Final Report Summary - HEBBIANNEWSPINES (Visualizing the structural synaptic memory trace: presynaptic partners of newly formed spines) A central goal of modern neuroscience is to identify the physical changes in a neuronal network that encode long-term memories. The most widely accepted concept of how learning on the level of neuronal circuits could be achieved is summarized in the famous paraphrase of Donald Hebb’s postulate: “Fire together, wire together”: Two neurons that show correlated activity during a given behavioral task should strengthen their connection by enhancing signal transmission at individual existing synapses but also by de novo generation of synaptic contacts.The majority of excitatory synapses in the brain are formed on dendritic spines – small protrusions from the parent dendrite of a postsynaptic neuron. We previously have extensively followed the formation and elimination of these synapses both in vitro as well as in vivo using chronic 2-photon imaging. We now know that the turnover and total number of both dendritic spines and presynaptic boutons change dramatically during modification of sensory experience, while the gross dendritic morphology remains strikingly stable. Given that some spines are likely to last for a lifetime it is tempting to speculate that this form of structural synaptic plasticity is the neuronal substrate of long-lasting memories. The easily comprehensible link of learning and synapse formation became part of popular science knowledge. However, up to now all observations so far are entirely correlative: We and others have previously shown that the induction of long-term synaptic potentiation (LTP) in vitro leads to an increase in the number of spines specifically at dendrites in the vicinity of the stimulation electrode. We could also show in vivo that new spines are formed and stabilized during retinal lesions and monocular deprivation (MD), both of which are classical paradigms to study cortical plasticity. However, whether new spines indeed follow the full set of rules for Hebbian plasticity remains to be established: While it seems clear that simultaneous pre- and postsynaptic activity is needed for the structural changes to occur, it is still entirely unclear if the presynaptic neuron forming a functional contact on a newly formed spine indeed was the one that showed correlated activity with the postsynaptic cell. In fact, since it has never been shown which signals are transmitted via a newly formed synapse in vivo, it has yet to be demonstrated that new spines indeed could carry a memory trace related to a specific behavioral or sensory task.The goal of the EU-funded HebbianNewSpines project was to functionally identify the presynaptic partners of newly formed spine synapses in vitro (approach A) and in vivo (approach B). We used hippocampal slice cultures to study whether neurons that ‘fire together’ indeed also ‘wire together’ by forming new stable spine synapses (approach A). In vivo we asked, using the well-established MD paradigm in the visual cortex, if the information carried by a newly formed or eliminated spine synapse follows our predictions derived from observations of functional changes on the population level (approach B). Approach A: Using a combination of stable and specific optogenetic long-term stimulation and optophysiological imaging techniques we studied (i) if newly formed spines show any initial preference for axons stimulated during LTP induction or, alternatively, whether new spines promiscuously form contacts with random presynaptic partners, (ii) if continuous activity in these axons is needed to stabilize transient spine synapses, and (iii) we aimed to follow the time course of synapse formation from a newly formed spine to a functional synapse. The following project milestones were reached:1. We developed a dual-color marker that leads to the expression of a stoichiometrically fixed level of a structural and functional indicator (mTurquoise2-P2A-GCaMP6s), enabling us to visualize (sub-)cellular morphology together with single synapse activation by NMDA-receptor mediated Ca2+ influx (Fig. 1A,B).2. We developed a robust all-optical LTP paradigm based on direct stimulation of presynaptic boutons expressing ChR2. We can therefore be certain that only ChR2+ axons will be stimulated and thereby act as active potential targets for newly-formed spines.3. We find that all-optical LTP leads to the formation of new persistent spines over the time course of 5 hours.Approach B:We have previously shown that following an MD-induced shift in ocular dominance (OD) new spines are formed in the binocular region of mouse visual cortex, even after the classical critical period has ended. A subset of these spines persists for an extended period of time. We hypothesized that after a second MD these new spines are re-used and thereby might constitute the ‘memory trace’ of a previously established but subsequently overridden shift OD. Our specific questions were: (i) Does the strengthened response to stimulation of the open eye indeed utilize newly formed spine synapses? To this end, we aimed to visualize the activation of new stable spines in response to visual stimulation in vivo. Then, by separation of visual inputs (i.e. left vs. right eye) this approach should allow the assignment of structure (de novo spine formation and stabilization) to function (strengthening of open eye input) for the first time on a single synapse level in vivo. Being able to determine the response characteristic (i.e. OD) of an individual new spine would be the first step in demonstrating that the formation of new synapses is indeed involved in changing this neuron’s response properties and would strongly support the notion that the acquisition and retention of long-term memory occurs via spine formation. The following project milestones were reached:1. We are now able to achieve sparse viral transfection of either defined subsets of neurons using a combinatorial viral approach.2. We have developed a dual color construct that leads to approximately stoichiometric coexpression of a bright structural marker (mRuby2) together with the ultrasensitive GeCI GCaMP6s. This new tool allows us to obtain high quality structural and functional data of imaged neurons but also introduces further benefits like ratiometric signal analysis and superior movement correction. 3. We are now able to image the binocular tuning properties of individual spines and of large cell populations.We are confident that now we have the necessary tools at hand to approach the important question of the relevance of de novo synapse formation in the near future. In the long run this should allow us to answer whether newly formed or eliminated synapses are mere bystanders of learning and memory or if they play a central role in reshaping cortical networks during neuronal plasticity.An up-to-date website with project information and resources is accessible through: http://www.neuro.mpg.de/37634/rose
