Apoptosis is characterized by chromatin condensation and DNA fragmentation in the absence of release of cytosolic enzymes such as lactate dehydrogenase (LDH). In contrast, necrosis is characterized by cell swelling, membrane disintegration with cytosolic enzyme release, and absence of chromatin condensation. Staining of cells with Hoechst H33342 dye is a routine method for identifying apoptotic nuclei. However, this process is tedious and prone to individual bias. Therefore, we investigated the suitability of an artificial neural network (ANN) to recognize and distinguish between apoptosis and necrosis. Using a human endothelial cell line (ECV304), we trained an ANN with DNA-stained apoptotic and necrotic nuclei obtained from cells exposed for 8 h to cycloheximide (Chx; 100 microM)/tumour necrosis factor-alpha (TNF; 50 ng/ml) or tert-butylhydroperoxide (t-BH; 2 mM), respectively. After this training step, the ANN correctly assigned necrosis induced by t-BH and apoptosis induced via Chx/TNF, Chx/CD95 activation, Pseudomonas exotoxin A/TNF, or X-rays. In all cases, apoptosis and necrosis as assigned by the ANN correlated with DNA fragmentation and LDH release.
