Recently we found that retinoic acid receptors (alpha, beta, gamma) and retinoid X receptors (alpha, beta, gamma) are variably expressed in MCF10A model of breast cancer development and progression. Here we employed this model to assess both in vitro and in vivo sensitivity of cells to retinoids and the role of RARbeta2 in mediating the antitumor potential of retinoids. In vitro, we found that transformation of the benign MCF10A cells into premalignant MCF10AT and malignant MCF10CA1a cells increased their sensitivity to 4-(Hydroxyphenyl)retinamide (4-HPR) and all-trans-retinoic acid (atRA) but not to 9-cis-retinoic acid (9cRA) and LGD1069 and this was associated with loss of induction of RARbeta2 by retinoids. RARbeta2 expression in premalignant MCF10AT cells decreased their proliferating activity and increased their sensitivity to atRA. In vivo, when transplanted into the mammary fat pads of nude mice, MCF10A cells did not grow, MCF10AT cells formed slow-growing tumor nodules, and MCF10CA1a cells were highly malignant, grew quickly and infiltrated the surrounding tissues. Of the retinoids used, only 4-HPR suppressed the growth of slow-growing hyperplastic and premalignant MCF10AT but not of the malignant MCF10CA1a tumor nodules. These data may have clinical implication in selecting women with hyperplastic and premalignant breast lesions that may benefit the most from clinical trials with retinoids.