Final Report Summary - OMFLICSS (Functional long-term imaging of single bouton-spine pairs during optogenetically controlled synaptic plasticity) The brain has to process and store a large amount of information throughout life and many memories are preserved for decades. Yet, the brain has to forget in order to separate important information from less important ones. Thus, some memories are stable throughout the entire lifetime of the organism, whereas other memory traces are soon erased. At the same time, the brain is highly plastic and therefore can quickly adapt to changes in the organism's environment. Long-term synaptic plasticity has been put forward as a mechanism underlying the processing and storage of information with long-term potentiation (LTP) and long-term depression (LTD) strengthening or weakening synapses, respectively. However, it is unclear whether synapses can maintain their strength at a specific level over the time scales of memory and how plasticity affects stability of single synapses.Intracellular recording techniques such as whole-cell patch clamp can be used to measure the strength of unitary connections with high precision, but last only for several hours, eventually killing the cell under scrutiny. Time scales relevant for long-term memory can be assessed by extracellular (field) recordings, but these report average responses of many cells and hundreds of synapses. Due to these methodological limitations, we do not know whether the strength of an individual synapse drifts over time, or how specific activity patterns affect the long-term survival of a synapse.In recent years, with the advent of two-photon microscopy, evidence has accumulated that synapses in the living brain are constantly undergoing morphological modifications including synapse elimination and de novo formation. Spine turnover has been especially well characterised in the rodent neocortex. Here, cortex-dependent learning tasks promote the stabilisation of existing spines but also de novo spine formation, suggesting that learning and memory directly links to structural changes at single synapses. However, it is less clear how (electrical) activity relates to synapse morphology at time scales relevant for memory formation.To investigate identified synapses over extended time periods, we developed an optogenetic, non-invasive experimental setup enabling us to manipulate and measure plasticity at single synapses. Functional synapses between hippocampal Schaffer collateral axons expressing a light-gated ion channel (channelrhdodopsin-2) and postsynaptic CA1 pyramidal neurons were identified by two-photon imaging of evoked spine calcium transients using the genetically encoded calcium indicator GCaMP3. In addition, we visualised morphology of presynaptic terminals by expressing a red-fluorescent protein (tdimer2) in the stimulated axons. Spine morphology in postsynaptic CA1 pyramidal neurons was observed by expressing a cyan fluorescent protein (mCerulean).Investigation of LTD - a correlate of memory-deletion - at single synapses revealed that they are frequently removed from the network within days following induction of depression. In addition, life expectancy of depressed synapses was dependent on their strength at the time of LTD induction. Stronger synapses were more resistant to elimination than weak synapses. These findings suggest that the strength of a single synapse is not simply reduced to a certain level which remains constant over time as may be inferred from recordings of large populations. Rather, synapses are either completely removed or they recover. As a consequence, information flow in the network is re-directed by changing the synaptic wiring pattern. Synapses that are not relevant in the context of a memory trace get removed while those that are meaningful are preserved. We also show that formation of new synapses is not affected by LTD and therefore, while some routes of information flow get eliminated new routes may still get established. In addition to this homosynaptic form of LTD, we found a heterosynaptic component that lead to depression of nearby synapses. Spines neighboring depressed synapses on the same dendrite experienced depression although they were not active during the LTD induction protocol. This heterosynaptic effect was limited to the dendrite where synaptic transmission occurred during LTD-induction. Single dendritic branches rather than single synapses may therefore serve as the unit of information processing and storage.In summary, we analysed plasticity at single, identified synapses for the first time at time scales relevant for learning and memory. Our findings show that synaptic communication is regulated in a digital rather than analog way in the hippocampus. Synaptic connections which are rarely used may be broken up when they get depressed, leaving space for establishment of new ones.