Abstract： Objective To explore the intervention effect and the underlying molecular mechanism of 3-Methyladenine on behavioral damage of neonatal rat with prolonged seizures. Methods Forty-five 6-dayold SD rats were randomly ( random number) divided into the recurrent prolonged neonatal seizure group ( RS group), the 3-MA-treated seizure group and control group. The volatile agent flurothyl was used to induce 30 min seizure attack. At postnatal day 6 (P6), recurrent seizures were induced once daily for successive 6 days in both RS group and 3-MA group. In 3-MA group, 3-MA (2 μL) was injected daily before seizures induced. Neural-behavior changes were observed with double-blind method including swimming development, open field test and Morris water maze analysis. Bcl-2 and Beclinl protein levels in hippocampus were detected by western blot method at P50. Results The total scores of swimming behavior in RS rats were decreased significantly compared with those of control and 3-MA rats ( control: 7. 44 ±1. 13, RS: 5.06±1.63, 3-MA: 7.33 ±1.08, F=16.19, P＜0. 01) . The start-latency time of open filed behavior in RS rats ( 13. 33 ±6. 69) were increased significantly compared with that of control (7. 11 ±2. 37) and 3-MA rats (9. 91 ±4. 23) (F=4. 39, P＜0. 05). The escape latency was significantly longer in rats of RS group than that of control and 3-MA rats on the 4th and 5th days (P ＜ 0.05). The level of Bcl-2 in hippocampus of RS group (0. 587 +0. 139) were significantly lower than that of control (0. 782 +0. 083) and 3-MA groups (0. 799 + 0. 163) (F =4. 7 1, P ＜ 0. 05 ). There were no significant differences in the level of Beclin1 protein in hippocampus among the three groups ( F =0. 27, P ＞ 0. 05 ). Conclusions Pretreatment with autophagy inhibitor 3-MA in acute phase of neonatal seizures could significantly improve neurobehavioral capacity, which might be associated with the increased in the level of Bcl-2 protein in hippocampus.