**2. Neurophysiological Changes in Local Motor Circuits during Aging** *2.1. TMS Studies*

In the last decades, TMS has been extensively used to investigate the physiology of brain aging in a safe and non-invasive way, particularly with regards to the sensorimotor system. It is well known that a single TMS pulse results in multiple descending volleys, i.e., an early direct (D) followed by indirect (I) waves [29]. The physiological mechanisms underlying the generation of I-waves are still unclear and different hypotheses have been made, ranging from oscillating activation of corticospinal tract (CST) neurons, to reverberation of activity within interneuronal circuits and converging pyramidal output cells [30–32]. The motor-evoked potential (MEP) arises from the temporal summation of these descending volleys at the spinal level, but its basic characteristics (i.e., latency and amplitude) also reflect the combination of excitatory and inhibitory events occurring in a more complex synaptic network, at different levels of the motor pathway [31,33]. The easiest and most reproducible single-pulse (sp) TMS measure, used to probe the excitability of M1, is the resting motor threshold (RMT), which is defined as the stimulation intensity required to elicit MEPs in resting muscles of at least 50 µV in 5 out of 10 trials [30,34]. The RMT can be considered as an estimate of global cortical excitability, it reflects axonal excitability, likely of the intracortical elements activated by TMS, and it has been reported to have high test-retest reliability in healthy older adults, in particular when using monophasic rather than biphasic TMS pulses [35]. It is sensitive not only to cortical excitability, but also to the scalp-to-cortex distance, as this alters the amount of energy required to bring CST neurons to threshold [36]. Given that age-related cortical atrophy can increase the coil-to-cortex distance, the RMT can be used as a measure to explore brain aging [37,38]. Indeed, several studies reported an increase in RMT in older adults [39,40]; however, this has not been replicated by other authors [41–43]. This inconsistency can be due to the possibly different degree of individual cortical atrophy in elderly subjects and to the lack of a structural neuroimaging assessment in the various studies, which would allow a coil-to-cortex distance measurement in all participants and, accordingly, an ad-hoc RMT normalization. Another hypothesis proposed to explain the lack of consistency in agerelated differences in RMT is that this variable changes in discrete stages during the life span. Supporting this idea, Shibuya and colleagues (2016) recently evaluated the RMT in 113 subjects with an age ranging from 20 to 83 years; they found that the RMT increases

from age 20 to 50 and then decreases, roughly following a quadratic function [44]. Up to date, a long-term longitudinal RMT follow up is missing; this could potentially have more value, compared to cross-sectional measurements, in clarifying electrophysiological correlates of the gray matter volume loss in the aging brain.

In addition to RMT, the stimulus-response relationship may be intrinsically altered in the aged motor cortex. It is well known that the MEP amplitude increases sigmoidally with stimulation intensity, and that voluntary activation of the target muscle shifts this inputoutput curve to the left [45,46]. Pitcher and colleagues (2003) explored the input-output curve in young and elderly subjects and demonstrated a rightward shift in the latter group, due to the higher stimulation intensity necessary to achieve the maximum MEP. The slope of the curve and the RMT were not different between the two groups [47]. These data have been interpreted as reflecting a reduction in the amount of spinal motoneurons activated by TMS or in the stimulus-induced neuronal synchronization with aging [47,48].

Besides spTMS, more complex TMS protocols, which consist in delivering two stimuli via the same coil (paired-pulse TMS; ppTMS) [49,50] or by coupling a TMS pulse with peripheral electric stimulation [51] have been also applied in older subjects. While spTMS measures generally reflect the degree of the overall corticospinal excitability [52], these protocols allow to investigate the physiology of intracortical circuits which rely on different neurotransmitter classes [30,53,54]. More in detail, in ppTMS a conditioning stimulus (CS) precedes a test stimulus (TS) by an interstimulus interval (ISI) which varies according to the protocol used, allowing the evaluation of intra-cortical inhibitory and facilitatory circuits. Short-interval intracortical inhibition (SICI) is elicited when a subthreshold CS is followed by a suprathreshold TS with an ISI of 1–6 ms and likely reflects inhibition mediated by GABA-A receptors, while intracortical facilitation (ICF) entails longer ISIs (6–30 ms) and is likely mediated by NMDA receptors [30,49,53,55]. The investigation of ppTMS measures in physiological aging has produced inconsistent results. SICI has been reported to be either decreased [41,56–59], normal [42,43,60,61] or even increased [39,62] in elderly. Differently, ICF was found to be generally normal in the majority of [41,44,58,62], but not all [39,61], studies, suggesting intact NMDA-related glutamatergic neurotransmission in these subjects. Long-interval intracortical inhibition (LICI) is another ppTMS measure, which refers to inhibition of a test MEP using a suprathreshold CS applied 50–200 ms before, and likely depends on GABA-B receptors [53,63,64]. Similar to what observed with SICI, the effectiveness of LICI was found to be variable with aging, being either increased [39] or decreased [43,58] in different studies. Thus, ppTMS researches have not yet clarified whether GABA-ergic neurotransmission is generally preserved or impaired in physiological aging. Finally, short interval intracortical facilitation (SICF) can be elicited by a suprathreshold first stimulus followed by a second stimulus at about RMT [65,66], by an ISI compatible with I-waves period (i.e., 1.3–1.5 ms). In this protocol, the observed increase in MEP is likely due to summation of different I-waves at the M1 level and mainly reflects the activity of non-NMDA glutamatergic neurotransmission [17,31]. Clark and colleagues (2011) described a higher SICF in elderly than young subjects at ISI 1.5 ms, while the opposite result was present at ISI 2.5 ms. These apparently contrasting data confirm that physiological mechanisms underlying the first and later SICF peaks differ [17,31,67,68] and may suggest that they have different sensitivity to the aging process [69]. More recently, however, Opie and colleagues tested SICF in young and older adults in more detail and compared I-wave intervals that were optimal for I-waves summation in each group. Although they confirmed the reduction in the second SICF peak in the elderly, a similar change was also found in the other peaks, thus not corroborating the previous hypothesis and leaving the question open [70,71]. Additionally, these studies found a delayed latency of the third SICF peak with aging, which influences both plasticity induction and motor function [70,71].

Discrepancies in the effects of ppTMS protocols across studies can have several explanations. The first could be the choice of the ISI due the fact that paired-pulse protocol is associated with a relatively wide range of possible ISIs, and the timing optimal to obtain

the predicted changes in MEP may be different across subjects [30]. While some of ppTMS studies have used multiple ISIs [41,42,44,58], the majority employed only one ISI for all subjects [39,43,56,57,59–61]. As different subjects will probably manifest optimal responses at different ISI, the outcome of future experimental studies could be improved by the use of paired-pulse curves which may decrease the variability and improve the outcome in future studies. Furthermore, concerning SICI, many studies tested this measure only using an ISI of 3 ms, which, according to some reports [72], may be influenced by SICF; however, this has not been universally confirmed [66]. Another important factor contributing to the inconsistency of ppTMS protocols is the choice of CS intensity, which can be calibrated either based on the RMT, or the active motor threshold (AMT). The intensities used vary considerably, usually ranging from 50% to 95% RMT [39,41,42,44,55,56] or from 70% to 90% AMT [73–75], and this variability may contribute to divergent findings in SICI studies. Another point to take into account is that ppTMS measures may have different sensitivity to coil orientation [66,76], and that this pattern of susceptibility to current direction might change according to the age of subjects. In this regard, Sale and colleagues (2016) assessed SICI in young and older adults with TMS coil in posterior-to-anterior (PA) and, then, anterior-to-posterior (AP) position. Interestingly, they found no age-related differences in SICI with PA-directed TMS, while SICI was more effective in older than young subjects using AP-directed TMS [77]. Unfortunately, both the CS and TS were delivered with the same current direction, thus making it difficult to understand which one drives the effect. A further important issue is that ppTMS measures are known to be modulated by different brain states, including motor tasks [78,79], and that this change may be dependent on age. For instance, the amount of SICI during a reaction time task was related to manual dexterity in aging subjects [57]. Moreover, while SICI increases during motor stopping in young subjects, this phenomenon does not occur in older adults [80]. Analyzing taskrelated changes in ppTMS measures may thus increase their sensitivity in investigating possible age-related decline in intracortical motor circuits. Overall, future studies should standardize the methodology of ppTMS to help comparisons across different age groups.

Differently from ppTMS, short-afferent inhibition (SAI) has shown more consistent alterations in the aging brain. SAI is elicited by electrical stimulation of a peripheral nerve preceding contralateral M1 TMS by an ISIs slightly longer than the latency of the N20 component of the somatosensory evoked potential. This form of inhibition likely reflects intracortical sensorimotor interaction and involves cholinergic as well as GABA-A-ergic circuits [51,81–83]. Although a first study found that SAI was comparable between older and younger adults [42], later studies showed decreased SAI in the elderly [40,84–86], which was also correlated with the age of subjects [87]. These data overall point to a progressive impairment of cholinergic activity during aging.

#### *2.2. TMS-EEG Studies*

TMS-EEG allows to measure the perturbation induced by a TMS pulse on cortical activity, both at the local and network level. Thanks to its high temporal resolution, TMS-EEG can provide meaningful clues on the functional properties of cerebral circuits in physiological and pathological conditions [88–92]. Importantly, TMS–EEG allows the assessment of cortical physiology by discerning causal interactions from pure temporal correlations [93,94]. TMS-evoked EEG responses were first measured in the late nineties [95] and, subsequently, scalp topography and possible generator sources of TMS-evoked potentials (TEPs) have been described. It is now clear that TEPs represent genuine responses of the cerebral cortex to TMS, provided that indirect sources of brain activation, such as somatosensory and auditory stimulation, are appropriately reduced or suppressed [96–99]. TEPs occurring in the first 20–40 ms after the TMS pulse most likely reflect the responses of local cortical circuits, whereas longer-latency responses involve more distributed networks [100]. TMS-EEG studies which aimed to investigate age-related modifications in M1 generally showed decreased excitability, a finding which supports the outcomes of MEP investigations [42,47,101]. However TMS-EEG yields complex results, based on the metric

used for analysis [102]. For instance, the global-mean field power (GMFP), which provides a measure of the TMS-induced brain response on the whole scalp [103,104], has been shown to be smaller in the elderly, pointing towards a generalized cortical hypoexcitability. This is in agreement with analysis of single TEP peaks, which showed a decrease in the local P30 amplitude after M1 stimulation; however, an opposite trend was found in the ipsilateral prefrontal areas, reflected in an increased TEP amplitude. This has been interpreted as a true hyperexcitability of the prefrontal cortex, rather than a compensatory phenomenon to M1 hypoexcitability [102]. The result on P30 has not been confirmed by another study, which failed to identify an amplitude difference between groups, suggesting instead a decreased P30 latency in elderly subjects [105].

TMS-EEG allows the investigation of specific cortical neurotransmitter systems, with the assumption that different TEP components reflect selective activity of receptor classes. Data obtained in humans and non–human primates point to age-dependent changes in GABA-A and GABA-B receptor density and subunit composition, particularly in frontal cortices [106,107]. In particular, while GABA-ergic neurotransmission seems to become more efficient during the first years of postnatal development in the dorsolateral prefrontal cortex (DLPFC) [108], there is evidence for GABA-ergic neurons activity impairment in animal and human motor areas with aging [57–59,109–111]. Previous TMS-EEG data have found that the P30 and N45 peaks are related to GABA-A neurotransmission, whereas activity of GABA-B receptors shows a stronger link with later peaks, both in healthy and in diseased brains [82,90,100,112,113]. The amplitude of the N45 has been shown to be modulated by aging, albeit in opposite directions in two studies. It was reduced in Ferreri et al. [102], a data which supported a GABA-A-ergic neurotransmission deficit in M1 in older subjects, and increased in a following study [105]. As a further confounding factor, both P30 and N45 amplitude did not change when SICI, which measures GABA-A activity, was obtained by stimulation of DLPFC [114]. While the cause of these contradictory findings is currently unclear, it may reflect the variability observed for SICI in TMS studies in aging (see Section 2.1). Importantly, to date, no study has assessed SICI in M1 with concurrent EEG recording in older adults; this would help in better elucidating possible age-related changes in GABA-A mediated inhibition in the motor system. Only few studies used other ppTMS protocols during EEG recording in the elderly. Noda and colleagues (2017) tested NMDA receptor-mediated glutamatergic activity in the DLPFC using ICF; when comparing the results with a group of healthy controls, they found a reduction in N45 amplitude. The authors speculated that this result would reflect a functional decline in glutamatergic excitatory neurotransmission in older adults [114]. Finally, a recent study recorded TEPs during M1 LICI and demonstrated a greater inhibition of the N45 wave in older, compared to younger subjects; this may suggest that this component is particularly sensitive in detecting modulation of GABA-B-ergic inhibitory circuits which occurs during aging [105].
