Brain plasticity, memory and learning

“What could not be repeated at first is readily put together on the following day; and the very time which is generally thought to cause forgetfulness is found to strengthen the memory.” Quintilian^1.7

A third theory that has gathered good empirical validation over the decades states that a principal function for the existence of sleep is related to neuronal network plasticity (i.e., learning of any nature) and memory consolidation (Smith, 1996, Stickgold and Walker, 2005a, Rasch and Born, 2013, Hodor et al., 2014). As it is evidenced both from animal and human studies, after training on a certain motor or cognitive task (e.g., procedural or declarative memory, problem solving, etc.), sleep, either during a whole night or simply after a nap, significantly improves learning as compared to a similar period of wakefulness (Stickgold, 2005, Stickgold and Walker, 2005b). On the other hand, sleep deprivation may also hinder memory consolidation;

for example, SWS deprivation is detrimental for explicit learning of visuospatial content (Casey et al., 2016). Another proposed underlying mechanism refers to the importance of sleep for synaptic plasticity, either by strengthening of synapses activated during learning activities

(e.g., playing a musical instrument or a dance movement) or by a homeostatic synaptic downscaling during sleep after waking synaptic potentiation (Tononi and Cirelli, 2006, Cirelli, 2013, Cirelli and Tononi, 2015). In addition to structural neuronal plasticity, neurogenesis occurring in the dentate gyrus of the hippocampus could also be particularly boosted during sleep, a biological process that is significantly deteriorated under a sleep deprivation protocol (Tung et al., 2005, Meerlo et al., 2009).

After establishing the general importance of sleep for learning and memory consolidation in a causal way, the neuroscientific community aimed to discover the specific roles played by NREM and REM sleep pertaining different kinds of long-term memory aspects (Smith, 2001, Stickgold and Walker, 2005b, van der Helm et al., 2011). Contrary to the conception in which it is used in the common language, memory is not a unitary system but a rich constellation of different mental processes; as a consequence, long-term memory is usually divided into explicit or declarative (further subdivided into episodic and semantic;

consciously accessible and shareable) and implicit (further subdivided into procedural and emotional; unconscious and not shareable) (Rasch and Born, 2013, Casey et al., 2016). As it might be imagined, the neuroanatomical substrates underpinning different memory aspects are not the same, thus, motivating its separate study and their particular correlations to the different sleep phases (Stickgold and Walker, 2005b, Stickgold, 2013).

In the case of procedural memories, both motor (e.g., learning a gymnastic exercise) and cognitive (e.g., the procedure to solve a geometry problem) types precise REM sleep for appropriate learning and behavioral efficiency (emotional memories regulated by the amygdala have also been linked to REM sleep), whereas learning of declarative memories are predominantly influenced by NREM sleep (Smith, 2001, Meerlo et al., 2009, Rasch and Born, 2013). Among the brain structures associated with each memory aspect, for implicit memories, cognitive procedural memory include the striatum and the prefrontal cortex, cognitive motor memory involve the primary motor cortex, supplementary motor area and cerebellum, and emotional memory relates principally to the amygdala (e.g., important for

aversive conditioning learning), whereas for explicit memories, the hippocampus and surrounding structures (entorhinal cortex, parahippocampal gyrus and thalamus) assume an essential role (Smith, 2001, Born et al., 2006). In addition, experimental evidence provided by studies in rats, but probably also transferable to humans, suggests that the previously learned task is replayed during sleep in the same neuronal circuits used during waking (e.g., including an interplay between the hippocampus and the cortex during SWS) in order to be fully consolidated and enhanced (Eckert and Tatsuno, 2015, Farthouat and Peigneux, 2015).

Pharmacological studies have also proved useful to further clarify the causal relationship between sleep and memory consolidation (Mednick et al., 2013, Rasch and Born, 2013). In this regard, administration of antidepressant drugs (such as SSRIs, MAOIs, tricyclic antidepressants and others), which are known to significantly suppress the amount of REM sleep, do not cause later impairment in executive functions or declarative memory consolidation and recollection, but might disrupt the learning of procedural and emotional memories (Smith, 2001, Goder et al., 2011). Drugs may not only impair memory consolidation, but also enhance it; in particular, it has been shown that administering zolpidem (a GABA agonist hypnotic) following a verbal memory task (i.e., an explicit memory task) and an immediate nap, significantly increases spindle density and performance on subsequent testing, allegedly by causal influence on spindles for semantic memory (Mednick et al., 2013). The same study also did not find any causal influence on spindle enhancing (whose frequency and amplitude was not altered) for procedural motor memory, but in turn, showed a worsening of perceptual memory (as assessed by a texture discrimination task) by a concomitant shortening of REM sleep produced by the drug.

A deeper investigation of declarative memory learning in humans has revealed that SWS is the most relevant and beneficial sleep stage for its long-term consolidation and the progressive reduction of SWS with age is connected to the memorization and retrieval deficits observed in the elderly (Gais and Born, 2004, Ladenbauer et al., 2016).

Endogenous neurochemical agents critically involved in declarative memory consolidation include the acetylcholine neurotransmitter and cortisol hormone (as well as other glucocorticoids). The acetylcholine decline associated with NREM sleep might be necessary for better transfer of signal output from the hippocampus and entorhinal cortex to the neocortex, as well as inside the hippocampus for feedback communication between the CA3 and CA1 regions (thought to be related to memory replaying) (Gais and Born, 2004). However, during wakefulness, acetylcholine actually improves declarative memory encoding, either directly or indirectly, as this neurotransmitter is relevant for attention (Sarter et al., 2003). On the other hand, the endocrine hypothalamic-pituitary-adrenocortical axis (HPA) is crucially involved in memory encoding by the action of cortisol released from the adrenal glands, which facilitates memory formation (e.g., an animal succeeding to escape from a dangerous place or after experiencing a stressful situation in humans), but at the same time, hinders memory retrieval (Osborne et al., 2015). Cortisol levels, which follow a circadian cycle, diminish until a minimum during the first hours of sleep, precisely when slow wave activity is at its maximum, which in turn inhibits the HPA axis, an inverse relationship that creates an optimal environment for information transferring from the hippocampus to the cortex for long-term storage (Bierwolf et al., 1997). This causal relationship is evidenced by experimental manipulation of cortisol levels in human studies, as infusion of cortisol during SWS significantly impairs declarative memory consolidation by hindering hippocampal-neocortical communication (Lavenex and Amaral, 2000, Gais and Born, 2004).