**3. Neurophysiological Changes in Wide-Range Networks during Aging**

#### *3.1. TMS Studies*

Long-range connectivity between different brain areas can be tested with TMS by delivering pulses with two different coils at specific ISIs. The most robust connectivity protocol is interhemispheric inhibition (IHI), which is obtained by conditioning a MEP with a preceding suprathreshold TMS pulse on the contralateral M1. IHI probably generated by local inhibitory interneurons within M1, which activate GABA-B receptors on principal cells [115] and are activated by interhemispheric excitatory pathways passing through the corpus callosum [116–118]. IHI is most pronounced at ISIs of 8–10 ms and 40–50 ms, which are referred to as short and long latency IHI, respectively (SIHI and LIHI). Most studies, but not all [119,120], testing IHI at rest, revealed that both SIHI [121,122] and LIHI [123–125] did not differ between older and young subjects. By contrast, there is evidence that elderly subjects have less lateralized cortical activation during various motor tasks, including hand grip [126]. In neuroimaging studies, young subjects demonstrate a decreased activity in the

M1 ipsilateral to the moving hand (iM1) during task, while this pattern is cancelled or even reversed in older subjects [127]. TMS research have further explored this phenomenon and found that LIHI from the contralateral M1 (cM1) to iM1 decreased during motor tasks in older compared to young participants, suggesting a reduced inhibitory drive from cM1 to iM1 with aging [121,128]. By combining neurophysiological and neuroimaging data, the authors clarified that the reduced inhibition of the iM1 underlies a progressive involvement of this area during simple motor tasks with aging [128]. Another study demonstrated that the iM1 also contributes to motor preparation in elderly. Indeed, TMS was delivered over iM1 before an index finger movement, and this caused delayed motor responses in older, but not in young, adults [129]. Besides IHI, the interaction between the two M1 can be functionally assessed by evaluating the ipsilateral silent period (ISP), defined as the interruption of voluntary electromyographic (EMG) activity induced by suprathreshold TMS applied in the iM1. ISP is thought to be mediated by transcallosal inhibition between the stimulated and the pre-activated cM1 [116,130]. A number of studies have reported changes in ISP in older subjects, such as delayed onset and decreased depth or area [131–134]. Taken together, findings on IHI and ISP indicate a decline in interhemispheric inhibition with increasing age, with recent evidence that this decline is uniform across the lifespan [134].

In addition to the interactions between the homologous M1, IHI can be extended to a widespread inhibitory system projecting from various cortical areas, including the dorsolateral prefrontal cortex, dorsal premotor cortex and somatosensory cortex, to cM1 [118]. In this regard, one study tested whether physiological aging alters the functional connectivity between the left dorsal premotor cortex (PMd) and right M1 [124]. Paired-pulse TMS was delivered immediately before a simple left index finger movement in young and older subjects, and the data showed that only in the latter group there was a facilitatory interaction between left PMd and right M1. Moreover, the degree of modulation was associated with faster responses [124]. A more recent study combined LIHI with neuroimaging and behavioral measures to assess interhemispheric connectivity between DLPFC and cM1 during the preparation of a complex bimanual coordination task in aging [125]. Interestingly, it was found that the ability to disinhibit functional connectivity between DLPFC and cM1 was impaired in older subjects, and this alteration was paralleled by decreased bimanual performance.

Overall, based on TMS data, it may be argued that the increased activity in ipsilateral M1 and premotor regions before/during simple movements reflects the involvement of additional areas as an attempt to preserve normal motor performances despite advancing age [128,135,136]. Conversely, the altered modulatory activity of DLPFC to contralateral M1 may underlie the decline of bimanual performance in older subjects [125].

#### *3.2. TMS-EEG Studies*

As previously stated, recording of EEG activity during TMS provides the possibility to non-invasively and directly probe brain connectivity [89]. Particularly, it has been suggested that the first part of the TMS-evoked EEG response reflects local excitation of the stimulated cortex (see Section 2.2), whereas the spatiotemporal distribution of later TEP components over the scalp reflect the activation of distant cortical areas, either via cortico-cortical connections or projections from subcortical structures [137–140]. Recent evidence has suggested that age and neurodegeneration influence late TEP components [12,102,105,114]. Among the late TEPs described in the EEG signals evoked by M1 stimulation, the N100 is the dominant negative peak and it has been related to the GABA-B-ergic neurotransmission [100,113]. In the elderly, its scalp distribution and source activation has been demonstrated to be significantly different from younger subjects, suggesting hypoexcitability in prefrontal and premotor cortices of the stimulated hemisphere, coupled with hyperexcitability in the median anterior EEG channels [102]. The neural generators of late TEP components are not entirely clarified. In previous studies [100], it had been suggested that they could be related to the activity of reverberant cortico-cortical as well as cortico-subcortical circuits driven

by GABA-B neurotransmission, and finally re-engaging the stimulated M1. Age-related differences in spatial distribution of the N100 have been confirmed by a more recent study: here, by using a LICI protocol, the authors demonstrated that the paired-pulse inhibitory effects on N100 wave are increased in older adults, thus suggesting a potentiation of preand post-synaptic GABA-B-mediated inhibition [105].

Previous TMS-EEG studies showed that TMS-evoked EEG signals strongly depend on the brain state at the time of stimulus delivery [94]; this can be determined by excitability of local circuits, or by the activity of diffuse neuromodulatory systems [141,142]. In agreement with this notion, features of EEG rhythms preceding a TMS pulse applied on M1 have been shown to influence MEP amplitude [143,144], and this process changes with aging [142,145]. It is known that MEP amplitude shows a degree of inter-trial variability, which depends on several factors, including fluctuations in excitability of cortical and spinal neurons [146,147] and corticospinal connectivity [148]. TMS-EEG allows to verify whether MEP amplitude variability also depends on cortico-cortical connectivity changes. In this regard, Ferreri and colleagues (2014) found that, in young subjects, MEPs are significantly larger when the ipsilateral M1-prefrontal cortex coherence in the beta−2 band and the ipsilateral M1 parietal cortex coherence in the delta band are high. However, elderly subjects showed higher M1-parietal cortex delta coherence than young participants, and this measure was unrelated to MEP size variations [145]. Since the delta rhythm may underlie functional disconnection between areas [149,150], the results of this study possibly reflect functional unbinding of M1 from the somatosensory cortices' inhibitory control. This mechanism may be compensatory to age-related decrease in cortical excitability and motor functions [145].

#### **4. Neurophysiological Changes in Plasticity and Metaplasticity Processes during Aging**

Physiological aging is characterized by a weakening of different brain functions, mainly linked to neuroplasticity processes, such as learning and memory [1]. In humans, various TMS-based protocols allow the assessment of synaptic plasticity mechanisms in a non-invasive way; the most widely used are paired associative stimulation (PAS) and thetaburst stimulation (TBS). PAS is based on associative synaptic plasticity [151] and its effects probably reflect spike-timing dependent plasticity, where the precise timing of pre- and postsynaptic neurons firing is crucial for the direction of long-lasting changes. If the inter-spike interval is positive (pre- before post-synaptic action potential), LTP occurs, whereas if the interval is negative (post- before pre-synaptic action potential), LTD is elicited [152–155]. In humans, the PAS protocol implies the combination of repetitive cortical TMS and peripheral nerve stimulation [156,157]. If M1 stimulation occurs around 25 ms after the electric median nerve stimulation at the wrist (PAS25), MEP amplitude is increased for 30–60 min (LTPlike effects), whereas if the inter-stimulus interval is shorter, i.e., 10 ms (PAS10), MEP amplitude is decreased (LTD-like effects) [156–158]. TBS is based on evidence in animal models demonstrating that high-frequency bursts of stimuli rhythmically delivered in the theta frequency range transiently modulate hippocampal neuronal firing [159], and that LTP/LTD-like changes can be recorded by measuring changes in post-synaptic responses following the stimulation [155,160,161]. The classical TBS paradigm consists of bursts of three TMS pulses at 50 Hz, repeated at 5 Hz [162,163]. If the pattern of stimulation is intermittent (iTBS), i.e., short trains of 2 s given every 10 s, cortical excitability is enhanced and MEP amplitude increases up to 30 min after stimulation (LTP-like effects). If the pattern is continuous (cTBS), i.e., bursts given continuously for 40 s, cortical excitability is inhibited and MEP amplitude decreases for 20–60 min (LTD-like effects) [155,163–166].

A recent meta-analysis on NIBS studies in the aged population suggested that there is a general trend towards decrease in motor cortex plasticity, with a certain degree of variability between different studies and different plasticity-inducing protocols [8]. In 2008, two reports by different groups showed, for the first time, that PAS-induced LTP-like plasticity of M1 may deteriorate with physiological aging [167,168]. Müller-Dahlhaus and colleagues verified the effects of PAS in a cohort of 27 subject with variable age (range: 22–71 years) and found that the magnitude of PAS effects was negatively correlated with age, with a smaller

MEP facilitation in elderly subjects [167]. A direct comparison of PAS effects between young and aged subjects was provided by Tecchio and colleagues (2008). Although it was somewhat confirmed that the long-lasting increase of M1 excitability after PAS is weaker in older than young participants, this effect was clearly driven by the female population [168]. The authors pointed to a possible impairment in intracortical excitatory network activity due to hormonal changes during menopause, a hypothesis which was confirmed in a later research [169]. In a following study, a larger number of healthy subjects was enrolled and divided in three groups based on age (young: 21–39 years, middle: 40–59 years, elderly: 60–79 years). The expected PAS-induced facilitation of MEP amplitude was observed in the young and middle groups, but not in the elderly group, further confirming the impaired LTP-like plasticity in M1 with aging [170]. Interestingly, age-related decline in response to PAS has been demonstrated to be restored by L-dopa [171], a finding that suggests that the alteration in PAS response observed in the elderly might have a functional, rather than a structural substrate. In contrast with PAS, only few studies investigated the influence of aging on M1 plasticity induced by TBS. Dickins and colleagues compared MEP amplitude changes after iTBS over the dominant M1 between 20 young (18–28 years) and older subjects (65–76 years). In contrast to PAS studies, M1 excitability increased in a comparable way between the two groups after iTBS [172]. Similar findings were also obtained in more recent research, indicating that iTBS-induced LTP-like plasticity of M1 is not affected by aging [173,174].

The different neurophysiological findings on age-related synaptic plasticity changes obtained using PAS and TBS may depend on the different types of plasticity mechanisms elicited by the different protocols, as described above in this section [155,157,163,175,176]. Moreover, PAS acts through a combination of sensory input and direct cortical stimulation activating the same M1 neurons, a process which reflects heterosynaptic plasticity due to sensorimotor interaction [155,157,177,178]. Conversely, TBS operates through the repetitive activation of the same synapses by M1 stimulation alone, which reflects homosynaptic plasticity mechanisms [155,163,176]. Impaired PAS-induced and normal TBS-induced effects may therefore suggest that synaptic plasticity processes are not diffusely impaired by physiological aging. Rather, the alteration is restricted to processes requiring the activation of specific intracortical circuits and/or sensorimotor interaction mechanisms.

The direction and amount of synaptic plasticity can be influenced by neuronal activity occurring immediately before or during the induction of plasticity [179,180]. These features of plasticity can be framed in the context of metaplasticity, which can be shortly defined as "the plasticity of synaptic plasticity"; this involves a wide range of mechanisms and, from a behavioral point of view, has an important role in the regulation of important brain functions, including memory and learning [27,180,181]. Metaplasticity in humans can be explored through different protocols, from priming an exogenous or endogenous plasticity-inducing protocol with NIBS, to delivering plasticity-inducing protocols with longer duration [64,73,75]. For instance, Opie and colleagues (2017) delivered iTBS 10 min after the application of sham TBS (sham TBS + iTBS), cTBS (cTBS + iTBS), or iTBS (iTBS + iTBS) in young and older participants. The results showed that, whereas priming iTBS with either cTBS or iTBS boosted M1 plasticity in young subjects, MEP facilitation after sham TBS + iTBS did not differ from iTBS + iTBS, and was even larger than cTBS + iTBS, in the older group [173]. Similar findings were also found using PAS: priming the stimulation with PASN20+2ms caused enhancement of plasticity in young but not in older subjects [182]. In another study by the same group, a visuo-motor training task was performed after facilitatory, inhibitory or sham PAS in young and older adults. While the baseline level of motor skill did not differ between sessions in young subjects, priming with PAS had a detrimental effect on skill acquisition in older ones [183]. Recent evidence suggests that the amount of NIBS-induced plasticity of M1 changes by concurrently modulating cortical gamma oscillations through transcranial alternating current stimulation (tACS), and this effect has been interpreted as reflecting gating phenomena [18,184–186]. However, gamma-tACS has been demonstrated to boost LTP-like plasticity induced by iTBS to a

larger extent in young than older adults and, in the latter group, the effect of gamma-tACS decreased with increasing age [174]. In summary, unlike synaptic plasticity mechanisms, which seem to be altered only in part by the aging process, the existing evidence point to a higher susceptibility of metaplasticity processes by physiological aging.

#### **5. Limitations, Perspectives and Conclusions**

The present review offers a summary of data obtained with TMS, either coupled with EMG or EEG, about physiological brain aging. Overall, the findings suggest that TMS can offer valuable insight into several functional derangements occurring throughout the life span, including a trend of decrease in brain excitability, altered long-range cortico-cortical connectivity and impaired associative plasticity and metaplasticity processes (Figure 1). There are, however, a number of caveats pertaining to acquisition of TMS-EMG and TMS-EEG variables, as well as their interpretation, which we deem important to discuss in this final section. The first, general issue is that, despite the relatively high number of studies investigating cortical excitability and plasticity during aging, their sample sizes are generally small. This factor would determine low statistical power and may, at least in part, explain the variability of results observed in the literature [8]. A second problem is related to the degree of cerebral atrophy which accompanies physiological aging. As this causes an increase in the distance between scalp and coil, TMS measures are necessarily affected by it [37,38]. Even the most reliable among them, such as the RMT [34], could yield limited information in a context where anatomical data are not available. Therefore, when possible, it is important to obtain structural brain information, along with neurophysiological assessment. In absence of the former, spTMS measures could still be able to track within-subject longitudinal changes, but the pathophysiological/clinical value of such follow up still need to be established. Paired-pulse TMS measures partially solve the confound represented by decreased brain volume, since they are usually calculated as ratios between conditioned and unconditioned MEPs. However, ppTMS protocols are affected by variability and reliability issues of experimental paradigms and output measures [175,187]. For instance, SICI considerably varies between individuals, even if the same CS intensities and ISIs are tested [66,188]. Therefore, to compare results across studies, it would be useful to explore a range of ISIs/CS intensities in large cohorts of subjects, so as to obtain data about maximal effects and thresholds. Moreover, since the putative circuits tested by different ppTMS protocols can interact [72], the interpretation of the effects of CS is not always straightforward. For instance, since SICI acts by suppressing I waves [189], an hypersynchronized and/or hyperexcitable state of excitatory M1 interneurons may secondarily result in SICI decrease [70,71,190]. This is possibly the case in amyotrophic lateral sclerosis [191] and Parkinson's Disease [13,17,192–194]; whether a similar scenario occurs in physiological aging is not known and is probably worth exploring.

The possibility of adding simultaneous EEG recording has substantially increased the range of variables that can be tested with TMS, thus expanding the amount of information that is possible to obtain. However, the nature of TMS-EEG signals has not been completely clarified yet. For instance, the information given by TEPs obtained from M1 TMS may not completely overlap with those provided by MEPs; indeed, the latter arise from excitation of PTN and associated circuitry, whereas the former probably reflect activity of a larger ensemble of cortical cells [89,100,195]. Therefore, caution should be used when comparing conclusions drawn by the two variables, especially for ppTMS protocols, which were devised for MEPs, and are still of uncertain interpretation in the TMS-EEG setting [112,196,197]. Another issue is related to the use of TEPs to measure brain connectivity. This is usually performed by measuring the spatial distribution of specific TEP components; however, this alone does not take into account volume conduction, and has not been assessed in conjunction with more common measures of EEG connectivity [198]. In addition, TEPs, especially in their late components (N100, P180), can be contaminated by EEG responses generated by indirect brain activation due to the somatosensory and auditory stimulation intrinsic to TMS, if adequate countermeasures are not properly taken

(e.g., suppression of the TMS click by the use of a masking noise, use of ear defenders, application of a foam layer under the coil) [99,199]. This should lead to careful review of past studies where sensory input by TMS was not properly masked; in particular, it should be noted that effective masking procedures were only seldom used in studies involving older adults. This should prompt to strict control of these confounding factors in future work.

**Figure 1.** Age-related changes in local motor circuits, wide-range networks and plasticity processes RMT: resting motor threshold; I/O curve: input-output curve; SICI: short-interval intracortical inhibition; ICF: intracortical facilitation; LICI: long-interval intracortical inhibition; SICF: short-interval intracortical facilitation; SAI: short-latency afferent inhibition; M1: primary motor cortex; TEP: transcranial evoked potential; SIHI: short-latency interhemispheric inhibition; LIHI: long-latency interhemispheric inhibition; cM1: contralateral M1; iM1: ipsilateral M1; ISP: ipsilateral silent period; PMd: dorsal premotor cortex; DLPFC: dorsolateral prefrontal cortex; PAS: paired-associative stimulation; iTBS: intermittent theta burst stimulation; cTBS: continuous theta burst stimulation; tACS: transcranial alternating current stimulation.

> In conclusion, we believe that TMS and TMS-EEG can give an important contribution to the understanding of the mechanisms underlying physiological brain aging, provided that technical pitfalls and interpretation biases are considered. Future studies should seek to integrate electrophysiological and structural data and clarify how these relate to impairment of daily activities in the elderly population, with the ultimate goal of reliably distinguishing physiological and compensatory processes from disease.

> **Funding:** This work was partially supported by the Italian Institute of Health, Grant GR-2016- 02361802.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Conflicts of Interest:** The authors declare no conflict of interest.
