- Motor practice results in highly efficient and seemingly effortless motor execution.
- Synaptic plasticity observed after seconds to minutes of practice increases the efficacy of information transfer within task-specific neural ensembles and is instrumental for behavioral improvements.
- Measuring corticomotor excitability in the primary motor cortex (M1) provides indirect markers of synaptic plasticity in humans.
- Synapses in motor cortex are potentiated during the day to form new memories but also by random stimuli, which might turn out to be behaviorally irrelevant.
- Specific consolidation processes during sleep are instrumental for strengthening task-relevant synapses (e.g., due to reactivation) and weakening others. These mechanisms are essential for increasing the signal-to-noise ratio within neural networks and for maintaining long-term synaptic homeostasis.
Motor learning is characterized by long-lasting performance improvements caused by practice that result in highly efficient and seemingly effortless motor execution. Skill-specific motor memories are stored in the human central nervous system and have been conceptualized as engrams, which are assumed to correspond to physical traces in the brain (1). One aspect of particular interest is how the nervous system strikes a fine balance between plasticity and stability so that, on the one hand, adaptation occurs quickly in an activity-dependent manner, with repetitive motor training triggering the first neuronal changes within the range of seconds to a few minutes. On the other hand, once a motor skill has been acquired — like riding a bicycle — it remains remarkably stable and can be retrieved even after years without practice.
In an attempt to bridge concepts from the cellular to the neural network and behavioral level and from animal models to humans, I will review evidence with regard to three important mechanisms that are proposed to be instrumental for understanding neuroplastic changes in the motor system.
First, skill acquisition and motor learning reflect changes at the cell and neural network level, which are driven by functional and structural changes enhancing synaptic transmission within task-relevant neural ensembles (Fig. 1).
Second, even though cellular changes are not directly observable in intact humans, macroscopic measurements have been argued to provide indirect markers of processes governing synaptic plasticity in the primary motor cortex (M1) (Fig. 2).
Third, most studies have investigated how motor memories are encoded and modulated while awake. I will review evidence supporting the hypothesis that sleep is essential for motor memory consolidation and, particularly, for differentiating between synapses that represent relevant motor memories, and need to be strengthened to ensure long-term storage, and other synapses representing behaviorally irrelevant information, which are weakened. One important implication of the proposed hypothesis is that sleep-dependent mechanisms are essential for maintaining synaptic homeostasis within M1 and, thus, for restoring the motor cortex’s capacity to undergo further neuroplastic changes while awake (Fig. 3).
MOTOR MEMORIES ARE STORED IN SYNAPSES OF THE CENTRAL NERVOUS SYSTEM
Motor memories are stored in the human central nervous system, where they can be rapidly formed and modified because of motor practice. The cellular components mediating this high level of plasticity are synapses that regulate information transfer from one nerve cell to another. Synapses have the capability to modify the efficacy with which information is transferred between cells, a phenomenon known as synaptic plasticity. In the adult brain, synaptic plasticity is triggered in an activity-dependent manner. Short-term plasticity, which can already be observed after a single action potential arrives at the presynaptic terminal, lasts for a few milliseconds up to a few minutes or less. By contrast, long-term mechanisms of synaptic plasticity, like long-term potentiation (LTP) and its functional counterpart, long-term depression (LTD), are cellular mechanisms that are activated quickly but alter synaptic transmission over days and years.
LTP Triggers Rapid Changes of Synaptic Efficacy That Drive Behavioral Improvements
According to Hebbian theory (2), which can be summarized as “neurons that fire together, wire together”, LTP only is observed when activity in the presynaptic and postsynaptic neuron coincides, i.e., depolarization occurs tightly linked in time (within approximately 100 ms). Research into the molecular mechanism underlying LTP has revealed that AMPA receptors (responding to α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid) and NMDA receptors (responding to N-methyl-D-aspartate) are essential for triggering this form of synaptic plasticity. AMPA receptors respond to the excitatory neurotransmitter glutamate and produce large excitatory responses at the postsynaptic membrane; thus, they transmit neuronal activation from one cell to the next. NMDA receptors act as coincidence detectors because they only are activated if glutamate is released by the presynaptic neuron while the postsynaptic neuron is sufficiently depolarized. Under these conditions, the NMDA receptor opens its pore allowing Ca2+ to enter the cell, where it activates intracellular signaling cascades that increase the synaptic efficiency for transferring excitatory signals by integrating new AMPA receptors into the postsynaptic membrane.
Even though there is a clear gap between understanding synaptic plasticity at the cellular level and explaining performance improvements caused by motor practice, several properties make LTP a plausible cellular mechanism underlying learning and information storage in the form of acquired behaviors and memory (see (3) for review):
- LTP is rapidly induced and develops incrementally over a period of 10–20 min;
- it is activity-dependent and increases the probability that presynaptic activity triggers postsynaptic action potentials;
- it is long lasting (under certain circumstances, more than a year (3));
- LTP is input specific and associative; and
- LTP is reversed by its counterpart LTD, ensuring that synaptic information transfer can be dynamically modified causing either strengthening or weakening of synapses.
Until recently, it has been difficult to causally link synaptic plasticity at the cellular level to behavioral indexes of learning and memory. In a seminal study, Nabavi et al. (4) used optogenetics, a method that allows the experimenter to selectively modulate neuronal activity in vivo and study how this influences the animal’s behavior. They were the first to show that experimentally strengthening synapses in a specific circuit by LTP is required for the formation of fear memories, and that weakening the same synapses by LTD can erase the memory, thus confirming the central hypothesis that synaptic changes form the basis of learned behavior.
LTP Mediates Motor Practice‐Dependent Changes of Sensorimotor Networks
Even though LTP and LTD have been most intensively studied in subcortical areas, they represent general phenomena that drive plasticity across the brain including the sensorimotor system. M1 contains a body map that is roughly segregated into representations of the lower limbs, upper limbs, and the head. However, within these defined subregions, the cortex is organized into distributed neuronal networks containing multiple overlapping motor representations that are functionally connected through an extensive network of horizontal connections rather than via a one-to-one mapping between neurons and muscles (5). M1 networks have been shown to be highly plastic in response to sensorimotor training, which leads to the strengthening of horizontal connections so that new functional neuronal assemblies are formed. More specifically, Rioult-Pedotti et al. (6) trained rats to reach and grasp food pellets with only one forepaw. Once the skill was overtrained, the efficacy of M1 networks was determined ex vivo by measuring local field potentials in response to electrostimulation that was applied either to the hemisphere contralateral to the trained forepaw (trained hemisphere) or to the ipsilateral hemisphere (untrained hemisphere). Interestingly, stimulation of the forepaw representation in the trained hemisphere evoked larger local field potentials than stimulation of the opposite untrained hemisphere, indicating that repetitive skill training increased the efficacy of information transfer within the M1 neuronal network. Not surprisingly, it was hypothesized that this use-dependent form of synaptic strengthening was mediated by LTP. If true, it suggests that synaptic plasticity can only occur within a defined modification range, i.e., synapses that have undergone LTP in the past become less likely to undergo additional LTP while they become increasingly susceptible to LTD expression (7). Indeed, when the trained M1 was compared with the untrained M1, the response to experimentally induced LTP was markedly reduced, whereas the response to LTD was enhanced (8). These results are consistent with the theory that motor training causes synaptic potentiation via LTP and moves synapses in the trained M1 nearer to the ceiling of their modification range.
LTP Triggers Structural Changes of Synapses
In addition to modifying synaptic function, motor training also has been shown to trigger structural changes. Excitatory synapses are usually formed between axons and so-called dendritic spines, small protrusions that emanate from dendrites and are the receivers in the neuron-to-neuron signaling chain. Even when no new memories are acquired, a subpopulation of small spines appears and disappears within tens of minutes so that connections to neighboring axons are constantly remodeled. Despite this constant “turnover” of dendritic spines, overall spine densities remain roughly stable in the adult brain (9–11). Not all small spines can form functional synapses, but in addition to randomly remodeling neural connections, the size of an individual spine also fluctuates because spine heads enlarge and shrink in a stochastic manner. Enlarging the spine heads has been associated with increasing the number of AMPA glutamate receptors in the membrane, enabling signal transmission in the smallest spines and increasing the efficacy of excitatory synapses (11,12). Thus, imaging dendritic spines allows studying both the rearrangement of cortical connections and changes in a putative structural correlate of synaptic strength.
Activity-dependent plasticity that leads to LTP influences both spine dynamics and spine head volumes. Training mice with a forelimb-reaching task triggers rapid growth of de novo spines (within 1 h of the first training trial) in those M1 neurons representing the trained forelimb (13). Subsequently, this rapid spinogenesis is followed by enhanced elimination that preferences removal of old spines that were already present before the motor training, whereas newly formed spines tend to be maintained long beyond the training period (i.e., up to 4 mo despite no additional training). This process ensures that synapses are rearranged in the motor cortex, forming ensembles that represent a specific motor task while overall spine density is maintained at a similar level to that observed in control mice. LTP not only triggers the rearrangement of cortical connections, but also causes particularly small spines to increase the size of their spine head so that more AMPA glutamate receptors are integrated into the membrane. These changes can already be observed within minutes after the start of training and increase the efficacy of signal transmission (11,12) (Fig. 1). It has only recently been shown that changes in spine head volume are causally related to memory formation. Hayashi et al. (14) labeled spines that were potentiated during motor learning and, subsequently, applied synaptic optogenetics to cause selective shrinkage of the spine heads. Interestingly, this experimental manipulation caused significant performance decrements on the trained task, indicating a causal link between the remodeling of small dendritic spine and behavioral motor memories. Even though small spines seem to play a crucial role for skill acquisition, medium and large spines that have larger spine heads that contain many AMPA receptors remain virtually unchanged by LTP induction, i.e., they are protected from further potentiation. These findings are consistent with the idea that large spines are highly stable and represent long-term memory traces that are protected from being integrated into new spine ensembles representing different memory contents (12). They also highlight that LTP is an important mechanism for triggering initial reorganization of motor networks. However, once a task is overtrained, high neuronal efficacy is achieved independent of LTP.
Even though synapses are likely to constitute the basic unit for information storage, memories are ultimately laid down in neuronal ensembles, i.e., networks of preferentially connected engram cells, which can span across different brain areas. It is this preferential connectivity pattern at the network level that represents the specific input-output mapping linking a stimulus to the acquired behavior. Two recent findings are interesting in this regard. First, a simulation study has suggested that even if synaptic connections are nonstationary and highly flexible, stable input-output mappings (representing a specific skill like riding a bicycle) can be achieved at the network level because the brain is highly redundant, i.e., the same pattern of neural activity can be achieved by many different patterns of synaptic connectivity (15). Second, work investigating synaptic changes in well-defined neuronal circuits underlying fear memories (1) has shown that engram cells undergo two phases of plasticity: early LPT changes synaptic weights by increasing the insertion and conductance of AMPA receptors when encoding a new memory. Late LTP further enhances the synthesis of AMPA receptors for several hours after the encoding period, resulting in synaptic strengthening via cellular consolidation. Perturbing synaptic strengthening by blocking late LTP has been shown to affect the retrieval of memories but not memory storage per se. Thus, the connectivity pattern within distributed cell ensembles can be hypothesized to be a flexible yet lasting substrate for simultaneously storing different memories. However, retrieval of a specific memory depends crucially on synaptic plasticity and cellular consolidation, at least when tested in memory systems that rely on the hippocampus (1).
MEASURING NEUROPLASTICITY IN THE HUMAN MOTOR CORTEX
In the human central nervous system, it is not possible to directly measure plasticity at the level of the synapse because all available techniques are restricted to macroscopic levels, i.e., they can only detect changes in neuronal activity or structure that occur across large neuronal populations. Changes in synaptic plasticity within M1 can be indirectly inferred when the system is probed with transcranial magnetic stimulation (TMS). Administering TMS over the human scalp depolarizes cortical neurons in the underling tissue and gives rise to action potentials. TMS that induces currents in the anterior-to-posterior direction is believed to excite predominantly interneurons in cortical layer II/III, which in turn activate pyramidal cells in layer V (16). Given that the stimulation intensity is sufficiently high, descending volleys are propagated via the corticospinal tract to the spinal cord, where they depolarize motor neurons and are measured via compound motor-evoked potentials (MEP) in the electromyogram of the targeted muscle. Thus, TMS evokes MEPs transsynaptically so that MEP magnitudes provide a summary measurement of the excitability of cortical networks formed by interneurons, which synapse on corticospinal neurons (16). Note that MEPs also are influenced by the state of the spinal segment of the pathway and, particularly, by the excitability of spinal motor neurons. Therefore, linking potential changes in corticomotor excitability induced by motor practice exclusively to cortical plasticity requires careful experimentation and additional control experiments (16).
Motor Training Causes LTP-Like Plasticity in Human Primary Motor Cortex
In humans, it has been shown repeatedly that improvements in motor performance due to repetitive practice that aims to reduce movement variability are paralleled by reorganization of the motor cortex (17,18). One approach for testing M1 reorganization in the laboratory is to evoke isolated thumb movements by single TMS pulses (Fig. 2). Once the baseline direction is established, participants practice voluntary thumb movements in the opposite direction for 30 min (i.e., if the TMS pulse evoked thumb extension at baseline, the participants produced thumb flexion movements). After training, TMS is applied with the same intensity and at the same location as during baseline. Importantly, network reorganization is inferred when the evoked movements are in or near the practiced direction (19). This form of plasticity can be interpreted as reflecting use-dependent reorganization of cortical networks because, after training, the same TMS input elicits different muscular responses than before training. This reorganization seems to emerge predominantly at the cortical level because control experiments, which stimulated pyramidal corticospinal tract neurons directly via transcranial electrical stimulation (i.e., circumventing intracortical networks in M1), revealed only minor changes (19). An alternative approach is to measure practice-induced changes in corticomotor excitability; for example, by “motor mapping,” which tests from how many scalp positions MEPs can be evoked, or by maintaining the coil over the exact same scalp position and analyze MEP amplitudes. For this latter approach, results from animal work predict that an increase in synaptic efficacy will be reflected by a TMS pulse of the same intensity evoking a larger MEP after training than before (Fig. 2). Indeed, a large number of TMS studies in humans report that this is in fact the case for many motor tasks (20–25). Even though use-dependent plasticity is considered a relatively basic form of plasticity, it has been shown to interact with other motor learning mechanisms occurring during error-based motor adaptation (26) or skill acquisition (27). In particular, it has been demonstrated that use-dependent plasticity is enhanced when triggered by goal-directed skill training that includes success-based reinforcement signals (27).
Several lines of research provide converging evidence that increased TMS-evoked responses early after motor practice might reflect an LTP-like mechanism in humans. First, animal work has shown that LTP induction crucially relies on NMDA receptor activity (28) and that LTP is suppressed by high activity in inhibitory circuits (in particular gamma-aminobutyric acid (GABA)ergic circuits) (29). In a seminal paper, Bütefisch et al. (17) demonstrated that the training-induced reorganization of a muscle representation measurable with TMS was markedly reduced when NMDA receptors were blocked or inhibitory GABAergic activity increased with pharmacological agents. These results exhibit striking parallels to findings in animal preparations, suggesting that use-dependent plasticity due to repetitive motor training in humans is driven by an LTP-like mechanism. Second, animal work also has demonstrated that if motor practice is sufficient to activate LTP, M1 neurons are shifted closer to the ceiling of their synaptic modification range (8). Several groups (20,24,25) have tested this prediction with a protocol that experimentally induced LTP or LTD using a paired-associative stimulation protocol after motor training. In agreement with the animal literature, these studies found that motor learning significantly diminished the effect of LTP induction and markedly increased the effect of LTD induction. Interestingly, these effects were observed only after the first training session. When training was continued until day 5, corticomotor excitability tested before task execution remained permanently increased compared with day 1. Moreover, executing the motor task did not change corticomotor excitability or impact subsequent LTP or LTD induction. Thus, similar to animal work (30), LTP-like plasticity triggered initial changes in the synaptic efficacy of M1, but it is not required after prolonged training, suggesting that the horizontal connections of M1 were permanently remodeled by stabilizing skill-specific synapses. Finally, it has been shown that perturbing M1 activity using a repetitive TMS paradigm abolishes performance improvements resulting from motor training (31). This perturbation effect was anatomically specific to M1, and performance improvements were only reduced when M1 was perturbed early after learning. These findings indicate that M1 is crucially involved in use-dependent aspects of motor learning, even though it is not completely clear whether this reflects a problem with memory storage or memory retrieval.
In summary, several lines of research have shown that human M1 changes in response to motor training and confirmed key predictions from mechanistic experiments investigating LTP induction in animal preparations. This led to the notion that motor learning in response to training is associated with changes in synaptic efficacy in M1, which is most likely driven by an LTP-like mechanism.
Based on the concept of use-dependent plasticity, it is important to note that the human brain is continuously shaped by daily motor actions. Therefore, reduced use or nonuse, e.g., due to experimental immobilization, also triggers rapid neural changes as indicated by reduced cortical excitability (32) and reorganized cortical maps (33), which are already observed after 10–24 h of nonuse.
Weak Association Between Physiological and Behavioral Markers Of Early Learning
Even though there is a large body of evidence demonstrating changes in corticomotor excitability shortly after motor training (particularly for repetitive tasks that require ballistic finger actions), few studies report a significant association with behavioral markers of motor learning. A notable exception is the study of Jensen et al. showing that 4 wk of skill training caused an increase in corticomotor excitability that correlated with behavioral improvements, while the same result was not observed for strength training (34). Moreover, it was shown that an increase of corticomotor excitability measured immediately after practicing a serial reaction time task is predictive of and causally related to an individual’s offline improvements when the task was retested 10 h after skill acquisition (35). Causality was demonstrated by experimentally increasing corticomotor excitability via theta burst stimulation over the M1 or prefrontal cortex, which resulted in enhanced offline improvements in performance. By contrast, there seems to be no robust relation between changes in corticomotor excitability and behavioral improvements caused by the preceding motor training (for a critical discussion, see (36,37). This might reflect that motor behaviors result from interactions between task-specific cortico-subcortical networks and spinal circuits (38), which all have the capacity to undergo changes in response to training (39,40) but are not necessarily reflected in macroscopic electrophysiological markers of M1 activity. In addition, both TMS based markers and behavioral measurements of skill are not a pure reflection of training-induced changes because they are “compound measurements” of several interacting neural elements in M1 and several interacting behavioral processes, respectively.
SLEEP IS ESSENTIAL FOR STABILIZING MEMORIES BUT ALSO FOR KEEPING THE BRAIN PLASTIC
Most of the studies reviewed earlier investigated neuroplastic processes while participants were awake, a state that is believed to mainly serve memory encoding. Here, I will focus on the role sleep plays in maintaining the right balance between stabilizing motor memories and restoring synaptic plasticity to promote memory consolidation.
Even though sleep-related consolidation was first demonstrated for declarative memories, several motor tasks (e.g., short motor sequence tapping tasks and adaptation learning) have been shown to benefit when learning is followed by a period of sleep (but note that not all motor tasks are equally susceptible to sleep-dependent consolidation as discussed in (13)). Beneficial effects of postlearning sleep are typically demonstrated in the form of “offline gains” in motor performance, i.e., skill levels are higher at a retention test performed after one night of sleep or a nap in the laboratory than at the end of learning (41). Sleep dependency can only be inferred when these offline gains are specific to sleep and absent or significantly smaller in a control group spending the same period awake. There is converging evidence that motor memory consolidation is especially dependent on nonrapid eye movement (NREM) sleep, which is characterized by large slow-wave activity (large amplitude oscillations at frequencies between 0.5 and 4.5 Hz) and sleep spindles (transient oscillatory bursts in the range of 11–16 Hz that last approximately 0.3–2 s). Correlational evidence indicates that offline gains are associated with higher slow-wave activity and increased spindle activity either in the motor cortex or in upstream cortical areas relevant for motor skill performance (42–44). Causal evidence for the relation between NREM sleep and memory consolidation in humans was revealed recently by Lustenberger et al. (45), who showed that motor memory consolidation can be improved by selectively enhancing spindle activity via a closed-loop system that detects sleep spindles online and applies time-locked transcranial alternating current stimulation over the frontal cortex to facilitate cortical synchronization in the spindle frequency range. Interestingly, stimulation-induced changes in fast spindle activity resulted in better motor memory consolidation as indicated by a retention test, and the extent of these two phenomena was correlated. Together, these findings suggest that sleep spindles are functionally involved in motor memory consolidation.
Two important mechanisms have been hypothesized to promote different aspects of memory consolidation (Fig. 3) and brain plasticity during NREM sleep: first, reactivation of memories during sleep according to the active system consolidation theory (41); and, second, synaptic downscaling as predicted by the synaptic homeostasis hypothesis (46).
Consolidation by Reactivating Motor Memories During Sleep
The first mechanism can be tested by the cued memory reactivation approach that links skill acquisition during wakefulness to specific olfactory or auditory stimuli, i.e., a motor task is practiced in the presence of a specific odor or sound. When asleep, presenting the cue is supposed to trigger a replay/reactivation of the associated motor memory trace, which might further strengthen the underlying synaptic connections.
Using this approach, it was shown that reactivating a motor sequence-tapping task by presenting odor (47) or auditory cues (48) during NREM sleep enhanced overnight performance changes when compared with a control group. In an elegant study, Laventure et al. (47) demonstrated that offline gains were enhanced only when the motor memory was cued during the N2 stage of NREM sleep (which caused a significant increase in spindle activity), whereas no such benefits were observed when cues were presented during rapid eye movement (REM) sleep. These results in humans are in line with animal experiments showing that motor cortex neurons form highly synchronized ensembles during the learning of a new motor skill, which are reactivated during postlearning sleep and, more specifically, in conjunction with spindle oscillations (44). The degree of sleep-dependent reactivation has been shown to be correlated with behavioral offline gains measured during a retention test. In addition, reactivation during NREM sleep seems to be related to forming new dendritic spines (49) that are specific for the acquired motor task, whereas REM sleep has been shown to be essential for selectively stabilizing task-relevant synapses and pruning others (49).
Restoring Synaptic Homeostasis During Sleep
There is increasing evidence that NREM sleep, specifically deep sleep, also is essential for maintaining synaptic homeostasis. Because the brain is highly plastic when awake, synapses are constantly strengthened to represent statistical regularities about the current environment, even if these alleged regularities turn out to be random or not important for behavior. Ever-increasing synaptic potentiation, however, has major disadvantages because it can saturate learning at the synaptic level, decrease the signal-to-noise ratio within a neural system, and increase cellular energy consumption. The idea that synapses get generally potentiated while awake was indirectly supported by studies showing that corticomotor excitability of the human cortex increases over the course of wakefulness (50,51). In addition, de Beukelaar et al. showed that acquiring one sequence-tapping task in the morning and another one in the evening triggered changes in corticomotor excitability that are consistent with LTP-like plasticity (i.e., TMS-evoked responses were larger after motor training than before), but the extent of these learning-specific changes was higher in the morning than in the evening. Interestingly, the capacity to exhibit larger changes in corticomotor excitability — indicative of more synaptic plasticity — was restored in participants who acquired the first motor task in the evening and the second motor task the next morning after one night of sleep. It has been hypothesized that deep NREM sleep, and particularly slow-wave activity, is essential for downregulating synaptic strength and maintaining synaptic homeostasis (46). In line with this synaptic homeostasis hypothesis, slow-wave activity is highest shortly after falling asleep, when the sleep need is still high, whereas it is markedly reduced at the end of the night after restorative processes have taken place (46). Interestingly, neural plasticity induced by practicing a specific motor task during wakefulness leads to more slow-wave activity during sleep, but only in task-specific brain areas (52). However, to demonstrate that slow waves are directly responsible for restorative processes, one has to establish a causal relation between these phenomena. Recently, we used a novel perturbation approach, where real-time closed-loop acoustic stimulation was timed to coincide precisely with the vulnerable downphase of electroencephalogram slow waves in M1 and investigated the consequences on behavioral and neurophysiological markers of neuroplasticity arising from dedicated motor practice (23). We showed that the capacity to exhibit learning-induced synaptic potentiation, measured by the increase in corticomotor excitability from pre- to postlearning, is reduced by wakefulness but is restored during unperturbed sleep, thus reproducing previous results (50). This restorative process is markedly attenuated when slow waves are selectively perturbed in primary motor cortex, demonstrating that deep sleep is a requirement for maintaining sustainable learning efficiency. Importantly, differences in slow-wave activity measured between the perturbed and the unperturbed night correlated with changes in corticomotor excitability in response to motor learning. Interestingly, perturbing slow wave activity in M1 did not erase the capability to improve behavioral markers of skill acquisition (e.g., average movement accuracy or speed); rather, it increased motor variability indicating that the motor system might have operated less efficiently than after unperturbed deep sleep (note that such subtle effects are not completely surprising given that slow-wave activity in M1 was changed by only 13%). Taken together, our findings (23) indicate that local deep sleep might play an important role for maintaining the brain’s capacity to respond efficiently to motor training at the next day and thus for ensuring long-term adaptability to the environment. In addition, downregulating the strength of synapses that were only weakly potentiated in the first place might cause additional memory benefits by ensuring a better signal-to-noise ratio within task-relevant neural networks.
Converging evidence from research in animals and humans indicates that acquiring a new motor skill triggers functional and structural changes that lead to more efficient information transfer within task-relevant networks. This is achieved by processes at the cellular level, most notably LTP. Synaptic plasticity is a dynamic process, where new synapses are continuously formed and erased in an activity-dependent manner until stable, task-specific connectivity patterns emerge within dedicated cell ensembles. Once these ensembles are stabilized, the resulting memory traces can be maintained for long periods without being reactivated. These processes explain how the brain maintains a fine balance between plasticity, which is important for learning, and stability, which is important for remembering skills that have been acquired long ago. Even though motor memories are usually formed while awake, a growing body of evidence indicates that memory traces are reactivated during sleep, causing additional strengthening of task-relevant synapses that benefit memory the next day. In addition, there is increasing experimental support for the hypothesis that weakly potentiated, task-irrelevant synapses are downscaled during sleep, thus maintaining the brain’s capacity to respond efficiently to motor training.
I thank D.G. Woolley and C. Lustenberger for their comments on the manuscript.
This study was supported by the National Science Foundation Switzerland (SNSF 320030_175616 and SNSF 320030_14956).
1. Tonegawa S, Pignatelli M, Roy DS, Ryan TJ. Memory engram storage and retrieval. Curr. Opin. Neurobiol
. 2015; 35:101–9.
2. Hebb DO. The organization of behavior. Organ Behav
. 1949; 911(1):335.
3. Abraham WC, Logan B, Greenwood JM, Dragunow M. Induction and experience-dependent consolidation of stable long-term potentiation lasting months in the hippocampus. J. Neurosci
. 2002; 22(21):9626–34.
4. Nabavi S, Fox R, Proulx CD, Lin JY, Tsien RY, Malinow R. Engineering a memory with LTD and LTP. Nature
. 2014; 511(7509):348–52.
5. Sanes JN, Donoghue JP. Plasticity and primary motor cortex. Annu. Rev. Neurosci
. 2000; 23:393–415.
6. Rioult-Pedotti MS, Friedman D, Hess G, Donoghue JP. Strengthening of horizontal cortical connections following skill learning. Nat. Neurosci
. 1998; 1(3):230–4.
7. Bienenstock EL, Cooper LN, Munro PW. Theory for the development of neuron selectivity: orientation specificity and binocular interaction in visual cortex. J. Neurosci
. 1982; 2(1):32–48.
8. Rioult-Pedotti MS, Friedman D, Donoghue JP. Learning-Induced LTP in Neocortex. Science
. 2000; 290(5491):533–6.
9. Holtmaat A, Svoboda K. Experience-dependent structural synaptic plasticity
in the mammalian brain. Nat. Rev. Neurosci
. 2009; 10(10):647–58.
10. Fu M, Zuo Y. Experience-dependent structural plasticity in the cortex. Trends Neurosci
. 2011; 34(4):177–87.
11. Kasai H, Fukuda M, Watanabe S, Hayashi-Takagi A, Noguchi J. Structural dynamics of dendritic spines in memory and cognition. Trends Neurosci
. 2010; 33(3):121–9.
12. Matsuzaki M, Honkura N, Ellis-Davies GC, Kasai H. Structural basis of long-term potentiation in single dendritic spines. Nature
. 2004; 429(6993):761–6.
13. Xu T, Yu X, Perlik AJ, et al. Rapid formation and selective stabilization of synapses for enduring motor memories. Nature
. 2009; 462(7275):915–9.
14. Hayashi-Takagi A, Yagishita S, Nakamura M, et al. Labelling and optical erasure of synaptic memory traces in the motor cortex. Nature
. 2015; 525(7569):333–8.
15. Ajemian R, D’Ausilio A, Moorman H, Bizzi E. A theory for how sensorimotor skills are learned and retained in noisy and nonstationary neural circuits. Proc. Natl. Acad. Sci. U. S. A
. 2013; 110(52):E5078–87.
16. Siebner HR, Ziemann U. The TMS book (Das TMS-Buch)
. 1st Edition. Berlin Heidelberg: Spinger-Verlag; 2007. p. 3–55.
17. Bütefisch CM, Davis BC, Wise SP, et al. Mechanisms of use-dependent plasticity in the human motor cortex. Proc. Natl. Acad. Sci. U. S. A
. 2000; 97(7):3661–5.
18. Bütefisch CM. Plasticity in the human cerebral cortex: lessons from the normal brain and from stroke. Neuroscientist
. 2004; 10(2):163–73.
19. Classen J, Liepert J, Wise SP, Hallett M, Cohen LG. Rapid plasticity of human cortical movement representation induced by practice. J. Neurophysiol
. 1998; 79(2):1117–23.
20. Rosenkranz K, Kacar A, Rothwell JC. Differential modulation of motor cortical plasticity and excitability in early and late phases of human motor learning
. J. Neurosci
. 2007; 27(44):12058–66.
21. Perez MA, Wise SP, Willingham DT, Cohen LG. Neurophysiological mechanisms involved in transfer of procedural knowledge. J. Neurosci
. 2007; 27(5):1045–53.
22. Zhang X, de Beukelaar TT, Possel J, et al. Movement observation improves early consolidation of motor memory. J. Neurosci
. 2011; 31(32):11515–20.
23. Fattinger S, de Beukelaar TT, Ruddy KL, et al. Deep sleep maintains learning efficiency of the human brain. Nat. Commun
. 2017; 8:15405.
24. Ziemann U. Learning modifies subsequent induction of long-term potentiation-like and long-term depression-like plasticity in human motor cortex. J. Neurosci
. 2004; 24(7):1666–72.
25. Stefan K, Wycislo M, Gentner R, et al. Temporary occlusion of associative motor cortical plasticity by prior dynamic motor training. Cereb. Cortex
. 2006; 16(3):376–85.
26. Huang VS, Haith A, Mazzoni P, Krakauer JW. Rethinking motor learning
and savings in adaptation paradigms: model-free memory for successful actions combines with internal models. Neuron [Internet]
. 2011; 70(4):787–801. Available from: http://dx.doi.org/10.1016/j.neuron.2011.04.012
27. Mawase F, Uehara S, Bastian AJ, Celnik P. Motor learning
enhances use-dependent plasticity. J. Neurosci
. 2017; 37(10):2673–85.
28. Morris RG, Anderson E, Lynch GS, Baudry M. Selective impairment of learning and blockade of long-term potentiation by an N-methyl-D-aspartate receptor antagonist, AP5. Nature
. 1986; 319(6056):774–6.
29. Hess G, Donoghue JP. Long-term potentiation of horizontal connections provides a mechanism to reorganize cortical motor maps. J. Neurophysiol
. 1994; 71(6):2543–7.
30. Kleim JA. Cortical synaptogenesis and motor map reorganization occur during late, but not early, phase of motor skill learning. J. Neurosci
. 2004; 24(3):628–33.
31. Muellbacher W, Zlemann U, Wissel J, et al. Early consolidation in human primary motor cortex. Nature
. 2002; 415(6872):640–4.
32. Burianova H, Sowman PF, Marstaller L, et al. Adaptive motor imagery: a multimodal study of immobilization-induced brain plasticity. Cereb. Cortex
. 2016; 26(3):1072–80.
33. Kolasinski J, Makin TR, Logan JP, et al. Perceptually relevant remapping of human somatotopy in 24 hours. Elife
. 2016; 5: pii: e17280.
34. Jensen JL, Marstrand PCD, Nielsen JB, Lundbye J, Jens B. Motor skill training and strength training are associated with different plastic changes in the central nervous system. J. Appl. Physiol
. 2005; 99:1558–68.
36. Carson RG, Ruddy KL, Mcnickle E. What do TMS-evoked motor potentials tell us about motor learning
? Adv. Exp. Med. Biol
. 2016; 957:143–57.
37. Bestmann S, Krakauer JW. The uses and interpretations of the motor-evoked potential for understanding behaviour. Exp. Brain Res
. 2015; 233(3):679–89.
38. Wolpaw JR, Tennissen AM. Activity-dependent spinal cord plasticity in health and disease. Annu. Rev. Neurosci
. 2001; 24:807–43.
39. Debaere F, Wenderoth N, Sunaert S, Van Hecke P, Swinnen SP. Changes in brain activation during the acquisition of a new bimanual coordination task. Neuropsychologia
. 2004; 42:855–67.
40. Puttemans V, Wenderoth N, Swinnen SP. Changes in brain activation during the acquisition of a Multifrequency Bimanual Coordination Task: from the cognitive stage to advanced levels of automaticity. J. Neurosci
. 2005; 25(17):4270–8.
41. Rasch B, Born J. In search of a role of REM sleep in memory formation. Neurobiol. Learn. Mem
. 2015; 122:1–3.
42. Miyamoto D, Hirai D, Fung CC, et al. Top-down cortical input during NREM sleep consolidates perceptual memory. Science
. 2016; 352(6291):1315–8.
43. Walker MP, Brakefield T, Morgan A, Hobson JA, Stickgold R. Practice with sleep makes perfect: sleep-dependent motor skill learning. Neuron
. 2002; 35(1):205–11.
44. Ramanathan DS, Gulati T, Ganguly K. Sleep-dependent reactivation of ensembles in motor cortex promotes skill consolidation. PLoS Biol
. 2015; 13(9):e1002263.
45. Lustenberger C, Boyle MR, Alagapan S, Mellin JM, Vaughn BV, Fröhlich F. Feedback-controlled transcranial alternating current stimulation reveals a functional role of sleep spindles in motor nemory consolidation. Curr. Biol
. 2016; 26(16):2127–36.
46. Tononi G, Cirelli C. Sleep and the price of plasticity: from synaptic and cellular homeostasis
to memory Consolidation and Integration. Neuron
. 2014; 81(1):12–34.
47. Laventure S, Fogel S, Lungu O, et al. NREM2 and sleep spindles are instrumental to the consolidation of motor sequence memories. PLoS Biol
. 2016; 14(3):e1002429.
48. Antony JW, Gobel EW, O’Hare JK, Reber PJ, Paller KA. Cued memory reactivation during sleep influences skill learning. Nat. Neurosci
. 2012; 15(8):1114–6.
49. Li W, Ma L, Yang G, Gan WB. REM sleep selectively prunes and maintains new synapses in development and learning. Nat. Neurosci
. 2017; 20(3):427–37.
50. de Beukelaar TT, Van Soom J, Huber R, Wenderoth N. A day awake attenuates motor learning
-induced increases in corticomotor excitability
. Front. Hum. Neurosci
. 2016; 10:138.
51. Huber R, Mäki H, Rosanova M, et al. Human cortical excitability increases with time awake. Cereb. Cortex
. 2013; 23(2):332–8.
52. Huber R, Ghilardi MF, Massimini M, Tononi G. Local sleep and learning. Nature
. 2004; 430(6995):78–81.