Tfm (testicular feminization) mutant mice lack functional androgen receptors. By studying liver tumor development in Tfm mice, we have shown that the greater susceptibility of male mice relative to female mice for liver tumor induction by N,N-diethylnitrosamine is androgen receptor-dependent. C57BL/6J normal and Tfm mutant mice were injected at 12 days of age with N,N-diethylnitrosamine (0.2 mumol/g, i.p.), and liver tumors were enumerated in 50-week-old animals. Normal males averaged 20 liver tumors per animal; Tfm males, 0.7; normal females, 0.6; and Tfm/+ heterozygous females, 1.5. The androgen receptor gene and the Tfm mutation are X chromosome linked. Because of random X chromosome inactivation, hepatocytes from Tfm/+ heterozygous female mice are mosaic with respect to the expression of mutant or wild-type receptors. To determine if testosterone acts directly as a liver tumor promoter, through the androgen receptor in preneoplastic hepatocytes, or by an indirect mechanism, we chronically treated these mosaic female mice with testosterone and measured the androgen receptor content of the resulting tumors. B6C3F1 Tfm/+ mosaic and +/+ wild-type female mice were injected i.p. at 12 days of age with N,N-diethylnitrosamine (0.1 mumol/g) and ovariectomized at 8 weeks of age. Half of the mice of each group subsequently received biweekly s.c. injections of testosterone (0.15 mg per mouse) for 30 weeks. Tumor multiplicity was the same for wild-type and Tfm/+ mosaic females treated with testosterone (31-32 tumors per animal at 38 weeks of age) and was increased relative to females not treated with testosterone (13-17 tumors per animal at 50 weeks of age). Testosterone treatment did not significantly increase the percentage of androgen receptor-positive tumors in Tfm/+ mosaic females: 58% of the tumors from Tfm/+ mosaic females treated with testosterone were receptor positive compared to 48% in Tfm/+ females not treated with testosterone and 92% in wild-type females treated with testosterone. Finally, the number of androgen receptors in the majority of liver tumors examined was greatly decreased relative to the surrounding normal liver tissue. We conclude that liver tumor promotion by testosterone requires a functional androgen receptor in the intact animal. However, this promotion is not cell autonomous; that is, the response of the preneoplastic hepatocyte is not dependent on the expression of functional receptor in the target cell.