Plasticity is a natural property of living organisms that is crucial for adaptation
and evolution. Over the last decades, the availability of sophisticated neuroimaging
techniques (in particular, functional magnetic resonance imaging (fMRI), and transcranial
magnetic stimulation (TMS)), has made it possible to explore in vivo the on-line functioning
of brain and its plasticity. However, the impressive visual impact of fMRI and the
quick and reliable measures obtained by TMS have tended to divert attention from one
of the oldest neuroimaging techniques available in humans: electroencephalography
(EEG). Recent advances in EEG signal processing and recording technology have started
to reverse this trend with remarkable demonstrations in humans of phenomena previously
only recorded invasively in animals such as the electrical correlates of memory consolidation
during sleep (for a review see Tononi and Cirelli, 2012). EEG has the advantage over
fMRI in that it can explore real-time spontaneous electrical brain activity with an
excellent temporal resolution (milliseconds). Coupled with new forms of mathematical
signal analysis (signal coherence and perform fractal dimension, small world and entropy
analyses), EEG can now be used to investigate functional connectivity and to explore
how neural networks are organized and interact (an example in Zappasodi et al., 2014).
A number of approaches have been employed to capture the EEG signatures of neural
plasticity. In the awake state these are often masked by the large amount of unrelated
activity that occurs at the same time. Experimental strategies to increase the signal:
noise ratio include increasing the topographic specificity of tasks required to induce
plasticity and increasing spatial resolution of the EEG signal by using a large number
of electrodes. Another opportunity is offered by sleep, when spontaneous electric
neural patterns are predictable, allowing novel events to be easily characterised.
This was the approach used by Tononi and colleagues (2012), who demonstrated that
slow delta waves (DW) during non-rapid eye movement (NREM) sleep are associated with
memory consolidation. DW are the most prominent EEG feature of human NREM sleep. They
originate in cortical neurons and have been proposed as possible effectors of sleep-dependent
synaptic plasticity. Experiments in animals have demonstrated that DW are the EEG
counterpart of near-synchronous transitions between states of hyper- and hypo-excitability
over wide neural networks (Steriade et al., 1993), with the amplitude and slope of
DW correlated to the number of neurons oscillating near-synchronously (Vyazovskiy
et al., 2009). This synchrony is, in turn, directly related to the number and strength
of synaptic connections among neural networks. In humans, high-density EEG studies
revealed that there was an increase in parietal cortex DW after learning a visuo-motor
task; conversely DW activity was reduced over sensorimotor cortex if the contralateral
arm was immobilized during the day. In this case, synaptic depression was confirmed
behaviourally by reduced motor performance and physiologically by reduced amplitude
sensory evoked responses (Huber et al., 2006). Sensitivity of sleep DW to neural plasticity
was also corroborated by demonstrating changes in DW after the induction of long-term
potentiation (LTP) and depression-like phenomena in the brain following repetitive
TMS.
Brain plasticity is affected in a number of pathological states. In particular, large
acute lesions, such as after stroke give rise to a state when the brain devotes much
of its energy to induce brain plasticity. In EEG studies, the best experimental model
to investigate plasticity is after unilateral monolesional stroke, particularly if
restricted to a confined area involving eloquent cortex, i.e., a cortex producing
a measurable clinical deficit after its lesion (e.g., middle cerebral artery territory
with upper limb sensory/motor deficits). In such cases it is then possible to take
into account lesion size, clinical impairment and EEG changes of the affected cortex
throughout the subacute and chronic phases of stroke. Additionally, EEG changes specific
to the affected cortex can be compared with activity of contralateral non-lesioned
hemisphere (even if non-lesioned areas activity can, in turn, be partially influenced
from modified transcallosal inputs originating from the lesioned one). Clinical studies
on acute stroke patients have demonstrated that DW of the affected hemisphere are
a very sensitive indicator of neuronal dysfunction that is related to both the lesion
volume and the acute neurological deficit (Assenza et al., 2009). Thus, spread of
DW from the affected to unaffected hemisphere is associated with poor prognosis (Finnigan
et al., 2008). Indeed, EEG or magnetoencephalography measures of DW activity in acute
stroke can provide additional predictive value of clinical recovery at 3 months over
and above that given by clinical examination alone (Assenza et al., 2013). Advanced
EEG signal analyses (the coherence analysis) elucidated the pathophysiology of DW
arising from the unaffected hemisphere. They come from an interhemispheric communication
breakdown of electrical signals between the two hemispheres that can interfere with
the unaffected hemisphere contribution to plasticity processes sustaining clinical
recovery (Graziadio et al., 2012; Assenza et al., 2013). These results confirmed old
animal experimental data reporting that EEG DW detection requires an intact cortex
(Steriade et al., 1993), as they appear, in stroke models, only when a subcortical
lesion occurs and not after a cortical lesion, which sets almost to zero the neural
activity.
Although DW during wakefulness in stroke patients is usually regarded as a negative
sign, the studies above suggest that they may also have a role in cortical plasticity
and stroke recovery. The frequency of oscillatory activity in a neural network is
inversely related to the number of synchronously active neurons. Thus the population
of neurons that oscillate in the delta rhythm is larger than participates in the alpha
rhythm. We postulate that neural plasticity in perilesional areas increases connectivity
between neurons, and thus increases the chance that they synchronise their activity
at the delta frequency (
Figure 1
).
Figure 1
Delta waves and brain plasticity.
In physiological conditions with preserved interneuronal connections, background electroencephalography
(EEG) activity oscillates in alpha frequency (8–12 Hz) band (A). When cortical plasticity
is locally stimulated (as in Assenza et al., 2014) new connections are created and
larger groups of neurons start oscillating simultaneously generating delta waves (B).
When a lesion hit the brain, spared neurons on the boundaries of the lesion try to
connect each other according to novel networks in the attempt to overcome the occurred
neuronal deficit and start oscillating in delta frequency band.
In order to test this possible relationship between DW and plasticity we used intermittent
theta burst (iTBS) TMS to manipulate levels of cortical plasticity in healthy volunteers
(Assenza et al., 2015). This method has been shown to transiently increase cortical
excitability for up to 30 minutes after administration (Di Pino et al., 2014). We
could therefore test whether increased plasticity in the healthy brain led to a period
of increased DW activity in the EEG. As predicted we found that iTBS (but not sham
iTBS) increased the power of DW, thus confirming in wakefulness the concept from Tononi's
experiments during sleep showing an association between synaptic consolidation and
DW. Thus, DW in lesional and controlateral areas of stroke patients may be not merely
a marker of network dysfunction, but more a sign of neuronal rearrangement accompanying
acute and chronic phases of recovery. Admittedly, the data so far only show correlations
between plasticity and DW activity and further complementary experiments are planned
to try to answer this question. In particular, present data do not permit to establish
the exact relationship between DW and plasticity (because of the lack of a correlation
between motor evoked potential amplitude and DW power and of a temporal segregation
of DW and motor evoked potential increase), but clearly reveal their physiological
association. Animal studies provided further data supporting an active role of DW
in the awake state. In stroke rats, Carmichael and Chesselet (2002) demonstrated that
DW in the controlesional hemisphere might function as an attraction guide for interhemispheric
fibers sprouting.
We note that several studies have changes in alpha and beta band activity in association
with learning and plastic phenomena, but the effects are smaller and less consistent
than those reported above. However, the implication is that plasticity phenomena may
manifest other changes in addition to those in the delta band EEG. Nevertheless, we
propose that data from delta band activity has huge potential which should be explored
in more detail to confirm its involvement in processes related to neural plasticity.
Our hypothesis is that DWs are a non-invasive and easily-recordable correlate of neural
plasticity that open new and exciting scenarios in neurological rehabilitation as
well as in a wide context of learning programs. It could also provide an objective
measure of obtained learning.
In conclusion, EEG is a reliable technique for exploring functional activity of the
brain and is sensitive to changes in neural plasticity processes related to LTP-like
phenomena and clinical recovery after brain lesions. Its use in this field of neuroscience
should be encouraged, in particular in searching traces of plasticity processes to
provide a biological marker to lead improvements of brain functioning.