In the second reporting period, we again made major progress towards reaching these highly ambitious goals. Most importantly, we published the first paper on functional electron microscopy (“flash and freeze”) in acute brain slices (Borges-Merjane et al., 2020, Neuron, cover article). Using the novel flash and freeze technique, we were able to show that optogenetic stimulation of hippocampal mossy fibers depleted the pool of docked vesicles in active zones of hippocampal mossy fiber synapses. This suggests that docked and readily releasable pool are overlapping, as often assumed, but never directly shown. The methods paper also provides a detailed recipe how structure – function analysis can be performed at other synapses. Second, using paired recordings between mossy fiber terminals and postsynaptic CA3 pyramidal cells, we found that posttetanic potentiation (PTP), a major form of presynaptic plasticity at this synapse, is not primarily caused by increased release probability, as previously assumed, but rather by an augmented size of the readily releasable vesicle pool (Vandael et al., in revision). Following depletion, the pool not only recovered back to the control value, but rather became larger in comparison to control conditions. This pool overfilling may contribute to posttetanic potentiation observed in our paired recording experiments. Based on this observation, we suggest a new mechanism of short-term memory in the hippocampus, in which information is stored as “pool engrams”. We also found that PTP has anti-associative induction properties and a uniquely low induction threshold. Third, we extensively characterized the activity of granule cells in vivo in awake mice during a spatial navigation task. We discovered that granule cells fire action potentials only very sparsely, but if they fire, they often generate bursts and higher order activity patterns, termed “superbursts”. Our data base, comprised of 73 morphologically identified granule cells, represents the largest data set of in vivo recordings from rigorously identified dentate gyrus granule cells. We also discovered that the activity of granule cells varies over a wide range, consistent with a log-normal distribution of firing frequency. To quantitatively analyze synaptic activity, we developed a new method for detection of EPSPs based on the principles of machine learning and optimal filtering (Zhang et al., in preparation). In comparison to conventional methods, including template fit and deconvolution, the method was up to 3-fold more powerful. Surprisingly, we found that both active and silent cells received spatially tuned synaptic input. Fourier analysis indicated that the input showed place-like tuning, grid-like tuning, or complex mixtures (Zhang et al, in revision). Finally, we used transsynaptic rabies labeling to examine the converging input on dentate gyrus granule cells and CA3 pyramidal neurons. We found that hippocampal neurons not only received input from superficial layers of the entorhinal cortex, but also from neurons in the entorhinal subplate (Ben-Simon et al., in preparation). Based on these results, we propose a revision of the trisynaptic circuit model of the hippocampal formation. Finally, we established a full-scale model of the dentate gyrus–CA3 network (Guzman et al., 2019, bioRxiv). We found that a winner-takes-all mechanism mediated by lateral inhibition in the dentate gyrus contributes to pattern separation (Espinoza et al., 2018, Nature Communications; Guzman et al., 2019, bioRxiv). Furthermore, we discovered that the detonation properties of the hippocampal mossy fiber synapse shift the balance between pattern separation and pattern completion towards pattern completion.