Processes of functional plasticity such as hippocampal long-term potentiation (LTP) and long-term depression (LTD) are regarded as cellular mechanisms underlying learning and memory formation. LTP and LTD are used as suitable models for the investigation of the latter processes. During the last decades LTP was intensively investigated. However, less is known about LTD and its relation to learning. Here, studies were performed to investigate whether electrically-induced LTD within rat CA1 hippocampal slices in vitro shares common cellular features with LTD in the intact animal, with particular emphasis being placed on mechanisms required for its late maintenance. My initial studies have led to the development of stimulation protocols which were able to reliably induce different forms of LTD in vitro. Depending on the induction protocol, either a transient protein synthesis-independent early-LTD (with duration of up to 3-4 h) or a de novo protein synthesis-dependent late-LTD (lasting for at least 8 h) could be induced in the hippocampal slices in vitro. Both forms required NMDA-receptor activation during their induction. Furthermore, LTD was input-specific, i.e., the expression was shown only by those synapses specifically stimulated by a low-frequency protocol. Thus, phenotypically LTD in vitro was characterized by analog induction properties as LTP. Recently, it was described that the induction of LTP can mark a specifically activated synapse by a synaptic tag to capture synapse non-specific plasticity-related proteins (PRPs) and thus maintaining input-specific LTP for prolonged periods. My studies show that in rat hippocampal slices in vitro, the induction of protein synthesis-dependent late-LTD is also characterized by processes of synaptic tagging and that heterosynaptic induction of either LTD or LTP on two sets of independent synaptic inputs S1 and S2 can lead to late-associative interactions between LTD- and LTP-inputs: early-LTD in a synaptic input S2 was transformed into a late-LTD, if late-LTP was induced in a synaptic input S1 of the same neuronal population within a distinct time interval. The synthesis of process-independent PRPs by late-LTP in S1 was sufficient to transform early- into late-LTD in S2 when process-specific synaptic tags were set. We have named this new late associative property of cellular information processing as cross-tagging , since process-unspecific PRPs can be captured by either LTP- or LTD- synpatic tags thus transforming a normally transient LTP or LTD in a long-lasting form. The tag as well as the PRPs are characterized by a relatively short half-life of several minutes up to a few hours before they degrade most likely by processes such as dephosphorylation. The question now arose whether the tags or better: the tag complex can also be reset in an activity-dependent manner, thus preventing the processing of PRPs with the result of transient short-lasting plasticity. Early-LTP was used to study this and we found that low-frequency stimulation shortly after early-LTP induction (5 min) reset the tag preventing any lasting forms of LTP and thus, preventing the formation of a cellular memory trace. Next, we searched for the possible tag candidate or PRP -molecules. The role of a PKC isoform is widely speculated as a candidate molecule involved in the synaptic tag-complex. So we investigated the putative role of protein kinase M-zeta (PKM zeta) required for the protein synthesis-dependent phases of late-LTP/-LTD, synaptic tagging, or synaptic cross-tagging. PKM zetainhibition, after LTP induction reversed late-LTP maintenance, and subsequently depressed tetanized inputs. In contrast, LTD maintenance was unaffected, but its induction was blocked. PKM zeta inhibition prevented synaptic tagging of LTP, but during cross-tagging, the inhibitor reversed late-LTP, while early-LTD of a second, independent synaptic input was converted into late-LTD. Thus PKM zeta is specific to the synaptic tagging mechanism of LTP, but not LTD. Our data provide evidence that PKM zeta activity has dual functions: (1) it is specifically involved in LTP-maintenance and LTP-tagging, but not in LTD-maintenance and LTD-tagging and (2) it is required for processes necessary for the induction of both LTP and LTD.