Mirror neurons and their relationship with neurodegenerative disorders

2.2.1 MN functions in humans

2.2.2 Inter-subjectivity, empathy, and social cognition

Experimental evidence strongly suggests that the MN system (along with other cerebral circuits, such as the social cognition network), contributes to understand not only other people's actions but also their emotions (Agnew, Bhakoo, & Puri, 2007; Gallese, 2003; Gallese, Keysers, & Rizzolatti, 2004; Pineda & Hecht, 2009; Wicker et al., 2003). In this sense, a crucial role of MN would be to foster empathy. Empathy is a subjective experience. It can be viewed as the process by which a person resonates and tunes to others’ affective and cognitive states (Levy &  Feldman, 2019). It plays a key role in social relations and allows humans to connect and understand each other and to create social and cultural groups. Studies on emotional states demonstrate that a large-scale neural network—including the ventral premotor cortex and the inferior frontal gyrus (two reputed MN areas), in addition to classical cerebral regions involved in feeling emotions, such as the anterior insula and the amygdala—is active for instance during facial expression observation and imitation (Budell, Jackson, & Rainville, 2010; Carr et al., 2003; Grosbras & Paus, 2006; Iacoboni, 2009; Rizzolatti & Craighero, 2004). In children, activity in the frontal part of the MN system, produced by observation and imitation of emotional expressions, correlates with measures of empathic behavior and interpersonal skills (Pfeifer, Iacoboni, Mazziotta, & Dapretto, 2008). Likewise, it is now proposed that the core deficits of autism, which are motor, language, and social impairments, are indicative of dysfunction of the MN system (Dapretto et al., 2006; Fishman, Keown, Lincoln, Pineda, & Müller, 2014), and various works suggest a role of MNs in schizophrenia (Kato et al., 2011; Saito et al., 2017; Schilbach et al., 2016). In contrast, as suggested by Linkovski, Katzin, and Salti (2017), caution in theoretical generalization of experimental procedures is needed. As an example, researches on MNs and schizophrenia do not report a correlation between social dysfunction severity and the abnormality of the MN network (Enticott et al., 2008; McCormick et al., 2012).

2.2.3 Language

Some studies have shown that MNs have a main role in different aspects of language acquisition (Theoret & Pascual-Leone, 2002), speech perception (Glenberg et al., 2008) and production (Kühn & Brass, 2008), and evolutionary language development (Arbib, 2008). Tettamanti et al. (2005) demonstrated that listening to sentences describing actions performed with the mouth, the hand or the leg activated a left fronto-parietal-temporal network that included the pars opercularis of the inferior frontal gyrus (Broca's area), the intraparietal sulcus, and the posterior middle temporal gyrus, in a somatotopic manner. Based on these studies, it has been suggested that cerebral areas establishing a correspondence among performed and observed actions in monkeys (F5Area) match to those that are appointed to language production in humans (Broca's Area; Molnar-Szakacs, Iacoboni, Koski, & Mazziotta, 2005, Tettamanti et al., 2005). However, a recent study raised doubts about this presumed matching (Cerri et al., 2015). Other neuroimaging studies revealed zones in the human premotor cortex that activate both when participants see actions being performed and when they read sentences concerning those actions (Aziz-Zadeh et al., 2006). Furthermore, Hauk, Johnsrude, and Pulvermüller (2004) found a somatotopically organized activation in the motor and premotor cortex when subjects passively read action-associated words (i.e., leg-associated action words lead to activations in medially situated areas while arm or face-associated action words lead to lateral areas activation).

More generally, the role of the MN system in evolutionary language development is recurrently proposed (Chwilla, Virgillito, & Vissers, 2011; Fischer & Zwaan, 2008; Meteyard, Zokaei, Bahrami, & Vigliocco, 2008; Rizzolatti & Arbib, 1998), even if it is still under discussion (Caramazza, Anzellotti, Strnad, & Lingnau, 2014).

2.2.4 Memory

Several behavioral investigations support the idea of the embodied nature of memory processes, that is the idea that memory is the recall of a sensation stored after an action production generating that sensation. According to this view, memory is considered a dynamic process instead of a static storage such as a library or a video archive and learning is influenced by sensorimotor coupling. For instance, when normal subjects are asked to retain lists of objects in memory while performing a congruent or incongruent action with these objects, performing an incongruent action impaired memory performance, in comparison to congruent action. Interestingly, the quantity of manual experience with the object regulates the quantity of interference (Yee, Chrysikou, Hoffman, & Thompson-Schill, 2013). Accompanying words or sentences in a foreign language with the corresponding gestures conducts to better learning (Macedonia, 2014); gestures activate motor neurons and could activate MN, which are a subset of motor neurons. Neurophysiological studies have indicated a direct involvement of the MN network in the imitational learning of familiar elementary movements (Iacoboni et al., 1999) and in the acquisition of new motor skills (Buccino et al., 2004; Gatti et al., 2013; Mattar & Gribble, 2005). Stefan et al. (2005) demonstrated that a specific motor memory like the one induced by practicing movements is formed by observation. Viewed together, these studies suggest that MNs could affect memory by improving encoding. In contrast, the MN system seems to influence not only encoding, but also retrieving; autobiographical memories in body-congruent and body-incongruent positions relative to that of the original experience also influences performance (Dijkstra, Kaschak, & Zwaan, 2007). Gelbard-Sagiv, Mukamel, Harel, Malach, and Fried (2008) showed, in patients with pharmacologically intractable epilepsy implanted with depth electrodes, that neurons in the medial temporal lobe are reactivated during spontaneous recall of previously observed audiovisual sequences (episodic memory). According to the authors, these neurons could match the vision of actions executed by other people with the memory of those same actions completed by the observer. This concept is in agreement with the idea that during action accomplishment, a memory of the performed action is formed, and this memory trace is reactivated during action observation. The same group (Mukamel et al., 2010), by recording extracellular activity, described neurons with mirror properties in the hippocampus (that is neurons which responded to both observation and execution of actions). This important observation suggests that in humans, multiple neuronal systems might be endowed with mechanisms of mirroring others’ actions; the functional significance of the mirror mechanism might vary according to the position of the MN, thus supporting different functions, memory included. Even if these data need to be confirmed by further studies, it opens exciting perspectives for a novel interpretation of memory mechanisms.

In conclusion, MNs represent the neuronal population which link perception and action, cognition and motility. For a summary of the literature about MN functions in humans, see Supporting Information Appendix A.

3.2.1 Neurophysiological studies

Many studies on MNs in subjects with PD have used the “Action Observation” (AO) approach. This approach includes both observation and execution of an action (Ertelt & Binkofski, 2012) and it is based on the core property of MN. The subthalamic nucleus is directly linked with the frontal (motor and premotor) cortex through the so called “direct” pathway of the motor circuit (Nambu, 2004) and by the cortico-strio-pallido-subthalamic projection (the “indirect” circuit). The direct pathway conveys a monosynaptic excitatory input from the frontal areas to the subthalamic nucleus. Cortical activity triggered in the MN network by AO could therefore easily propagate to the basal ganglia and particularly to the subthalamic nucleus through these pathways. This entry might modulate basal ganglia output in order to facilitate/inhibit competing motor programs according with the expected results of action (Alegre et al., 2010).

According to these views, Alegre et al. (2010) have shown that the subthalamic nucleus displays variations in activity during both movement observation and movement execution. The authors recorded EEG and local field potentials in 18 patients with PD through surgically implanted electrodes for deep brain stimulation (DBS). Oscillations in electrical activities were recorded during AO and two control conditions. During AO (observing wrist extension movements of an examiner seated in front), a significant bilateral decrease of beta range activity in the subthalamic nucleus was recorded in both on and off medication states (even if higher in the first one). This result suggests a substantial conservation of the MN network in PD, with a possible influence of the medication state. In line with this evidence, another study regarding DBS-implanted patients (Marceglia et al., 2009) found a change in subthalamic nucleus oscillatory activity during AO.

Other studies however showed opposite results; in a study, µ-rhythm desynchronization usually recorded during movement observation was reduced in patients with early PD (Heida, Poppe, Vos, Putten, & Vugt, 2014). Tremblay, Léonard, and Tremblay (2007) tested 11 PD patients in the on condition and the same number of healthy elderly controls, by recording motor evoked potentials (MEP) amplitudes in four conditions: rest, AO (a video depicting a hand cutting a piece of paper with scissors), MI, and active action imitation. The MEP amplitude (recorded at the first dorsal interosseous and abductor digiti minimi) increased in PD patients during active imitation but neither during AO nor during action imagery. This would indicate a failure to involve the motor system more at the covert than at the overt phase of action execution. This would be due, according to the authors, to a deficit in motor activation affecting critical nodes in the motor cortical circuit physiologically involved in action preparation. Tremblay et al. hypothesize that these nodes could be the supplementary motor area and the inferior parietal cortex. However, these remarks may also be explained by the involvement of the frontal-subcortical circuits that link the basal ganglia to the premotor cortex (Alegre, Guridi, & Artieda, 2011; Alegre et al., 2010; Alexander, Crutcher, & DeLong, 1990).

Contrasting results in PD studies could depend on several factors, namely, most importantly, the stage of the disease (not always precised in the different studies), then the recruited population (e.g., a bias depending on the recruitment of very small samples cannot be discarded), the dopaminergic state and maybe the clinical form of the disease (tremor vs. akinetic form), an information which is seldom reported.

3.2.2 fMRI studies

In an interesting neuroimaging study, Anders et al. (2012) recruited eight pre-symptomatic carriers of a single mutant Parkin gene, who exhibited a slight but significant decrease of dopamine metabolism in the basal ganglia. Indeed, it is well known that PD has a long pre-symptomatic stage, during which the brain compensates for dopaminergic degeneration by augmenting motor-related cortical activity (Morrish, Sawle, & Brooks, 1996). This work aimed to study whether comparable compensatory mechanisms were operative in non-motor basal ganglia-cortical gating circuits. As execution and perception of facial gestures are thought to be related to MNs in the ventrolateral premotor cortex (Hennenlotter et al., 2005; Leslie et al., 2004), Parkin mutation carriers first underwent fMRI while observing neutral and affective dynamic facial expressions and then performed a facial emotion recognition task. As expected, recruited participants showed a significantly higher activity in the right ventrolateral premotor cortex during execution and perception of affective facial gestures than healthy controls. Furthermore, Parkin mutation carriers showed a slightly reduced ability to recognize facial emotions that were inversely proportional to the increase of ventrolateral premotor activity. According to the authors, these findings are consistent with the hypothesis that compensatory activity in a MN area during processing of affective facial gestures can lessen impairment in facial emotion appreciation in subclinical Parkin mutation carriers. A failure of this compensatory mechanism could conduct to the impairment of facial expressivity and facial emotion recognition observed in clinically evident PD. In contrast, it is possible that the stimulation of MNs could favor these compensation mechanisms at least in the first phases of the disease, thus allowing PD patients to have better social interaction.

Additional support to the putative link between MN area impairment and the well-known emotional deficit of PD patients (Péron et al., 2012) comes from a similar recent study by the same group (Pohl et al., 2017). In this investigation, 13 PD patients and controls performed an emotion recognition task during fMRI measurement. Subjects watched video clips displaying emotional, non-emotional, and neutral facial expressions or were requested to make these facial expressions. Patients completed the emotion recognition task a little worse than controls, but only for the most difficult expressions to be interpreted. Inferior frontal and anterior inferior parietal MN areas activated during observation and execution of the emotional expressions in both groups, but at a lesser extent in PD patients; furthermore, activation of the right anterior IPL positively correlated with patients’ emotion recognition ability.

Péran et al. (2009) compared fMRI cortical activation during the production of action verbs with activation during object naming, in 14 PD patients using a common set of objects pictures. Data revealed the participation of a large cortical circuit during action verb generation, with differences from the object name production situated above all in the premotor and prefrontal cortices. These results suggest an important role of the frontal cortex in action verb generation. They also suggest that a motor striatal-frontal loops impairment leads to the compensative enrollment of a cortical network. Abrevaya et al. (2016) found similar results when subjects listened to action verbs and nouns. The verb lexical category elicited connectivity between primary motor areas and anterior areas (implied in action observation and imitation), in normal controls, while activated posterior areas in PD patients, thus suggesting that patients might afford alternative pathways to process words when motor substrates are altered.

3.2.3 Behavioral studies

Poliakoff, Galpin, Dick, Moore, and Tipper (2007) evaluated the effect of movement-relevant visual stimuli on reaction times in mild-to-moderate PD patients. In the first experiment, participants had to classify visual stimuli (graspable door handles and the control bar condition) according to shape and orientation. No difference was detected in the overall reaction times for patients and controls. However, the spatial compatibility effect (that is faster reaction times when the response hand and the stimulus direction were compatible) was larger in the handle than in the bar condition for controls, but not for PD patients. In the second experiment, the two groups observed video clips of finger or object movements and had to answer as quickly as possible if an X appeared at the end. Both patients and normal participants reduced significantly their reaction times after observing finger compared to objects movements, indicating partial preservation of MN network in PD. However, patients with PD showed a spatial compatibility effect only for non-living object movements leading the authors to propose that in PD the MN system is not completely preserved and that external cues would act through low-level visual processes.

However, other findings suggest a normal coupling between perception and action systems in PD. In an interference task, healthy controls and PD patients, (tested in off period) completed horizontal and vertical arm movements, while watching a person, or a moving dot, performing similar movements in the same-congruent or orthogonal-incongruent plane (Albert, Peiris, Cohen, Miall, & Praamstra, 2010, see also Alegre et al., 2011). No difference was found between patients and controls.

Castiello, Ansuini, Bulgheroni, Scaravilli, and Nicoletti (2009), examining PD patients’ motor imitation ability, showed kinematic facilitation effects only when the model was a patient with PD—who performed slower and less fluid movements—in contrast to a healthy model. Therefore, differently than normal controls, patients could re-enact their motor representations only when the visual model belonged to the patient's motor repertoire. In line with this, the authors have proposed that basal ganglia play a role in setting frontal and parietal cortices; this role is important not only for the execution of actions but also for the internal simulation of observed behaviors, given that the internal state of the simulated action is available.

Reaserchers have also studied the ability of PD patients to perceive and recognize human action. Precisely, when displayed with limited visual attributes (a point-light walker) persons with PD showed reduced visual sensitivity to biological motion (Jaywant, Shiffrar, Roy, & Cronin-Golomb, 2016). Likewise, when observing point-light human figures that carried communicative and non-communicative gestures, patients were impaired, relatively to normal controls, in describing the meaning of non-communicative gestures, while they normally perceived communicative gestures. However, men were more compromised than women and their ability to recognize both types of gestures was reduced.

3.2.4 Neuropsychological studies

Action naming is an ability that depends on the inferior frontal gyrus, a “classical” MN area (see Kemmerer, Rudrauf, Manzel, & Tranel, 2012). Action verb impairment (with relative preservation of noun processing) has been repetitively described in PD. In the same vein, it has been reported that motor-language coupling (i.e., the influence between verbal processes and voluntary body movements, see García & Ibáñez, 2014) is altered in PD. For a detailed list of publications on these topics one can refer to the reviews by Cardona et al. (2013), Silva, Machado, Cravo, Parente, and Carthery-Goulart (2014), and Birba et al. (2017).

Neuropsychological studies (Jacobs, Shuren, Bowers, & Heilman, 1995; Livingstone, Vezer, McGarry, Lang, & Russo, 2016; Marneweck, Palermo, & Hammond, 2014; Ricciardi et al., 2017), similar to the already cited study of Pohl et al. (2017), have also reported the existence of a significant relationship between voluntary control of facial muscles and emotion recognition deficits in PD patients, an interesting finding at the light of the embodied cognition framework. If the MN system is implicated in both production and perception of facial emotional expression, one can really expect an alteration of MN in PD, even if the MN network is not the only neuronal system involved in emotion perception (Wang, Larson, Bowen, & van Belle, 2006). A possible involvement of the MN network in PD empathy deficits is also suggested by a study of Nobis et al. (2017). This research shows a relationship between the performance on the reading the mind in the eyes test (RME, Baron-Cohen, Wheelwright, Hill, Raste, & Plumb, 2001), a test measuring empathic abilities, and the severity of motor symptoms, along with disease duration, in PD patients. It is also interesting to mention that similar data to those reported in PD have been obtained in Huntington disease (Clark, Neargarder, & Cronin-Golomb, 2008; Sprengelmeyer et al., 1996; Trinkler et al., 2017).

Overall, fMRI data (Anders et al., 2012; Pohl et al., 2017) and neuropsychological studies listed above considerably support the contention that empathy deficits in PD are linked to an impairment of the MN network.

3.2.5 Action observation training for PD rehabilitation

Several works found that action observation training (AOT), which includes observation and execution of an action (Ertelt & Binkofski, 2012) and is evidently based on the stimulation of the MN system (Shih et al., 2017), has a positive effect on PD rehabilitation. During a typical rehabilitation session, patients must observe a specific daily action presented through a video clip on the computer screen and then perform what they have observed. Typically, 20 daily actions are performed and chosen based on their ecological value (e.g., drinking coffee, reading the newspaper) during a rehabilitation treatment that lasts 4 weeks (5 days a week; see Buccino, 2014). This supplementary top-down therapeutic tool was first developed for patients with stroke, with the aim to improve brain plasticity and functional outcomes (Carvalho et al., 2013) but subsequently it has also been used in motor deficits of children with cerebral palsy (Bassolino, Sandini, & Pozzo, 2015; Buccino et al., 2012; Sgandurra et al., 2011) and in PD (Buccino, 2014). Most of the studies in PD have focused on walking disturbances.

3.2.6 Motor imagery

Motor imagery is a cognitive process in which subjects imagine performing a movement without actually doing it (Jeannerod, 2001; Jeannerod & Frak, 1999). MI would mirror the result of conscious access to the intention to move (Jeannerod & Decety, 1995). Since the last decade of the 20th century, many studies have shown that this process involves many areas also recruited during action execution (e.g., Abbruzzese, Trompetto, & Schieppati, 1996; Decety, 1996). Among these areas, there are regions thought to be part of the MN network in humans, such as the premotor cortex and the IPL. Differently than AOT, which is an implicit and automatic online process, MI can only be performed offline that is when one is disconnected from other potential interactions with other subjects or objects. It is also clear that there are obvious behavioral differences among AO, MI, and action execution; they are linked, for example to motivation, attentional processes, the recollection of a sensitive initial state to be able to recall kinesthetic sensation for MI—which is not present in AO—,and the absence of a risk linked to a real execution in AO and MI. Recently, a review has recapitulated data from neuroimaging experiments examining MI, AO, and related control tasks implicating movement execution (Hardwick, Caspers, Eickhoff, & Swinnen, 2018). These behavioral and neural differences should be taken into account when planning a rehabilitative training in different neurodegenerative diseases.

Several studies investigated MI in PD with different approaches (electrophysiological, neuroimaging, and neuropsychological ones)—results are once again controversial. A TMS research (with an increase of cortical excitability; Tremblay et al., 2007) and a neurophysiological study (Cunnington et al., 2001) recording movement-related potentials (usually associated with voluntary movements preparation and execution) reported an impaired facilitation by MI in PD. Conson et al. (2014) tested whether the most affected body side influenced PD patients' ability to mentally manipulate images. PD patients were specifically compromised in judging laterality of the hand corresponding to their own affected side when presented with back-facing human figures, in comparison with normal subjects. However, no difference was found for front-facing figures. Authors hypothesized that two kinds of whole-body transformation could exist, namely an “embodied” one, for back-facing figures, and a “perspective” one, activated for front-facing bodies. However, their interpretation is questionable and their results could be also interpreted in another way—a defective “embodied” mechanism could succeed in judging front-facing figures, but not back-facing ones, because this kind of judgment is less frequently requested (and therefore exerted) in everyday life. It must be also kept in mind that aging affects MI abilities, particularly for the non-dominant side of the body (Saimpont, Mourey, Manckoundia, Pfitzenmeyer, & Pozzo, 2010; Saimpont, Pozzo, & Papaxanthis, 2009); this could exert a confounding effect in studies focusing on MI, PD, and laterality.

Recent investigations testing MI through ad hoc assessment batteries did not found differences between PD patients and normal controls (Heremans et al., 2011; Maillet et al., 2015) both during the on and off phases (Peterson, Pickett, & Earhart, 2012, see Abbruzzese, Avanzino, Marchese, & Pelosin, 2015 for a complete review). A position emission tomography investigation showed that MI in PD was associated with a normal activation in the supplementary motor area and a significant activation of the ipsilateral inferior parietal cortex (both in the “off” and the “on” state) and the ipsilateral premotor cortex (when “off” only). Inferior parietal cortex hyperactivation could compensate the reduced activation of other areas in comparison with controls, including the right dorsolateral prefrontal area, which deals with the working memory component of the imagery task (Cunnington et al., 2001). Other studies (Helmich, Lange, Bloem, & Toni, 2007; Maillet et al., 2015) hypothesized that neural compensation mechanisms can help patients to maintain a good performance in MI tasks.

3.3.1 MRI studies

Most of the research into MN and ALS (a form of motoneuron disease [MotND] in which both central and spinal motor neurons degenerate) or FTD (a common presenile dementia which can display a spectrum of clinical syndromes, ranging from behavioral impairment to language or motor deficit, tightly linked to ALS both from the clinical and genetic point of view) were conducted using magnetic resonance imaging (MRI). For a summary of the literature searching about MN and FTD/ALS, see Supporting Information Appendix C.

Li et al. (2015) conducted a study on 30 patients with ALS and 30 matched healthy controls. The participants performed fMRI while observing a video of repetitive flexion-extension of the fingers at three frequencies or difficulties. AO activated brain areas belonging to the MN system in both ALS and healthy subjects. In ALS patients, however, the dorsal lateral premotor cortex, inferior parietal gyrus, and supplementary motor area were more activated in comparison with controls. Augmented activation within the primary motor cortex, the dorsal lateral premotor cortex, the inferior frontal gyrus, and superior parietal gyrus was linked with hand movement frequency/complexity in patients. This finding indicates a constant compensatory process happening within the motor-processing network of ALS patients, in order to compensate the loss of function. However, a limitation of this study is the fact that it was restricted to action observation.

Another study suggesting a compensatory response of the MN system during AO processing in ALS was performed by Jelsone-Swain, Persad, Burkard, and Welsh (2015). Using fMRI, while subjects observed an actor's hand rhythmically squeezing a ball or squeezed themselves a ball, the authors recorded greater activity in ALS patients than in controls in MN regions, particularly the right frontal inferior operculum and the right frontal and left parietal lobes. In the second part of this same study, Jelsone-Swain et al. (2015) investigated ALS patients’ ability to recognize actions of other people. Participants watched a short video of an actor pantomiming an action with his hands; they had either to passively observe it or to “actively recognize” the action by choosing the correct action from two phrases displayed on the screen. The contrast analysis of the cerebral activity during active recognition versus passive observation displayed greater activity in various anterior and posterior regions in the normal group; on the contrary, only in the ALS group the right occipital activity increased. Patients were then separated into two groups according to their performance at the recognition task—only the best performers showed activation in bilateral frontal superior gyrus. Their performance was also proportional to performance at the RME test. Thus, social cognition would be affected in some ALS patients, an impairment which may be related to a MN system dysfunction (Jelsone-Swain et al., 2015).

Jastorff et al. (2016) investigated brain areas which process emotional stimuli in patients diagnosed with bvFTD. They used both behavioral testing and structural and resting state imaging. bvFTD patients performed worse than controls in emotion detection and in emotion categorization tasks. Their performance in emotion categorization inversely correlated with atrophy of the left inferior frontal gyrus (IFG), a MN area. Functional connectivity analysis also found lessened connectivity of this area. An explanatory hypothesis could be that, due to the atrophy of the IFG, the absence of motor mirroring could prevent the recognition of emotions. Similarly, Brioschi-Guevara et al. (2015) discovered that the atrophy of prefrontal and premotor regions at structural MRI inversely correlated with the ability of bvFTD patients to infer intention and emotional beliefs of others. Marshall et al. (2016) found that grey matter correlates between impaired emotion recognition and automatic imitation depended on the different clinical form of FTD. The brain regions involved delineated a distributed neural network previously associated with embodied cognition. We can conclude that data from bvFTD patients highly support the role of the MN system in emotion recognition.

Finally, another study in ALS deserves to be reported. Verstraete, Veldink, van den Berg, and van den Heuvel (2013) investigated the longitudinal effects of the disease on brain networks using diffusion tensor imaging. Their analyses revealed an expanding sub-network of impaired brain connections over time, with a central role of the primary motor cortex and loss of structural connectivity mainly propagating to frontal and parietal brain areas (regions also belonging to the MN network). In fact, some authors suggest a possible role of the MN network alteration in the pathogenesis of ALS and FTD. According to Eisen, Turner, and Lemon (2013), the ALS/FTD complex might be associated with a disruption of MN. In particular, the damage would involve MN projecting to the pyramidal tract, along with their sophisticated braking mechanism, which allows a subject to observe an actor and use his own motor repertoire to recognize and categorize the observed movement, without activating his own movements (Kraskov, Dancause, Quallo, Shepherd, & Lemon, 2009). In ALS/FTD the MN damage would lead to an impairment of different evolutionarily interlinked functions, associated with the different clinical forms of the disease, in particular: (a) hand function specialization and the associated development of bipedalism (associated with the “classical” form beginning from the upper limbs or to the pseudo-polineuritic form, beginning from the lower limbs), (b) sound production, swallowing, and breathing (which the authors designate as “the brainstem functional complex”; this impairment would be associated with the “bulbar” ALS variant), and (c) cognitive functions linked to social behavior and communication ability, through gestures and language (associated with FTD). Previously, Bak and Chandran (2012) had already proposed that dementia associated with MotND could be interpreted as the fifth major clinical form of the disease, together with bulbar, thoracic and upper and lower limb presentation. Accordingly, the authors underline that in ALS the most impaired cognitive functions are those with the tightest functional relations to the motor system, such as verb and action processing (see also Bak, 2013).

3.3.2 Neuropsychological studies

A series of neuropsychological researches showed an impairment of action rather than object naming in ALS (Bak & Hodges, 1997; Bak, O’Donovan, Xuereb, Boniface, & Hodges, 2001; Grossman et al., 2008; York et al., 2014) and FTD (Cappa et al., 1998; Hillis, Oh, & Ken, 2004). Combining neuropsychological assessment with morphological MRI, York et al. (2014) showed that performance in action verb judgment was related to gray matter atrophy in bilateral frontal regions, including motor association cortex and pre-frontal regions. Similarly, Grossman et al. (2008) had shown that cortical atrophy in premotor areas associated with the representation of the face, arm, and leg correlated with performance on measures requiring action knowledge. This supports the hypothesis that the difficulty shown by ALS patients in naming actions is partly related to the degradation of action-related conceptual knowledge represented in the motor-associated cortex.

Fiori et al. (2013) investigated controls and ALS patients’ responses during a task testing the effects of biomechanical constraints on MI. Effects present in normal subjects (slower responses and lower accuracy when participants judged the laterality of a hand displayed in a position difficult to reach with a real movement awkward positions) were compromised in ALS, at least for the proximal muscles.

Vannuscorps, Dricot, and Pillon (2016), longitudinally examined a patient with basal cortical degeneration, a clinical picture related to FTD, and found contra-indicative data compared to previously cited studies. Throughout 4 years, the patient exhibited worsening action production disorders with associated growing bilateral atrophy in cortical and subcortical regions linked to sensorimotor control (tsuperior parietal cortex, primary motor and premotor cortices, inferior frontal gyrus, mostly on the right side, and basal ganglia). Differently, the patient's performance in processing action-related concepts (e.g., action naming and action comprehension) was spared during the same period. These data would defy the idea that action concept processing is based on the same cognitive and neural networks underlying the sensorimotor control of actions (Meteyard, Cuadrado, Bahrami, & Vigliocco, 2012). However, the moderate involvement of MN areas on the left side could have allowed the maintenance of behavioral performances.

3.4.1 EEG spectral analysis

Recently, 74 adult subjects with MCI undertook EEG recording and high-resolution MRI by Moretti (2016). Alpha3/alpha2 frequency power ratio and cortical thickness were computed for each subject. This ratio represents the increase in high alpha frequencies relatively to low alpha power; it has been demonstrated as a reliable EEG marker of hippocampal atrophy and its increase has been found in MCI subjects who will convert in AD (Moretti et al., 2011). High EEG alpha3/alpha2 frequency power ratio was correlated with atrophy of cortical regions pertaining to the posterior MN network (IPL) areas in MCI subjects. Thus, a possible pathological uncoupling of the MN system would justify the cognitive deficits in prodromal AD. The location of the IPL at the junction of the parietal, temporal, and occipital lobes makes it ideally situated to perform cross-modal integration (hearing/vision/proprioceptive); its impairment would induce both praxis, language, calculation, and topographical deficits characterizing AD, a hypothesis which agrees with traditional neurological views (Greene, Killiany, & Alzheimer's Disease Neuroimaging Initiative, 2010).

3.4.2 TMS studies

In a study by Cotelli et al. (2006), TMS of both left and right dorsolateral prefrontal areas—with a frequency of 20 Hz leading to an increase of cortical excitability—ameliorated action naming but not object naming in 15 AD patients. Similar results were obtained for a mild AD group (Cotelli, Manenti, Cappa, Zanetti, & Miniussi, 2008), while an improved naming accuracy for both action and objects was found in the moderate to severe group. These results raise the interesting possibility of improving language performance in AD via magnetic stimulations of the motor system, but are difficult to interpret. The role of the dorsolateral prefrontal cortex in action naming must be further elucidated. A possibility might be the fact that this cerebral area can intervene by selecting and combining motor representations in the MN system (Vogt et al., 2007); therefore, stimulation of dorsolateral prefrontal cortex might facilitate MN activation and consequently naming of action (if we accept the embodied nature of language). In contrast, according to a recent study (Lanzilotto, Gerbella, Perciavalle, & Lucchetti, 2017) in the dorsolateral prefrontal cortex there are also neurons with mirror-like properties, which react to both self and other people head rotation.

3.4.3 fMRI studies

Rattanachayoto, Tritanon, Laothamatas, and Sungkarat (2012) examined the MN system abnormalities in MCI and AD through a fMRI study. Ninety-two subjects (five MCI, seven mild AD, and 80 cognitively normal) were studied. Participants had to observe a video showing the movement of a hand (tearing a piece of paper) or a control condition (observing a fixation point). Observing the hand movement elicited significant activations of the inferior bilateral frontal lobule and the IPL, but brain activations of controls were significantly higher than those in the MCI and mild AD groups (with no significant difference between these last two groups). Similar to Moretti's study (2016), this study suggests that a dysfunction of the MN system might play a role in cognitive impairment due to MCI/AD, even if we must be cautious in this conclusion because the number of MCI and AD subjects examined was low.

Recently, another study investigated the MN network functioning in people with MCI and AD, alongside with its relationship to cognitive performance (Farina et al., 2017). In this study, three matched groups of 16 subjects each (normal elderly, amnesic MCI with hippocampal atrophy, and AD) were assessed with a fMRI task specifically created to test MN and a focused neuropsychological battery. The MN network appeared largely conserved in aging, while it seemed involved according to an anterior–posterior gradient in neurodegenerative cognitive deterioration. In AD the MN network appeared clearly deficient. The authors speculate that the conservation of the anterior part of the MN network, found in MCI, could possibly compensate the initial decay of the posterior part, preserving cognitive performance, particularly in task (such as RME) related to abilities thought to be linked to MN (see Figure 2 for a graphical representation of the results of study). It is interesting to note that an fMRI study focusing on mindreading (or theory of mind; Baglio et al., 2012) similarly found a preserved performance at RME in people with MCI, despite a reduced activation of some posterior regions of the mind reading network. In the study of Farina et al. (2017) an increased activation of the left Broca area (B44), thought to be a node of the MN system, was observed. This result suggests that the preservation and maybe hyperactivation of MN prefrontal areas could play a compensatory role.

Another study, conducted by Peelle, Powers, Cook, Smith, and Grossman (2014), tested a semantic judgment task, involving shape and color of natural kinds or manufactured items. AD patients were impaired in this task, exhibiting significantly reduced activation in the left temporal-occipital cortex than healthy elderly people and diminished activity in the left inferior frontal cortex. According to the authors, these results agree with a sensory-motor approach of semantic memory; knowledge associated with object concepts would be stored in brain regions that are responsible for perceptual and motor functions of a certain material (Barsalou, Simmons, Barbey, & Wilson, 2003).

Finally, an fMRI study by Lee, Sun, Leung, Chu, and Keysers (2013) aimed at evaluating neural activities during the processing of facial expressions in AD patients. The patients showed weaker activations in left cerebral regions associated with MN or empathic simulation (ventral premotor cortex, anterior insula, and frontal operculum) compared to matched controls. Notably, levels of brain activation in those regions forecasted the level of affect in the AD group. Thus, neural changes in areas associated with motor and emotional simulation (comprised MN regions) could be relevant in the progress of AD symptoms and explain affective impairment associated with this disease.

3.4.4 Behavioral studies

In a first study Bisio et al. (2012) tested if the implicit imitation is preserved in mild to moderate AD patients. AD patients and healthy elderly controls had to observe a dot moving on a screen and to point to its final position when it stopped. The dot speed similarly influenced the actions of AD patients and of healthy elderly participants, suggesting that perception-action matching is not precluded by the disease. In contrast, only patients had an anticipatory motor response, that is they began to move before the end of the stimulus motion, in contrast to the request by the examiner. The imitation of the stimulus speed suggests a conserved ability to match the internal motor representations with that of the visual model; however, the uncontrolled motion initiation would indicate AD patients’ inability to voluntarily inhibit response production. The authors hypothesize that the first part of their results would point to the relative preservation of the MN network (responsible of perception-action matching). However, MN parietal areas such as IPL would be more affected than frontal ones, leading to impaired movement inhibition.

In the second experiment (Bisio et al., 2016), the authors tested voluntary imitation in AD and how this ability was influenced by the nature of the observed stimulus. To this end they compared the capability to reproduce the kinematic features of a human model with that of a dot moving on a screen. The dot's motion (at three different velocities) reproduced the characteristics of a human vertical movement. Participants had to observe the movement of the dot until it reached its final position. When the dot stopped, the subjects were asked to replicate its movement. The human model was a young person previously trained to make vertical simple arm movements at three different velocities. Results showed that, when asked to imitate the velocity of the stimulus, AD patients presented an intact capacity to reproduce it. This capacity improved when the stimulus was a human agent. These data suggest that high-level cognitive processes linked to voluntary imitation could be conserved in mild-moderate stages AD and that voluntary imitation capacities could profit from the implicit interpersonal communication existing between the patient and the human model. These findings make available new clinical views and innovative methods in training programs for people with early dementia. The conservation of the motor resonance mechanisms, not reliant on conscious awareness, finds an intact basis upon which clinicians could model both physical and cognitive interventions for healthy elderly, MCI, or early AD patients. Furthermore, by stimulating the voluntary imitation of everyday activities, the caregivers might help the patients to maintain intact, as long as possible, the ability to easily move in their familiar environment and to feel that they are still part of their family community.

3.4.5 Neuropsychological studies

Some studies pointed out a more severe loss in action than in object naming in AD patients (Druks et al., 2006; Kim & Thompson, 2004; Robinson, Grossman, White-Devine, & D'Esposito, 1996), even if at a minor extent than in FTD (Cappa et al., 1998). In contrast, other researchers reported that nouns were more impaired than verbs in AD patients (Williamson, Adair, Raymer, & Heilman, 1998) or did not report any difference between nouns and verbs (Masterson et al., 2007; Rodríguez-Ferreiro, Davies, González-Nosti, Barbón, & Cuetos, 2009).

Other studies have shown deficits among individuals with AD in recognizing facial emotions (Bediou et al., 2009; Guaita et al., 2009; McLellan, Johnston, Dalrymple-Alford, & Porter, 2008), mainly in the recognition of happy, sad, and fearful expressions (Kohler et al., 2005). These data, alongside with the fMRI study by Lee et al. (2013), appear coherent with a putative impairment of the MN network in AD, since this network (and particularly the inferior frontal gyrus and the ventral premotor cortex) is involved in recognizing other people emotions. However, we must recognize that data about the relationship between emotion recognition and the MN system in AD are scarce in comparison with PD and FTD.

De Scalzi, Rusted, and Oakhill (2015) found a conservation of the action compatibility effect in people with AD; when patients are requested to make judgments on sentences that designate a transfer of an object toward or away from their body, they are faster to answer when the response needs a movement in the same direction as the transfer defined in the sentence. This raises the possibility to exploit the motor systems to improve language comprehension of AD (e.g., by requiring performing the action while listening to a sentence).

3.4.6 Rehabilitative studies

AOT, which is a mean to stimulate the MN network, has been proposed as a rehabilitative motor tool (see Bassolino et al., 2015 for a review). However, AOT could also play a role in maintaining cognitive functions, if one accepts the indissociable nature of cognition from the motor system. So far, the only published research about AOT with AD is from Eggermont, Swaab, Hol, and Scherder (2009). Nursing home residents with dementia observed either videos showing hand movements (19 subjects) or a documentary (25 subjects) for 30 min, 5 days a week, for 6 weeks. At the end of treatment, patients showed improvements in attention and facial recognition (which depends on the superior temporal sulcus, a region tightly linked to the MN system). A preliminary research in progress, performed on small groups (nine each) of patients diagnosed with mild-to-moderate AD, compared AOT with multidimensional stimulation (Spada, 2012). The data analysis showed a significant improvement between test and retest in a naming action task. However, recently an intervention focused on MN in AD patients showed negative results (Caffarra et al., 2016, oral communication at XI Sindem National Congress, Florence, Italy). If the MN network is already disturbed in the pre-dementia phase, as suggested by Moretti’s (2016) and Farina et al.'s (2017) studies, AOT would result ineffective as a rehabilitation technique in moderate AD, and it should be rather performed in the MCI or at least in the early phase to achieve positive results.