Fluorescence in situ hybridization (FISH, 15) and high density, non-selective long term perfusion culture were used to study aspects of genetic stability and productivity in recombinant CHO cells. We analysed the distribution and structure of recombinant amplicons in chromosomes of CHO cells used for the production of proteins. In the presence, but not in the absence, of methotrexate (MTX) we found a high proportion of cells (40-60%) with multiple and/or unusually structured and extended chromosome regions containing amplified sequences. Removal of MTX from culture media resulted in the rapid decline in the frequency of cells containing amplified sequences exhibiting multiple and heterogeneous integrations. In cloned lines cultivated in the absence of MTX, a well defined signal motif on a specific chromosome, interpreted as the "master integration" unit, became increasingly abundant over time until almost all cells contained that signal motif. We used fluorescence in situ hybridization (FISH) with chemically modified DNA probes complementary to the integrated sequences to verify the stability of master integrations in cells cultured for extended periods in medium lacking MTX. In a second approach, we studied long-term stability of recombinant sequences during non-selective perfusion culture of CHO populations with non-cloned, MTX resistant heterogeneous populations of cells producing CD4IgG chimeric molecules. Due to the mode of transfection and primary selection the resulting cell lines consisted of subpopulations of cells derived from various independent integration events. The amplification and MTX selection procedure used consecutively generated multiple, structurally different amplicon types in the cell population, thus increasing the degree of heterogeneity. High density perfusion culture of these cells in the absence of MTX was used to maximize growth rates and was thought to select against cells with reduced growth rates, due to the highly amplified state of their introduced DNA, with concomitantly high productivity. However we found no evidence for such a selection; cells showed no reduction in copy number or total loss of amplified sequences at the end of the culture. More significantly, specific productivity of these cell lines grown under non-selective conditions did not decrease over the 100 days observation period.