TS samples. We conclude that signal losses with centromeric probes are largely, but not entirely, explained by nuclear slicing.(ABSTRACT TRUNCATED AT 250 WORDS)