We compared two methods originally devised to purify cytoplasmic granules from granulocytes for their capacity to produce cytotoxic granules from natural killer cell lines, suitable for use in target cell apoptosis assays. Both methods utilised nitrogen cavitation to efficiently lyse cells, followed by density gradient fractionation on Percoll to separate the granules from other organelles and granule debris. The first method, originally described by Millard and colleagues, employed DNase I to reduce the viscosity of the initial cell lysate, but the resulting granule fractions were found to contain residual nuclease activity that made them unsuitable for use in apoptosis assays that measure DNA fragmentation. An alternative method described by Borregaard and colleagues utilised a cell relaxation buffer without DNase I. Cytotoxic granules isolated from the NK tumor cell line YT by this protocol were localised predominantly to the densest Percoll fractions, with a density of approximately 1.13 g/ml. These granule fractions were rich in perforin and enzymatically active granzyme B, and induced potent Ca(2+)-dependent lysis and DNA fragmentation of Jurkat cells. Corresponding fractions from non-cytolytic cells, or YT granule extracts incubated with EGTA were unable to mediate significant target cell damage. Cytotoxic granule extracts purified through the Borregaard method were therefore free of nonspecific nuclease activity, and most suitable for studying the mechanism of target cell death induced through the perforin/Ca(2+)-dependent granule pathway.