In female mammalian cells, dosage compensation for X-linked genes is achieved by the transcriptional silencing, early in development, of many genes on just one of the two X chromosomes. Several properties distinguish the inactive X (Xi) from its active counterpart (Xa). These include expression of Xist, a gene located in the X-inactivation center (Xic), late replication, differential methylation of selected CpG islands and underacetylation of histone H4. The relationship between these properties and transcriptional silencing remains unclear. Female mouse embryonic stem (ES) cells have two active X chromosomes, one of which is inactivated as cells differentiate in culture. We describe here the use of these cells in studying the sequence of events leading to X-inactivation. By immunofluorescent labeling of metaphase chromosome spreads from ES cells with antibodies to acetylated H4, we show that an underacetylated X chromosome appears only after 4 days of differentiation, and only in female cells. The frequency of cells with an underacetylated X reaches a maximum by Day 6. In undifferentiated cells, H4 in centric heterochromatin is acetylated to the same extent as that in euchromatin but has become relatively underacetylated, as in adult cells, by Day 4 of differentiation (i.e. , when deacetylation of Xi is first seen). The overall deacetylation of Xi follows Xist expression and the first appearance of a single, late-replicating X, both of which occur on Day 2. It also follows the silencing of X-linked genes. Levels of mRNA from four such genes, Hprt, G6pd, Rps4, and Pgk-1, had all fallen by approximately 50% (relative to the autosomal gene Aprt) by Days 2-4. The results show that properties that characterize Xi are put in place in a set order over several days. H4 deacetylation occupies a defined place within this sequence, suggesting that it is an intrinsic part of the X-inactivation process. The stage at which a completely deacetylated Xi is first seen suggests that deacetylation may be necessary for the maintenance of silencing but is not required for its initiation. Nor is it required for, or an immediate consequence of, late replication. However, we note that selective deacetylation of H4 on specific genes would not be detected by the microscopical approach we have used and that such selective deacetylation may still be part of the silencing process.