Spontaneous disappearance is, unfortunately, a rare event in oncology, but neuroblastomas are more likely to play this trick than other tumours. What makes them regress? Paradoxically, the overexpression of two oncogenes — HRAS and TRKA — has been associated with a good prognosis, but the mechanism of regression has remained a mystery. Possibilities include apoptosis and differentiation, but Chifumi Kitanaka and colleagues now bring a third possibility onto the scene: non-apoptotic cell death.

Japan has a population-screening programme for neuroblastoma in infants, which permits the detection of neuroblastomas that might otherwise spontaneously regress. The authors first compared tumour samples from mass-screened patients with clinically detected tumours from older children who had advanced-stage neuroblastomas. Patches of HRAS staining occurred more frequently in mass-screened tumours than in clinically detected tumours, and these patches frequently colocalized with areas of degenerating cells. However, these regions did not stain for two classical apoptotic markers — active caspase-3 or fragmented 3′ DNA ends (TUNEL assay). Instead, staining with the periodic-acid–Schiff reagent and electron-microscopic analysis hinted that the cells might be eating themselves (autophagy).

The authors then turned to neuroblastoma cell lines to determine whether HRAS expression could kill these cells by a non-apoptotic mechansism. Expression of either wild-type HRAS or RASV12, a constitutively active HRAS mutant, caused the cells to round up and fragment, whereas expression of an inactive HRAS mutant had no effect. Furthermore, the degenerating cells looked quite different from those that were induced to apoptose by stauropsorine or serum withdrawal. Again, no DNA fragmentation was apparent in TUNEL assays, and electron microscopy revealed increased numbers of lysosomes, characteristic of autophagic degeneration.

Is caspase activation required for this peculiar form of HRAS-mediated cell death? Caspase inhibitors did not prevent HRAS-induced cell death, but did block staurosporine-mediated apoptosis. Furthermore, poly(ADP-ribosyl) transferase (PARP), which is invariably attacked by active caspases, was not fragmented in RAS-expressing neuroblastoma cells. Finally, overexpression of the anti-apoptotic protein BCL-XL did not block HRAS-mediated cell death, but did prevent staurosporine-mediated death. Active HRAS, then, can kill cells by a mechanism that is distinct from apoptosis.

Expression of both HRAS and the gene for the nerve growth factor (NGF) receptor TRKA is a better indicator of good prognosis than either gene alone, so the authors wanted to know whether TRKA could augment HRAS-mediated cell death. NGF increased the proportion of HRAS-mediated cell degeneration, but only when TRKA was overexpressed. The magician has an assistant.

Can we learn some magic tricks from this study? If there are other ways to activate this apparently autophagic pathway, perhaps one day we'll be able to make neuroblastomas — and maybe other tumours, too — disappear even if they don't express favourable markers.