We report that a small population of neurons expresses both choline acetyltransferase and classical estrogen receptor immunoreactivity and they are found primarily in the bed nucleus of the stria terminalis. In short-term ovariectomized ageing mice (24 months, n = 5) there were 41.0 +/- 4.1% fewer of these double-labeled cells than in young (five months, n = 5) short-term ovariectomized C57BL/6J mice. To study cholinergic neuron estrogen responsiveness, young mice (n = 8) were ovariectomized at puberty (five weeks). After three months half of the mice (n = 4) were given physiological levels of 17beta estradiol for 10 days. Bed nucleus double-labeled neurons increased by 32.9% (P < or = 0.003) in the young mice given estrogen. In a gel shift assay, double-stranded oligonucleotides with putative estrogen response elements from the choline acetyltransferase gene were used as competitors against estrogen receptor binding to consensus estrogen response elements. A sequence with 60% homology to the vitellogenin estrogen response element was found to compete at 500- and 1000-fold excess. Young mice (five months) with ovaries demonstrated significantly (P < or = 0.04) better performance in the spontaneous alternation T-maze test than did old (19 month) mice with ovaries (young = 66.3 +/- 3.3% correct choices; vs old = 55.0 +/- 4.0% in old mice with ovaries). Young mice (five months old), ovariectomized for one month and treated with estrogen, showed significantly more spontaneous alternation than ovariectomized controls (69.1 +/- 2.8% vs 58.3 +/- 3.9%; P < or = 0.04). Estrogen also increased spontaneous alternation in old, short-term ovariectomized mice (61.5 +/- 2.7% vs 48 +/- 3.3%; P < or = 0.005). In either young or old ovariectomized mice, estrogen increased spontaneous alternation to levels seen in young animals with ovaries. Estrogen increases the number of choline acetyltransferase-immunoreactive and choline acetyltransferase/estrogen receptor-immunoreactive cells in old or young mice lacking estrogen, and enhances working memory in old or young mice lacking estrogen. Our data suggest that estrogen may act at the level of the choline acetyltransferase gene, but in view of the limited distribution of cholinergic cells expressing the classical estrogen receptor, it is unlikely that these cells can account for a memory enhancing effect of estrogen replacement.