Résumé : Using a method based on [3H]tamoxifenaziridine ([3H]TAZ) labeling, sequential immunoadsorption with anti-ER monoclonal antibodies, sodium dodecyl sulfate-polyacrylamide gel electrophoresis(SDS-PAGE) and fluorography, we observed a striking change inthe estrogen receptor (ER) electrophoresis pattern of the transplantable MXT mouse mammary tumor. Early, ER "rich" tumors (approximately 100 fmol/mg prot) displayed classical cytosolic 67 and 50 KDa bands. These bands disappeared in favor of a "cytosolic" 35 KDa band during progression towards undifferentiated ER "poor" tumors (approximately 25 fmol/mg prot). Although we can not rule out that this 35 KDa peptide results from in vivo ER proteolysis, it seems unique in view of the following: 1. It is immunoadsorbed not only by an anti-ER monoclonal antibody (H-222) directed to the hormone-binding domain, but also by an anti-ER monoclonal antibody (H-226) which interacts with an epitope in the A/B region close to the DNA-binding domain and is mainly exposed under activation conditions. 2. It does not bind [3H]estradiol([3H]E2) and a tentative to restore its [3H]E2 binding capacity with calmodulin and ATP was unsuccessful. The observation of similar approximately 35 KDa ERs in the nuclear fraction of early tumor transplants and in control uterus suggests that this peptide is already in an activated form. Structural alterations of ER and/or associated "anchorage" nuclear proteins may beat the origin of its cytosolic localization. Moreover, the fact that the addition of calmodulin and ATP to late MXT transplants cytosols fails to increase their [3H]E2 binding capacity indicates that the low ER content of these tumors does not result from a deficiency in the phosphorylation status of the receptor.