Genetic modification of embryonic stem (ES) cells provides a powerful system for the correlation of in vitro and in vivo gene function, an issue relevant to transplantation, gene therapy and tissue engineering. ES cells also present unique obstacles to genetic manipulation, particularly with regard to epigenetic modulation of transgene expression. Strategies used to circumvent problems associated with methylation, histone deactylation, and chromatin remodeling include modification of vector sequences, insulators, and targeting transgenes to sites of “open” chromatin structure. Systems combining gene targeting with non-viral promoters have shown particular promise. In these studies, a variety of transgenes were introduced into Ainv15 ES cells and doxycycline (dox)-induced expression evaluated by flow cytometry and quantitative PCR. Ainv ES cells express a reverse tetracycline transactivator (rtTA) from the ROSA26 locus and allow Cre-mediated insertion of transgenes into the HPRT locus adjacent to a tetracycline response element (TRE). As expected, transgene expression in undifferentiated ES cells could be regulated by dose and time of exposure to dox. However, expression levels varied dramatically for different transgenes; variability was transgene-specific and did not directly correlate with transgene size or the presence of YFP sequences. In addition, regardless of initial expression level, all transgenes were transcriptionally silenced during ES cell differentiation as embryoid bodies (EBs). Distinct patterns of silencing occurred when cells were induced for 24 hours immediately preceding analysis or continuously from the start of EB differentiation. In both cases, a pronounced loss of expression occurred between days 4–7; however, EBs treated with dox for 24 hours maintained a population of cells capable of maximal induction through day 6, whereas continuously treated cells showed a marked decline in this population by day 4. These results suggested two distinct processes: a primary loss of inducibility independent of dox exposure and a secondary refractoriness of expressing cells to reinduction. The reproducible patterns further suggested that loss of TRE activity or general silencing of the HPRT locus were responsible. To distinguish between these possibilities, expression of HPRT and NeoR, which flank transgenes in targeted Ainv cells, were evaluated. In contrast to the loss of transgene expression, HPRT and NeoR expression remained stable or increased between EB day 3–7, indicating that silencing was not due to general chromatin remodeling within the locus. Expression of rtTA was also stable, excluding the possibility that silencing of the ROSA26 locus was responsible. We conclude that silencing in this system is due to a loss of TRE activity between EB days 4–7, and speculate that inhibitory factors produced by ES cells at this stage of differentiation are responsible. The findings underscore the complexities of gene regulation in ES cells, and raise the possibility that the capacity of cells to express transgenes may in part reflect intrinsic cell potential. In addition, if cells of particular lineages produce factors that interfere with TRE activity, there are obvious implications for using the Tet-On system for inducing gene expression in those lineages.