Posted comment on ´Network-wide reorganisation of procedural memory during NREM sleep revealed by fMRI` by S. Vahdat, S. Fogel, H. Benali and J. Dovon and published in eLife 2017;6:e24987 doi 10.7554/eLife.2498
Vahdat and colleagues described in their article their investigation of how learnt motor sequences formed memory traces during sleep and how there was a gradual shift of memory representations from a temporarily activated cortical pattern to a downscaled, more interconnected subcortical pattern.
For their experiments, Vahdat`s team used 13 healthy volunteers and recorded the development of memory traces of their subjects in real-time using the techniques of functional magnetic resonance imaging (fMRI) and electroencephalography (ECG). The subjects performed two different finger movement tasks one week apart. In the motor sequence learning task (MSL), they practiced a self-paced, known 5-item finger sequence task and this was compared to another task in which the subjects were asked to produce simultaneous movements of all four fingers at the same average frequency. This was termed the motor control task (CTL). The initial training session was held at 22.30 (termed ´learning session`, S1) and this was followed by a retest session at 09.00 the following morning (termed the ´retest session`, S2). Resting state conditions were met by the subjects staying awake with their eyes open and these were carried out before and after each practice session in the evening (termed RS1 and RS2) and the following morning (RS3 and RS4). A simultaneous EEG-fMRI recording scan was made lasting a maximum of 2.5 hr while the subjects slept in the scanner. Offline BOLD-fMRI imaging was carried out during the task learning, sleep and repetition of the task. Functional connectivity of brain areas was estimated using the overall co-activation of brain areas within a network during the different fMRI sessions.
The results of Vahdat and colleagues` experiments showed that specific neuronal firing activation patterns representing the learning of the motor sequences occurred in brain areas during the learning practice session (S1). This sequence related pattern was termed the ´learning pattern`. It was also observed that there was greater activation with the MSL rather than the CTL and although the speed of performance appeared faster for the MSL than the CTL, when the speeds of the MSL and CTL were intentionally matched, then no difference in average speed was observed. The summary of the activation peaks observed during the learning pattern was found to be related to the pattern observed representing consolidation. When the subjects were retested after sleep (S2), the authors found that the performance of the MSL had improved. They interpreted this as evidence of memory consolidation taking place during sleep. No such improvement was observed with the CTL. They also measured the accuracy of the performance of the MSL by looking at the percentage of incorrect key presses and it was found that in general there were a very low number of errors even after the overnight sleep. This was explained by the motor sequence being explicitly known by the subjects. An investigation into performance variability also found a significant decrease only in the MSL.
Vahdat and colleagues then looked at the pattern of activation of brain areas associated with learning and retesting of the motor sequence and found that the pattern of activation was similar for the learning session (learning pattern) and the retesting (consolidated pattern). Each had similar patterns of activity in the sensorimotor core areas, but there were significant differences between the cortical and subcortical areas. It was found that the consolidated pattern had increased activity (approx. 4 times as much) in sub-cortical areas (mostly basal ganglia and cerebellar areas) and decreased activity (approx. half as much) in cortical areas (mostly fronto-parietal sensorimotor regions). A volume-based analysis of the activated areas showed that two cortical clusters (including superior parietal lobule and anterior intraparietal sulcus) had greater activation during the learning session than during retesting and conversely, subcortical regions (including putamen and cerebellar regions) had greater activation during retesting. An estimate of functional connectivity (where strength was given as the connectivity index, CI) showed that there was a significant effect for both learning and retesting patterns with the latter being greater. A significant effect was also observed for the CTL. The authors showed that the effect was specific for consolidation by observing four recognised brain networks: the default mode, visual network, and left and right fronto-parietal. The CIs calculated for the resting states and during NREM sleep for both conditions showed no change in CI values. Therefore it was concluded that the changes observed were not due to overall changes caused by time, learning or sleep, but were specific for the tasks being undertaken.
Investigation of the CI analyses carried out using dual regression analysis of the fMRI allowed the authors to identify specific brain areas within the connectivity patterns that were recruited for each condition during the learning session and the Stage 2 NREM sleep period. Vahdat and colleagues found that transient reactivation in a cortically dominant pattern formed during learning was followed by downscaling of the functional connectivity in these areas followed by a gradual reorganisation of the representation towards a subcortically dominant consolidated trace during NREM sleep. The functional connectivity of the ventrolateral putamen area was significantly increased within the consolidated network during NREM sleep in MSL compared to CTL. A gradual increase in the strength of connectivity was observed and this was significantly related to the amount of gain seen with the motor performance. However, the pattern of the brain area connectivity during learning did not show significant change during the NREM periods with either task. Therefore, although the consolidated patterns were not increased immediately after training it was already significantly elevated before the retesting session indicating that the memory consolidation process took place during the sleep phase and was not caused by the practice during the retesting session after sleep.
Vahdat and colleagues also found that the V-VI areas of the cerebellar cortex also showed increased functional connectivity for both tasks during NREM sleep and the retesting session and this was also associated with the overnight behavioural gains observed. The functional connectivity within the posterior parietal cortex (more activated during the learning session) also demonstrated increased values in the resting state condition following the MSL.
Therefore, the authors concluded from their investigations that there is a gradual shift of the patterns of neuronal activation representing the motor sequences during the memory process from a transiently activated cortical pattern in learning to a downscaled, more interconnected and more subcortically dominant one during NREM sleep and after during recall. This indicated that NREM sleep is necessary for two complementary processes to consolidate human motor memory. They are: the suppression of the initial memory trace formed during learning; the restoration and reorganisation of the newly-learnt information in a more stable state following. The ventrolateral putamen appears to play a central role in the emergence of the consolidated pattern during NREM sleep. The authors concluded by indicating why sleep deprivation leads to impairment of motor skills and motor memory plus also indicated the possibility of enhancing newly learned skills by manipulating brain circuits during NREM.
What makes this article interesting is that it supports the view that sleep plays a beneficial, some might say even essential, role in the formation of long-term memories not only for explicit and implicit memories, but also as shown here for motor memories. Vahdat and colleagues demonstrated in their experiments that the learning of a five-finger sequence of movements produced as expected a specific pattern of neuronal activation of appropriate brain areas relating to motor memory. This activity pattern at learning represents the different complementary functions required for such tasks eg. control of muscle contraction and expansion and movement plus cognitive functions associated with visual pathways, attention, reasoning etc. that all occur simultaneously during the learning period. During the sleep phase (in particular the NREM phase) consolidation of the memory formed leads to greater performance of the motor sequence during the retesting phase. This was accompanied by reduced levels of variability in the performance and lower levels of recall error. With regards to the neuronal firing patterns, not only did the authors notice an increase in connectivity between the relevant areas in the retesting phase, but they also observed a shift of level of activity from the higher cortical areas (50% of the activity observed earlier for frontoparietal sensorimotor areas) to increased levels of activity in the subcortical areas (4 times increase in the basal ganglia areas and cerebellar V-V1) and particularly in the area of the ventrolateral putamen. Therefore, in order to understand their observations we need to look further at two areas in particular: the brain regions involved in motor movements and memory of those motor movements; and the role of sleep in the memory consolidation process.
Brain region activity related to motor learning and memory
Vahdat and colleagues in their article showed that there was a change in area activity between the learning and consolidation (retesting) phases of their experiment with a shift from cortical motor areas to subcortical structures eg. striatum and a gradual increase in connectivity between all participating areas. Area activity relates to the different functions occurring during the learning and retesting stages and these can be divided into two sets of processes: those relating to motor movements which one can assume are the brain areas associated with muscle movements and the second set which are those areas associated with cognitive functions such as visual processing, learning, reasoning, monitoring etc. plus those associated with emotional status and conscious awareness. This division is supported by McDougle, Bond and Taylor who separated the cognitive demands of motor memory into explicit and implicit learning all sub-serving the task performance. They described fast and slow processes with fast being associated with the explicit memories involved.
With regards to motor movements and motor memory of motor movements, it is known that the neuronal activation pattern follows area activity associated with what is known as the ´motor loop` – a group of interacting, sometimes interrelated brain areas where some areas influence excitatory effects on others in the loop and some inhibitory effects. Therefore, the greater the activity of some brain areas means other areas exhibit higher levels of inhibition and so, a change in activity of one area alone can have an overall effect on the performance of the motor loop as a whole. Another feature of this system is that specific activity leads to specific movement eg. force and direction since somatotopic maps are in present and also that because of the population coding scheme, the larger the active population, the finer the control of the movement. In the experiments discussed here the neuronal activation patterns representing the motor movements alone can be assumed to be the same whether the individual is in the learning phase or retesting phase.
Vahdat and colleagues found in their experiments that the brain areas, ventrolateral putamen and lobule VI of the cerebellar cortex, were mainly involved in the reorganization process following the learning phase of the motor sequence. This change in bias of function was linked to a change in functional connectivity of the putamen within the consolidated pattern during the NREM sleep phase, as well as during the post-sleep resting-state periods, and was thought to be related to the level of improved behavioural performance in the retesting stage following sleep. These results support the important role of the putamen in the motor loop and the observations that abnormal activity of this area is involved in certain disorders/diseases linked with abnormal or restricted movements eg. Parkinson`s disease, Tourette syndrome and stroke. Activation of the putamen is affected by the other member areas of the motor loop.
The motor loop consists of a group of brain areas whose activation patterns create the conditions required to perform motor movements. There is a group controlling strategy (eg. the higher levels of the brain such as the neocortex and basal ganglia), another controlling the tactics such as the movement of the muscles, the contractions, the organisation of sequences all based on previous experience (eg. the motor cortex and cerebellum) and the last group, the actual execution which involves the activation of the muscles (eg. the spinal cord and brain stem). Therefore, specific activation of the frontal (sensory cortex) or prefrontal, parietal or motor cortex areas leads to excitatory firing of the appropriate areas of the basal ganglia caudate region (through the somatotopic map condition). These activated caudate neurons then cause specific firing of areas of the globus pallidus or more importantly relating to the investigation described here, areas of the putamen, which then causes firing of the global pallidus. The globus pallidus is connected to the ventral lateral nucleus of thalamus (VLo), but global pallidus firing is inhibitory and therefore, any activation of the putamen increases the inhibitory effect on the VLo. The VLo is connected to the dorsal lateral thalamus which at rest is inhibited and hence, activity of this area undergoes excitation as the putamen activity relaxes the enforced inhibition. Át this point, the signal carries on through the motor loop to the higher cortical areas of supplementary motor cortex (area 6) and the premotor area. Areas 4 and 6 then send excitatory axons to the corticospinal tract which then activates the spinal cord. The other area involved in the motor loop is the M1 motor area whose layer V pyramidal cells receive input from other cortical areas and the thalamic ventral lateral nucleus. The output from such pyramidal cells activates the lower motor neurons such as those of the spinal cord and the subcortical sites for example associated with motor processing (brain stem). Hence, firing of the motor loop goes from the strategic areas to the execution areas.
Both of the two areas of the motor loop investigated by Vahdat and colleagues, that of the putamen and cerebellum, are capable of their wide-ranging involvement in motor movements and motor memory due to their complex physiology. The putamen has extensive connectivity to other brain areas, for example to the substantia nigra (an area linked with Parkinson`s disease), the globus pallidus, the thalamus (inhibitory effect from the putamen and excitatory effect to the putamen), and to the cortex (sends information to the putamen in multi-fibre pathways plus the area has numerous parallel circuits for cortico-subcortico-cortico communication loops). A look at the area at the cellular level shows a wide range of axons and dendrites which are highly arborized and exhibit a topographical organisation eg. anterior to posterior, and lateral to medial for functional and somatotropic gradients, diffuse terminal outputs, segregated terminals from adjacent regions and finely interconnected terminals from distal cortical regions in an overlapping function. It also has a varying range of active neurotransmitters, eg. GABA, which controls the inhibitory effect on the thalamus and dopamine, which acts presynaptically and influences the substantia nigra.
The functioning of the motor loop areas relates to each of their roles in motor movement and motor learning and memory and therefore, specific movements produce specific patterns of neuronal activation whether in learning or retesting. The other group of brain areas where activation could be observed relate to the cognitive demands placed on the individual because of the task. These include visual processing, and areas associated with cognitive functions such as learning, reasoning, monitoring etc. plus those associated with emotional status and conscious awareness. Even after sleep in the retesting session activity of this group of higher brain areas constitutes 50% of that reached during the learning cycle. Therefore, some cognitive processes are still active and repetition of the activation pattern is appropriate to memory recall of the learnt sequence. With regards to the putamen and its increased activity in retesting said to be linked to consolidation of the motor memory during sleep, we have already seen that learning is linked to putamen activity. This area integrates with other areas to complete tasks and an injection of muscimol (a GABA agonist) leads to a decrease in learning. In the case of reinforcement learning, then cholinergic interneurons in the area fire during stimulus with an impulse rate of 0.5 -3 and with regards to rule based tasks, then lesions of the area disrupt the recall due to the damage of the required hypothesis testing system.
The cognitive demands of the task not only relate to memory input, formation and recall they are also linked to emotional state which undergoes many changes during the learning and recall stages with each stage associated with different brain area activation. This means that the overall neural representation of the sequenced task will vary according to the emotional state observed at the time. However, only the overriding emotional status will be recorded in the form of an emotional tag alongside the information and this is the emotional status recalled along with that information. The concurrent emotional status observed due to real-time system activation will be expected as relaxed during re-testing due to successful recall. The changes in emotional status will also be mirrored by a shift in level of conscious awareness between the learning and retesting stages which will also be reflected by the activation pattern of the appropriate areas. As the learning of the motor sequence progresses then it is likely that there is a shift in conscious awareness to subconscious. Both conscious awareness and emotional status will be subconscious and neutral with any changes initiated through feedback (eg. monitoring for errors in the recalled sequences), a lack of the continuation of the sequence initiating the required prompts/reminders and increased speed although essentially the task sequence is order dependent and not time-dependent. The retesting session although proceeding with recall of learnt material with no variability demanded still requires visual input and other sensory input eg. body positioning, balance and error monitoring and therefore, increased activation in these areas should be observed.
Therefore, the activation of brain areas during the task given by Vahdat and colleagues reflects the requirements for motor movements plus for the cognitive demands placed on the individual due to the nature of the task. Whereas in other types of memory, the more connectivity observed the better the recall, ie. more details, more associations, in the case of motor memory, the more the connectivity then the finer the motor control.
Role of sleep in motor memory implied from the investigation
Sleep is believed to aid the consolidation of the learning neuronal firing patterns and this can have a beneficial effect on recall. For example, researchers have shown that a 10 minute nap can lead to a better recall of a story. In the experiments performed by Vahdat and colleagues regarding motor memory, sleep aids the consolidation process of the motor memory and a shift of strength of area activity is observed with lower levels observed for higher brain areas eg. the reasoning, strategy areas to the lower areas that of the putamen and cerebellum. There are many studies on sleep and memory and there are various hypotheses that cover the area of memory consolidation, including the trace reactivation, synaptic homeostasis, and systems consolidation hypotheses. The Trace Reactivation Hypothesis assumes that the repeated reactivation of a recently formed memory representation during sleep replayed in the slow-wave-sleep phase leads to a gradual strengthening of the learning-related connections (Guidotti and colleagues) especially in the hippocampus (an area known to exhibit activity even if the movement is not being carried out) and hence, to long-term storage of the memory trace.
Procedural memory has been found to be positively linked to sleep both in the sleep spindles and slow wave phases of sleep. This hypothesis is in contrast to the Systems Consolidation Hypothesis that proposes that sleep engages an active reorganization process that stabilizes the unstable neural representations formed of a new skill into a consolidated memory trace. In this hypothesis, protein synthesis is required. In the third hypothesis, the Synaptic Homeostasis Hypothesis it is proposed that local neuronal networks are potentiated and eventually become saturated during learning. In order for new information to be encoded, sleep is involved in the restoration of these local networks by downscaling the strength of synaptic connections. The authors of this article support this view by showing that there is a shift from the higher cortical areas to lower brain areas.
All three hypotheses however can be linked to the process involved in the consolidation of the memory traces even if the processes may be more complicated than suggested. The repetition of the sequence leading to the learnt sequence and then its recall in the retesting stage means that the temporary changes occurring in the initial learning stages cause physiological changes that mirror the translation of the memory from short-term to long-term memory. Consolidation comes through recall and retesting and it is disputed whether protein synthesis is required for this process or not, but it is known that histone acetylation is required.
In the experiments described here retesting occurs after sleep and the changes in connectivity imply that the consolidation process during sleep is the same as that which occurs when repetition is carried out whilst the individual is awake. The advantage of the consolidation process occurring during sleep is that the cognitive resources of the brain are minimised during sleep by the reduced sensory input, eg. visual pathway shut-down. Therefore, sleep is linked to reducing cognitive load allowing resources to be focused on the consolidation of memories formed from events recently experienced during the waking period. This implies that there is a point at which the cognitive demand outweighs the available resources and priorities have to be set. This point should be explored. We know at what point activity is not high enough for conscious awareness (through default network observations and minimally conscious individuals) or cognition (through dementia or coma sufferers), but the point at which the cognitive load is so great that the brain shuts down or sets its own cognitive priorities gives another angle to how the functioning of the brain could be manipulated.
Therefore, to conclude motor movements and learning of these movements rely on the activity of certain brain areas in a particular pattern. Motor memory just like episodic memory is dependent on learning and recall processes and the consolidation of memories can occur during sleep with replaying of the activation patterns specific to the movements during the SWS sleep phase. Vahdat and colleagues have shown that in the case of motor memories there is a shift of neuronal firing strength from the cortical areas to the putamen and cerebellum during sleep and hence, consolidation is linked to some extent to these shifts. This can be explained by the role of the putamen and cerebellum in the motor loop, the reduction of involvement of the higher areas in the formation of strategy and tactics of the motor movements required for the task in the recall phase compared to the learning phase and the decreased levels of conscious awareness and relaxed emotional status exhibited during the retesting phase. Learning impairments said to occur due to the lack of sleep can be caused by the inhibition of the replay process observed in the SWS and sleep spindle phases and are likely to affect both episodic and motor memories. They are likely to be linked to failed changes in area activation strength from the higher cortical areas to the subcortical as observed by Vahdat and colleagues in their experiments.
Since we`re talking about the topic……………………
…..it is said that new information can be acquired by an individual during sleep. Would a repetition of the above experiment when new information is presented during the sleep phases (eg. introduction of a conditioning sound or smell) have an effect on the observed shift of activation pattern strength from the higher cortical areas to the putamen and cerebellum as observed by Vahdat and colleagues?
…caffeine is said to enhance memory consolidation, but has an adverse effect on sleep. Can we assume that the connectivity changes observed in this experiment would still occur, but there may be increased performance levels due to better memory retention?
…would an intervening task have an effect on the activation patterns observed by Vahdat and colleagues and the shift of activity strength to the putamen would still be present since stimuli are said to be more familiar after an intervening task?
…it is said that mindful meditation leads to less reliable memory recall. Does it have an effect on motor memory and would mindful mediation cause the same shift in activation strength of the cortical and basal ganglia areas observed by Vahdat and colleagues in their experiments?