Tag Archives: learning

While you were sleeping: Neural reactivations promote motor learning

Originally published on the PLOS Neuroscience Community

Do you recall that moment when you first learned to ride a bike? After days of practice, it finally clicked. Almost effortlessly, your legs cycled in perfect harmony as you maneuvered gracefully around turns and maintained impeccable balance. You may have learned this skill decades ago but it will likely stick with you for the rest of your life. How did your brain accomplish this remarkable feat of transforming a series of forced and foreign actions into an automatic, fluid movement sequence? A new PLOS Biology study by Dr. Dhakshin Ramanathan and colleagues explored the neural substrates of motor memory acquisition, reporting that reactivation of task-related neural activity patterns during sleep promotes motor skill learning in mice.

Sleep enhances motor learning

To evaluate motor learning, the researchers trained mice to perform a task in which they had to reach for a food pellet. Since sleep is known to be important for memory consolidation, the mice were allowed to sleep before and again after the reach task. The mice performed the task a second time after sleeping to assess how sleep affected their performance. Accuracy on the reaching task improved over the course of the first training period, whereas the mice responded more quickly after sleeping; hence, “online learning” (during the task) improved accuracy and “offline learning” (while sleeping) improved speed. No changes occurred when the mice were sleep-deprived between task sessions, pinpointing sleep – rather than the passage of time – as the source of the performance boost.

Behavioral paradigm (Ramanathan et al., 2015)

Behavioral paradigm (Ramanathan et al., 2015)

Sleep-dependent neural changes

The researchers next explored the neurophysiological basis for this sleep-dependent learning, recording neural activity from the forelimb area of motor cortex. As co-author Dr. Karunesh Ganguly explained,

“Studies had previously studied hippocampal-based memory systems. It remained unclear specifically how the motor system (i.e., procedural memory) processes memories during sleep.”

When the mice performed the task the second time, after sleeping, the onset of neural firing (time-locked to the reach) peaked earlier, and this activity was more strongly modulated by the task. Notably, firing onset did not change after sleep deprivation, confirming that sleep was necessary for this temporal shift in the neural response.

Learning-related reactivation

Elsewhere in the cortex and hippocampus, during rest neurons will fire in a particular sequence matching the temporal pattern during a prior experience, an event known as “replay” that is thought to support formation of memory for that experience. The researchers speculated that reactivation or replay in motor cortex may similarly promote motor learning. Neural activity patterns identified from the reach task were more prevalent during sleep after the task, showing – as predicted – reactivation of task-related activity after motor learning. When mice performed the reach task on multiple days, over the course of all days the degree of reactivation during sleep correlated with reduced reaction time, linking stronger neural reactivation with behavioral improvements.

A) Reactivation of task-related neural activity after learning. B) Reactivation correlates with improvements in reaction time. (Ramanathan et al., 2015)

A) Reactivation of task-related neural activity after learning. B) Reactivation correlates with improvements in reaction time. (Ramanathan et al., 2015)

Since the authors observed neural reactivation during motor learning, they next wondered whether the temporal sequence of this reactivation may be an important element of the memory code. The neural activation pattern during sleep more closely matched the task-related activity pattern after learning than before, although some of the temporal information in the sleep sequence was lost. Although prior studies have shown a role for hippocampal or cortical replay in memory consolidation, Dr. Ganguly raises the important distinctions that here, they “did not find evidence of ‘replay’ (i.e., sequences) but ‘reactivation’ (i.e., synchronous bursts).”

Learning-related plasticity was evident even at the single neuron level. Those neurons with the highest task-related activity were most strongly reactivated during sleep, and those showing the strongest reactivation also happened to show the most dramatic shift in the onset of their response to reaching. The authors speculate that this increased temporal coupling of neural activity to the task could facilitate binding the neurons into a distributed “movement complex” that aids formation of the motor memory.

Locking to spindles and slow waves for widespread plasticity 

Burst of high-frequency activity – known as spindles – and slow wave oscillations have both been implicated in offline learning. If task-related reactivations during sleep are important for memory consolidation, the authors reasoned, they may be temporally linked to spindles or slow waves. After learning the reach task, reactivations were in fact more closely time-locked and phase-locked (i.e., occurred at a particular phase of the cycle) to fast spindles. Reactivations were also more strongly time-locked and shifted their phase-locking to slow oscillations. Thus, during sleep, neural activation patterns related to the motor task were not only more prevalent after training, but their timing was also refined to coincide with particular neural events that may facilitate memory formation. Since spindles may be involved in synchronizing long-range cortical activity, locking task-specific reactivations to spindles could tie them into neuroplastic changes throughout widespread brain networks supporting consolidation.

At a recent talk on sleep-dependent memory consolidation, the speaker compared the neural reorganization that takes place during memory formation to a house renovation. Just as it’s more comfortable and effective for us to check into a hotel while our house is renovated, learning may be more effective when the brain checks out – into the quietude of sleep – while neural reorganization occurs. This may explain why sleep is so important for learning, yet it doesn’t explain how the brain stores new memories during sleep. Past studies have identified neuronal reactivation, coordinated with spindles and slow waves, as critical for forming declarative memories. Dr. Ramanathan and colleagues’ findings suggest that these mechanisms also occur in the motor cortex to support a radically different form of memory – the kind that helped you learn to ride a bike many years ago.

Clarifying the neural dynamics of sleep-dependent learning holds profound implications not just for those of us hoping to learn to play a new instrument or refine our dance moves. It may also help us better understand the remarkable neuroplasticity that underlies rapid motor learning during early development, and hold potential to promote recovery from motor impairments following brain injury. Dr. Ganguly is optimistic about the possible applications of their findings:

“Motor learning is likely an essential process during rehabilitation. Surprisingly little is known about the role of sleep and replay during recovery. With further study, one could imagine using sleep and offline processing to maximize the learning during rehabilitation.”

References

Ji D, Wilson MA (2007). Coordinated memory replay in the visual cortex and hippocampus during sleep. Nat Neurosci. 10:100-107. doi: 10.1038/nn1825

Ramanathan DS, Gulati T, Ganguly K (2015). Sleep-Dependent Reactivation of Ensembles in Motor Cortex Promotes Skill Consolidation. PLOS Biol. doi: 10.1371/journal.pbio.1002263

Stickgold R (2005). Sleep-dependent memory consolidation. Nature. 437:1272–1278. doi: 10.1038/nature04286

Image credit https://www.flickr.com/photos/echoforsberg/

Advertisements
Tagged , , , ,

#PLOS #SfN15 Recap: Hidden variables of behavior

Originally published on the PLOS Neuroscience Community

The Society for Neuroscience meeting is unique in both is breadth and depth. There are sessions on literally everything Neuro, each delving with exquisite detail and nuance into their given topic. While this level of focus is great for those seeking comprehensive coverage of their niche, it can be daunting for those looking for a broader sampling of the field’s cutting edge. The Hidden Variables of Behavior symposium was one of the rare sessions to stray from the single-track convention to elegantly bridge seemingly disparate topics, methodologies and applications, producing a standout session with exceptionally broad appeal. It accomplished this by exploring a theme that is perhaps the unifying motivation underlying nearly all Neuroscience research: how does the brain engender behavior? How does neural activity give rise to the thoughts, interactions with our environment, and engagements with others that define our experiences? In an enthralling series of talks by Loren Frank, Mark Schnitzer, Yang Dan and Catherine Dulac, the symposium covered topics ranging from learning and memory to sleep and social behavior. This session had it all.

Rapidly alternating representations of present and past in hippocampal-cortical networks

frankLoren Frank kicked off the symposium by exploring how the hippocampus supports our ability to remember the past and plan for the future. Hippocampal cells have a remarkable ability to replay past experiences via high-frequency oscillations during sharp waves known as ripples. When an animal traverses a path its hippocampal neurons will fire in a characteristic sequence that codes its trajectory; later, at rest or while sleeping, this spiking sequence will repeat, with the sequence sped up approximately twenty times the original rate! Disrupting hippocampal ripples impairs sequence learning, indicating that they’re critical for acquiring memories. However, the mechanisms, at both regional and whole-brain levels, by which sharp wave ripples (SWRs) help to consolidate memories are unclear.

Place_Cell_Spiking_Activity_Example

Much attention has been paid to neurons in hippocampal subregions CA1 and CA3, which are excitable by high-speed motion and positively modulated by SWRs. However, Frank’s group identified a new group of hippocampal neurons – CA2P and CA2N – that are also positively and negatively modulated by SWRs, respectively. Notably, the CA2N population has an exceptionally high level of baseline activity and preferentially fires during rest or low speed motion. Because of their distinct function, these rebellious cells may be crucial for ongoing processing of the current state while maintaining representations of the past and future.

Although some (including yours truly) may hold a hippocampo-centric view of memory, Frank reminds us that memory is “not just a hippocampal thing.” Looking to the rest of the brain, his group found that SWRs recruit 35% of prelimbic regions, including cells that are both excited and inhibited by SWRs. Similar to the distinct populations of CA2P and CA2N cells, prefrontal cortex neurons may activate during either high-speed motion or immobility. This balance of excitation and inhibition in the hippocampus and surrounding cortex may promote rapid transitions between representations of the past and future, and facilitate their integration for learning and planning.

Large-scale ensemble neural dynamics underlying learning and long-term associative memory

schnitzer_mark_107Mark Schnitzer continued with this theme, presenting intriguing findings regarding the spatiotemporal properties of neural adaptations subserving learning. However, equally impressive are the advanced imaging tools his lab is developing to explore these issues. Their techniques allow neural recordings in behaving animals at unprecedented spatial depths and extents over long time scales. For instance, their current best is recording 1202 hippocampal cells in a freely moving mouse. Someone give this man the “I-recorded-the-most-neurons” award!

Using these tools, Schnitzer has been exploring hippocampal morphological and physiological changes that contribute to learning. CA1 neurons are a likely target for spatial learning, as they show place-cell activity, preferentially responding to particular regions of an animal’s environment. Surprisingly, dendrites in subregion CA1 are remarkably stable, suggesting that dendritic plasticity is unlikely to be the critical factor underlying learning. However, CA1 spine turnover is relatively rapid – on the order of 8-10 days – in comparison to cortical spines, of which 50% are permanent over a month. Schnitzer explained that although these cells are temporally stochastic in that they sometimes take breaks from their place-coding activity, when they return to the neuronal “spatial ensemble” they always return to encode the same place. What’s more, CA1 spatial representations are refined by learning, becoming both more accurate and reliable in their coding. I’ll be eagerly following Schnitzer’s work to see how their ongoing methodological innovations and applications advance our understanding of the hippocampal dynamics supporting long-term memories.

Neural circuits for sleep control

YIR_RH_DanYang Dan turned from this fast-paced discussion of rapid neural plasticity, spatial navigation and learning to examine neural regulation of sleep. Historically, neurons that trigger alertness and waking have been easy to identify, but researchers have struggled to track down those “sleep neurons.” Past lesion and c-fos studies have shown that hypothalamic – particularly preoptic – neurons are important for inducing sleep.

Combining optogenetics with electrophysiology, Dan’s lab has expanded upon these findings to pinpoint both the responsible cell types and their specific sleep-inducing effects. In particular, activating GABAergic preoptic cells projecting to the tuberomammillary nucleus (also of the hypothalamus) promotes non-REM sleep initially, and REM sleep later. The midbrain’s ventrolateral periaqueductal gray also promotes sleep, but only the non-REM type. Dan’s findings together suggest that mutual inhibition across these key hypothalamic and brainstem regions regulates transitions across three general brain states of waking, REM sleep and non-REM sleep.

Long-term changes in the representation of social information in the mouse medial amygdala

DulacAfter all this talk about sleep, my hypothalamic sleep neurons had begun batting the morning’s adenosine antagonists. Fortunately, Catherine Dulac’s captivating talk exploring the bases of social interactions and sex-specific behavior kept me alert and engaged. Two key circuits working in concert to process social information, she explained, are the olfactory and vomeronasal systems. This latter system in particular may act as a switch to promote appropriate (and suppress inappropriate) sex-specific behavior.

Dulac’s research, fusing molecular, genetic and electrophysiological techniques, has identified the medial amygdala as a critical stop along the vomeronasal circuit for mediating sex-specific social signaling in mice. Medial amygdalar encoding of social cues is not only sexually dimorphic, but is also regulated by salient social experiences including mating and co-housing. Furthermore, the efficiency of medial amygdalar signaling also changes after mating in a sex-specific manner, increasing in males but decreasing in females. Together, Dulac’s work has pinpointed the medial amygdala as an indispensible hub within an extensive neural circuit that regulates social behavior and in turn, is modulated by sexual and social experience.

Every SfN has at least one session that reminds me why I love the brain and re-ignites my passion for Neuroscience. This year, the Hidden Variables of Behavior symposium was it! It may be a year away, but I’m eagerly awaiting #SfN16 for similarly inspiring talks.

For an abbreviated play-by-play, visit my Storified live-tweeting of the symposium’s highlights.

References

Anderson EB, Grossrubatscher I, Frank L (2015). Dynamic Hippocampal Circuits Support Learning- and Memory-Guided Behaviors. Cold Spring Harb Symp Quant Biol. 79:51–58. doi: 10.1101/sqb.2014.79.024760

Attardo A, Fitzgerald JE, Schnitzer MJ (2015). Impermanence of dendritic spines in live adult CA1 hippocampus. Nature. 523:592–596. doi: 10.1038/nature14467

Bergan JF, Ben-Shaul Y, Dulac C (2014). Sex-specific processing of social cues in the medial amygdala. eLife. 3:e02743. doi: http://dx.doi.org/10.7554/eLife.02743

Brennan PA (2001). The vomeronasal system. Cell Mol Life Sci. 58(4):546–555.

Ego-Stengel V, Wilson MA (2010). Disruption of ripple-associated hippocampal activity during rest impairs spatial learning in the rat. Hippocampus. 20(1):1–10. doi: 10.1002/hipo.20707

Weber F, Chung S, Beier KT, Xu M, Luo L, Dan Y (2015). Control of REM sleep by ventral medulla GABAergic neurons. Nature. 526:435–438. doi: 10.1038/nature14979


Tagged , , , , , , ,