The cellular mechanism underlying the genesis of the long-lasting alpha-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA)-receptor-mediated excitatory postsynaptic currents (EPSCs) at the mossy fiber (MF)-unipolar brush cell (UBC) synapse in rat vestibular cerebellum was examined with the use of whole cell and excised patch-clamp recording methods in thin cerebellar slices. Activation of MFs evokes an all-or-none biphasic AMPA-receptor-mediated synaptic current with a late component that peaks at 100-800 ms, which has been proposed to originate from an entrapment of glutamate in the MF-UBC synaptic cleft and is generated by the steady-state activation of AMPA receptors. Bath application of cyclothiazide, which blocks desensitization of AMPA receptors, produced a dose-dependent enhancement of the amplitude of the synaptic current (median effective dose 30 microM) and slowing of the rise time of the fast EPSC. N-methyl-D-aspartate-receptor-mediated EPSCs in UBCs were not potentiated in amplitude or time course by cyclothiazide (100 microM). The dose-response relations for the steady-state current evoked by glutamate acting at AMPA receptors in excised outside-out patches from UBC and granule somatic membranes was biphasic, peaking at 50 microM and declining to 50-70% of this value at 1 mM glutamate. When glutamate was slowly washed from patches to simulate the gradual decline of glutamate in the synapse, a late hump in the transmembrane current was observed in patches from both cell types. The delivery of a second MF stimulus at the peak of the slow EPSC evoked a fast EPSC of reduced amplitude followed by an undershoot of the subsequent slow current, consistent with the hypothesis that the peak of the slow EPSC reflects the peak of the biphasic steady-state dose-response curve. Estimates of receptor occupancy and glutamate concentration derived from the ratio of fast EPSC amplitudes, and the amplitude and polarity of the initial steady-state current in paired-pulse experiments, predict a slow decline of glutamate with a time constant of 800 ms, declining to ineffective concentrations at 5.4 s. Manipulation of cleft glutamate concentration by lowered extracellular calcium or delivery of brief stimulus trains abolished the slow EPSC and restored the undershoot to paired stimuli, respectively, in a manner consistent with a prolonged lifetime of glutamate in the cleft. The slow component of the EPSC was prolonged in duration by the glutamate reuptake inhibitor L-trans-pyrrolidine-2,4-dicarboxylate, suggesting that glutamate transport contributes to the time course of the synaptic current in UBCs. The data support the notion that the MF-UBC synapse represents an ultrastructural specialization to effectively entrap glutamate for unusually prolonged periods of time following release from MF terminals. The properties of the postsynaptic receptors and constraints on diffusional escape of glutamate imposed by synaptic ultrastructure and glutamate transporters act in concert to sculpt the time course of the resulting slow EPSC. This in turn drives a long-lasting train of action potentials in response to single presynaptic stimuli.