Synaptic tagging: homeostatic plasticity goes Hebbian

Distinct plasticity mechanisms enable neurons to effectively process information also when facing global perturbations in network activity. In this issue of The EMBO Journal, Dubes et al (2022) provide a molecular mechanism whereby individual synapses during periods of chronic inactivity are “tagged” for future strengthening. These results lend further support to the idea that local, nonmultiplicative mechanisms play an important role in homeostatic synaptic plasticity as has been demonstrated for Hebbian‐like synaptic plasticity.

Recent work shows how global perturbations in network activity can also act locally to support future synaptic activity.

Neurons embedded within neural networks possess different sets of plasticity mechanisms, which enable them to perform efficient information processing even when facing global perturbations in the overall network activity, e.g., during periods of chronically high or low activity. While Hebbian plasticity is associated with input‐specific changes in synaptic strength that are dependent on previous activation history, homeostatic synaptic plasticity (HSP) has long been thought to function as a global weight normalizer, which either potentiates or depresses synapses uniformly and independent of previous activation history in a direction opposite to the stimulation (Turrigiano, 2008). Such a multiplicative scaling mechanism has been reported to keep synapses in a functional dynamic range of plasticity, and it could explain how neuronal networks can balance between being plastic and stable at the same time. Physiologically, HSP has been studied in the context of sensory deprivation (Desai et al, 2002), and more recently, it has been observed to take place during sleep in a process that promotes memory consolidation (Diering et al, 2017). While HSP could consolidate the discrepancy between stable and plastic networks, the original view of a purely multiplicative scaling mechanism has recently been challenged. In fact, both theoretical (Rabinowitch & Segev, 2008) and experimental (Sutton et al, 2006) evidence argue for a contribution of local regulatory mechanisms in HSP. Local HSP might, for example, be important to endow neurons with the ability to perform computations at the level of dendritic segments, which in turn might boost information processing capacity (Poirazi & Papoutsi, 2020).

In this study, Dubes et al (2022) tackle two major questions with regard to local, input‐specific HSP: first, do excitatory synapses, much like in Hebbian plasticity, scale dependent on their previous activation history? Second, what might be the molecular mechanism, or “tag,” which marks synapses for preferential scaling? Initially, the authors use a well‐established experimental paradigm to induce homeostatic synaptic upscaling—48‐h bath application of tetrodotoxin (TTX) to primary rat hippocampal neurons. By comparing both electrophysiological and microscopy data, they show that the cumulative distributions of miniature excitatory currents (mEPSCs) and synapses containing the activity‐dependent postsynaptic scaffolding protein Homer1c do not scale uniformly, arguing for an input‐specific upscaling in response to TTX. In a search for potential “synaptic tags” during upscaling, the authors focus on the candidate synaptopodin (SP), a component of the spine apparatus and known regulator of HSP that is present only in a subset of spines (Vlachos et al, 2013). They observe that even at basal conditions, larger spine sizes and higher AMPA‐type glutamate receptor (AMPAR) content are correlated with the occurrence of SP, consistent with a tagging function of the protein. Fluorescence recovery after photobleaching (FRAP) experiments, in which RFP‐tagged synaptopodin (RFP‐SP) was co‐transfected with a super‐ecliptic (SEP)‐tagged GluA2 recombinant protein, was performed to study stabilization dynamics between the two proteins. After photobleaching SEP‐GluA2 on synaptic sites, a significantly lower recovery rate was observed when RFP‐SP was present, suggesting that SP‐positive spines have a greater ability in stabilizing surface‐expressed AMPA receptors. Thus, SP might fulfill its tagging function by controlling the synaptic recruitment of AMPARs.

What mechanisms could be responsible for the nonuniform upregulation of SP at large synapses? Given the well‐documented role of local translation in upscaling (Sutton et al, 2006), the authors considered the involvement of microRNAs (miRNA), which have been shown to control local translation at synapses (Schratt, 2009). The neuron‐enriched miR‐124 was reported to downregulate SP in the spinal cord, thereby representing an obvious candidate. Using reporter gene assays, Dubes et al (2022) were able to validate SP as a direct target of miR‐124 in neurons. They went on to show that TTX induces downregulation of miR‐124, whereas miR‐124 overexpression occludes the TTX‐mediated increase in SP‐positive spines, synaptic AMPARs, and mEPSCs. In a very elegant set of experiments, they established that miR‐124 exerts its effect on upscaling via the unique binding site within the SP 3′UTR. First, they replaced endogenous SP with a recombinant version containing either the wildtype or mutant (miR‐124‐binding site lacking) 3′UTR and found that transfecting the mutant SP construct occluded TTX‐mediated upscaling. Second, they transfected an antisense oligonucleotide directed against the miR‐124‐binding site (“target site blocker”) and found that it similarly occluded upscaling.

To investigate the site of SP synthesis, the authors performed puromycin‐proximity ligation assays (Puro‐PLA) that allow to visualize sites of local translation for specific proteins. They observed that SP‐PLA clusters were significantly upregulated in number and intensity upon TTX treatment specifically in the proximity of larger synapses, providing further support for local regulation of SP. In a final set of technically challenging experiments, the authors addressed a potential synapse‐autonomous mode of HSP and its dependence on local miR‐124‐dependent SP translation. In CA3 pyramidal neurons of organotypic hippocampal slices, they injected presynaptically a tetanus‐toxin light chain (TetTX) construct, which effectively silences presynaptic neurons by interfering with VAMP2‐dependent vesicle fusion. At the same time, the above‐described SP replacement constructs were infused in postsynaptically connected neurons identified by paired patch‐clamp recordings. After 48 h, the authors observed that dendritic spines of silenced synapses underwent input‐specific upscaling only when the wildtype SP 3′UTR construct was postsynaptically infused. By contrast, upscaling was neither observed in neighboring unmanipulated synapses nor when the mutated form of the 3′UTR was employed.

In summary, this study provides strong experimental support that homeostatic synaptic upscaling induced by TTX is an input‐specific, nonmultiplicative mechanism, whereby upscaling occurs preferentially at larger synapses. It further suggests that SP “tags” synapses for upscaling, and that local de‐repression of miR‐124‐dependent SP translation mediates associated spine growth (Fig 1).

Figure 1. Model for the role of synaptopodin (SP) in synaptic tagging during homeostatic plasticity.

At basal activity (left), a subset of spines is “primed” for growth due to one of the indicated tagging mechanisms. Chronic inactivity (right) then leads to miR‐124 degradation and local translation of SP at these specific spines, enabling the growth of the spine apparatus and spine expansion. See text for further details.

While this study represents a major advance in our understanding of local homeostatic regulation, it also raises several questions that need to be addressed in follow‐up studies. For example, what is the mechanism of selective recruitment/synthesis of SP at large synapses that later undergo upscaling? Are their mechanisms to specifically capture SP mRNA at spines, or is rather local miR‐124 degradation the determining factor? High‐resolution mRNA/miRNA in situ hybridization in combination with spine imaging should help to resolve this question. Moreover, it will be interesting to test whether similar mechanisms are engaged in other homeostatic processes, such as synaptic downscaling. Interestingly, SP seems to be downregulated in the synaptodendritic compartment during picrotoxin‐induced homeostatic synaptic downscaling (Colameo et al, 2021), consistent with a bidirectional regulatory role in HSP. Many miRNAs have been assigned important roles in HSP and local translation at synapses, and it is therefore tempting to speculate that miRNAs might have a more general role in the input‐specific control of HSP. The relevant miRNA‐target interactions still need to be discovered. Finally, the physiological implications of input‐specific HSP are largely elusive. Intriguingly, in the context of sleep, Seibt and Frank (Seibt & Frank, 2019) propose that experiences during wake promote Hebbian plasticity‐related synaptic changes and prime‐specific circuits for further regulation during sleep by plasticity‐associated products. These “tags” are then expressed during sleep to mediate up‐ or downscaling of specific synapses. This model contrasts with the more classical synaptic homeostasis hypothesis (SHY) model, which invokes uniform reductions in synaptic strength during sleep (Tononi & Cirelli, 2014). Further studies are needed to determine the synapse‐specific dynamic expression of SP or other candidate synaptic tags during different sleep phases, or in the stabilization of memory engrams in general. Since input‐specific HSP has the potential to become maladaptive, potential pathophysiological implications (e.g., Alzheimer and epilepsy) are also worth considering.

The EMBO Journal (2022) 41: e112383.