How memories persist is a fundamental neurobiological question. The most commonly studied physiological model of memory is long-term potentiation (LTP). The molecular mechanisms of LTP can be divided into two phases: induction, triggering the potentiation; and maintenance, sustaining the potentiation over time. Although many molecules participate in induction, very few have been implicated in the mechanism of maintenance. Understanding maintenance, however, is critical for testing the hypothesis that LTP sustains memory storage in the brain. Only a single molecule has been found both necessary and sufficient for maintaining LTP--the brain-specific, atypical PKC isoform, protein kinase Mzeta (PKMzeta). Although full-length PKC isoforms respond to transient second messengers, and are involved in LTP induction, PKMzeta is a second messenger-independent kinase, consisting of the independent catalytic domain of PKCzeta, and is persistently active to sustain LTP maintenance. PKMzeta is produced by a unique PKMzeta mRNA, which is generated by an internal promoter within the PKCzeta gene and transported to the dendrites of neurons. LTP induction increases new PKMzeta synthesis, and the increased level of PKMzeta then enhances synaptic transmission by doubling the number of postsynaptic AMPA receptors (AMPAR) through GluR2 subunit-mediated trafficking of the receptors to the synapse. PKMzeta mediates synaptic potentiation specifically during the late-phase of LTP, as PKMzeta inhibitors can reverse established LTP when applied several hours after tetanization in hippocampal slices or 1 day after tetanization in vivo. These studies set the stage for testing the hypothesis that the mechanism of LTP maintenance sustains memory storage. PKMzeta inhibition in the hippocampus after learning eliminates the retention of spatial memory. Once the PKMzeta inhibitor has been eliminated, the memory is still erased, but new spatial memories can be learned and stored. Similar results are found for conditioned taste aversion when the inhibitor is injected in the insular neocortex. Thus PKMzeta is the first molecule found to be a component of the long-term memory trace.