The basis of many anti-cancer therapies is the use of genotoxic agents that damage DNA and thus kill dividing cells. Agents that cause cells to override the DNA-damage checkpoint are predicted to sensitize cells to killing by genotoxic agents. They have therefore been sought as adjuncts in radiation therapy and chemotherapy. One such compound, caffeine, uncouples cell-cycle progression from the replication and repair of DNA [1] [2]. Caffeine therefore servers as a model compound in establishing the principle that agents that override DNA-damage checkpoints can be used to sensitize cells to the killing effects of genotoxic drugs [3]. But despite more than 20 years of use, the molecular mechanisms by which caffeine affects the cell cycle and checkpoint responses have not been identified. We investigated the effects of caffeine on the G2/M DNA-damage checkpoint in human cells. We report that the radiation-induced activation of the kinase Cds1 [4] (also known as Chk2 [5]) is inhibited by caffeine in vivo and that ATM kinase activity is directly inhibited by caffeine in vitro. Inhibition of ATM provides a molecular explanation of the attenuation of DNA-damage checkpoint responses and for the increased radiosensitivity of caffeine-treated cells [6] [7] [8].

The ATM protein kinase regulates the cell's response to DNA damage by regulating cell cycle checkpoints and DNA repair. ATM phosphorylates several proteins involved in the DNA-damage response, including p53. We have examined the mechanism by which ATM regulates p53's transcriptional activity. Here, we demonstrate that reintroduction of ATM into AT cells restores the activation of p53 by the radio-mimetic agent bleomycin. Further, p53 activation is lost when a kinase inactive ATM is used, or if the N-terminal of ATM is deleted. In addition, AT cells stably expressing ATM showed decreased sensitivity to Ionizing Radiationinduced cell killing, whereas cells expressing kinase inactive ATM or N-terminally deleted ATM were indistinguishable from AT cells. Finally, single pointmutations of serines 15, 20, 33 or 37 did not individually block the ATM-dependent activation of p53 transcriptional activity by bleomycin. However, double mutations of either serines 15 and 20 or serines 33 and 37 blocked the ability of ATM to activate p53. Our results indicate that the N-terminal of ATM and ATM's kinase activity are required for activation of p53's transcriptional activity and restoration of normal sensitivity to DNA damage. In addition, activation of p53 by ATM requires multiple serine residues in p53's transactivation domain. Oncogene (2001) 20, 5100 ± 5110.