Activation of caspases 3 and 9 is thought to commit a cell irreversibly to
apoptosis. There are, however, several documented situations (e.g., during
erythroblast differentiation) in which caspases are activated and caspase s
ubstrates are cleaved with no associated apoptotic response. Why the cleava
ge of caspase substrates leads to cell death in certain cases but not in ot
hers is unclear. One possibility is that some caspase substrates generate a
ntiapoptotic signals when cleaved. Here we show that RasGAP is one such pro
tein. Caspases cleave RasGAP into a C-terminal fragment (fragment C) and an
N-terminal fragment (fragment N). Fragment C expressed alone induces apopt
osis, but this effect could be totally blocked by fragment N. Fragment N co
uld also block apoptosis induced by low levels of caspase 9. As caspase act
ivity increases, fragment N is further cleaved into fragments N1 and N2. Ap
optosis induced by high levels of caspase 9 or by cisplatin was strongly po
tentiated by fragment N1 or N2 but not by fragment N. The present study sup
ports a model in which RasGAP functions as a sensor of caspase activity to
determine whether or not a cell should survive. When caspases are mildly ac
tivated, the partial cleavage of RasGAP protects cells from apoptosis. When
caspase activity reaches levels that allow completion of RasGAP cleavage,
the resulting RasGAP fragments turn into potent proapoptotic molecules.