Memories are stored in the complex network of neurons in the brain. With the help of innovative tools to manipulate the connections between neurons, memories in mice can now be erased with a beam of light. See Article p.333
More than a century ago, the German biologist Richard Semon proposed that memories leave physical traces in the brain, and coined the term 'engram' to describe such traces1. Although the concept has gained general recognition, the search for the engram is ongoing. In this regard, the synapse — a specialized connecting region between neurons — has received much attention, but there is still no direct evidence of a causal link between synaptic changes and memory formation. In this issue, Hayashi-Takagi et al.2 (page 333) fill this gap. Using ingenious protein engineering and live imaging, the authors identify which synapses are activated when a mouse learns a motor skill, and then weaken these synapses to erase motor memory.
Most synapses in the brain form between axons (neuronal 'output cables') and dendrites (input cables). Signals to excitatory synapses are usually received by micrometre-sized protrusions called spines that emanate from dendrites. The size of the spine head correlates with the strength of the synapse3. Spines may emerge, disappear or change in size during learning and memory formation, reflecting changes in the wiring of neuronal circuits3.
To investigate the causal relationship between the formation of motor memories and the structural potentiation of spines (spine formation or enlargement), Hayashi-Takagi et al. developed an 'optoprobe' called AS-PaRac1 that manipulates potentiated spines in response to light. The DNA construct for AS-PaRac1 encodes a light-activatable version of the small signalling protein Rac1, whose prolonged activity induces spines to shrink. The construct also incorporates the dendrite-targeting sequence of the gene Arc, which is expressed rapidly and transiently in response to neuronal activity, ensuring that the probe moves to dendritic spines that are undergoing structural potentiation. The AS-PaRac1 optoprobe is the first optogenetic tool to enable the manipulation of potentiated spines.
Hayashi-Takagi and colleagues expressed AS-PaRac1 in the motor cortex of mice and trained the animals to run on an accelerating rotating rod known as a rotarod. Light activation of AS-PaRac1 in potentiated spines after learning caused the spines to shrink, disrupting the animals' ability to run on the rotarod. This demonstrates the causal relationship between synaptic strength and motor memory in this context (Fig. 1).
Next, the authors showed that the effect of the probe is task-specific. When mice learnt to run on the rotarod and then learnt to walk on a thin beam, disrupting the spines that were potentiated during beam walking did not affect performance on the rotarod. Furthermore, AS-PaRac1 activation in spines that spontaneously potentiated two days after learning (presumably because of unrelated motor tasks) did not affect motor performance. Finally, when the authors retrained mice on the same task for which spine potentiation had been disrupted, most of the optically shrunken spines reverted to their original potentiated sizes. Together, these results suggest that distinct subsets of synapses are altered in a task-specific way during motor learning and memory formation.
In the long quest for the engram, neuroscientists have reached the consensus that the mammalian brain stores different memory traces in different subsets of neurons in specific regions. Methods for labelling, imaging, activating and silencing neurons in animals have enabled researchers to map the ensemble of neurons that correlates with a particular learning task, to manipulate their activities, and even to generate artificial memory traces4,5,6. However, a single neuron may participate in the processing and storage of more than one distinct piece of information7. Therefore, the engram of a particular memory involves not only the identity of the constituent neurons, but also the entire set of synaptic connections between these neurons. How memory is allocated at this synaptic level remains unclear.
To qualify as an engram, a synaptic circuit should satisfy several criteria. First, changes in synaptic structures and function should correlate with learning. Second, blocking such synaptic modifications should prevent memory formation, demonstrating the need for these changes. And third, artificially inducing synaptic changes should be sufficient to produce a memory without the need for behavioural training. Over the past decade, in vivo imaging has revealed8 that the dynamic formation and elimination of dendritic spines correlates with motor-skill learning and memory. Now, Hayashi-Takagi and colleagues have taken the next step, by establishing necessity — they show that undoing the synaptic changes that accompany motor learning does indeed disrupt the memory.
The development of genetic and optical tools such as AS-PaRac1 promises to enable dissection of the finer details of the engram. The use of promoter sequences that drive the expression of target genes in a cell-type-specific manner, as well as connectivity-specific labelling methods9, can help to unravel the roles in learning and memory of synaptic circuits formed by different types of neuron — revealing, for example, the relative contributions of excitatory and inhibitory neurons, or of neurons in different layers of the brain's cortex. When we have a deeper understanding of the molecular signalling events that occur at synapses during memory formation10, tools similar to AS-PaRac1 can be devised to modulate other components of the molecular machinery. Improved microscopy techniques can already target individual neurons or synapses11, rather than manipulating a population of neurons as a whole.
When used together, such technical advances will enable us to strengthen existing engrams, to facilitate the formation of new ones, and to generate synthetic memory traces at the synaptic level. We will then be able to study the interaction between different memory traces, as well as the mechanisms that translate an engram into behavioural outputs. These efforts should allow us to gain an understanding of the intriguing phenomenon of memory simply by shining a light on its physical basis.Footnote 1
Semon, R. Die Mneme als erhaltendes Prinzip im Wechsel des organischen Geschehens (Wilhelm Engelmann, 1904)
Hayashi-Takagi, A. et al. Nature 525, 333–338 (2015).
Holtmaat, A. & Svoboda, K. Nature Rev. Neurosci. 10, 647–658 (2009).
Han, J.-H. et al. Science 323, 1492–1496 (2009).
Garner, A. R. et al. Science 335, 1513–1516 (2012).
Ramirez, S. et al. Science 341, 387–391 (2013).
Jia, H., Rochefort, N. L., Chen, X. & Konnerth, A. Nature 464, 1307–1312 (2010).
Chen, C.-C., Lu, J. & Zuo, Y. Front. Neuroanat. 8, 28 (2014).
Luo, L., Callaway, E. M. & Svoboda, K. Neuron 57, 634–660 (2008).
Mayford, M., Siegelbaum, S. A. & Kandel, E. R. Cold Spring Harb. Perspect. Biol. 4, a005751 (2012).
Packer, A. M., Russell, L. E., Dalgleish, H. W. P. & Häusser, M. Nature Methods 12, 140–146 (2015).
Related links in Nature Research
About this article
Brain Research Bulletin (2017)
Frontiers in Bioengineering and Biotechnology (2016)