A selection strategy to obtain cells deficient in poly(ADP-ribose) polymerase was developed based on the fact that treatment with high levels of N-methyl-N'-nitro-N-nitrosoguanidine results in sufficient activation of poly(ADP-ribose) polymerase to cause NAD and ATP depletion leading to cessation of all energy-dependent processes and rapid cell death. In contrast, cells with low levels of poly(ADP-ribose) polymerase should not consume their NAD and might therefore be more likely to survive the DNA damage. Using this approach, we have cloned a number of cell lines containing 37-82% enzyme activity. The apparent decrease in poly(ADP-ribose) polymerase activity is not due to increases in NAD glycohydrolase, poly(ADP-ribose) glycohydrolase, or phosphodiesterase activities. Further characterization of the poly(ADP-ribose) polymerase-deficient cells indicates that they have prolonged generation times and increased rates of spontaneous sister chromatid exchanges.