Professional musicians undergo long-term training that includes the acquisition of motor skills required to play a musical instrument. Performing music, either playing a musical instrument or singing, requires complex information processing and therefore activates several brain regions simultaneously. There are numerous tasks that can be used to measure brain activity during music-related processing. In a positron emission tomography (PET) study, a melody generation task in amateur musicians has been shown to activate the prefrontal cortex (PFC; right BA 44 and left BA 45), motor areas [premotor cortex (PMC), supplementary motor area (SMA), and pre-SMA], bilateral superior temporal gyrus (STG), insula, basal ganglia, thalamus, and cerebellum (Brown et al., 2006). Blood oxygenation level dependent (BOLD) functional magnetic resonance imaging (fMRI) studies have shown the following results: tasks with the anticipation of sound sequences activate the PFC, premotor areas, and frontostriatal system (Leaver et al., 2009). A similar task that involved the expectancy of a musical ending complex has been shown to activate the superior frontal gyrus (SFG), middle frontal gyrus (MFG), inferior frontal gyrus (IFG), SMA, superior parietal lobule (SPL), STG, and striatum (Seger et al., 2013). In opera singers, singing activates the primary sensorimotor areas and basal ganglia (Kleber et al., 2010). Whereas different cortical areas are activated by different dimensions of musical information processing, such that the dorsal visual stream is activated by melodic processing, and the ventral visual stream is activated by rhythmic processing (Bengtsson and Ullén, 2006), the cortical motor areas (the PMC, SMA, and pre-SMA), basal ganglia, and cerebellum are commonly activated by many music-related tasks.

In general, acquiring skilled movements is accomplished through procedural learning, in which the basal ganglia play a central role (Graybiel, 2005). The basal ganglia are involved in the formation of motor skills, and several studies have demonstrated the activation of these areas during procedural motor learning (Graybiel, 2005). For example, an fMRI study has revealed that the sequence learning of finger movements activates the putamen and prefrontal and motor cortical areas (Lehéricy et al., 2005). Similar brain regions are activated during learning sound sequences (Leaver et al., 2009). The dorsal striatum, which consists of the caudate nucleus and putamen, receives convergent glutamatergic afferents from a large part of the cerebral cortex and functions as the gateway to the basal ganglia (Graybiel, 2005; Kreitzer and Malenka, 2008; Nagano-Saito et al., 2014). The signals in the basal ganglia are subjected to modulation by dopaminergic afferents from the substantia nigra pars compacta and ventral tegmental area. The major target of the nigrostriatal dopaminergic afferents is the dorsal striatum, and this system is particularly involved in motor control and cognition (Alexander et al., 1986). The mesolimbic dopaminergic afferents project to the ventral striatum (Graybiel, 2005). This pathway constitutes a major part of the reward system and is activated in association with reward expectancy (Alexander et al., 1986). Dopamine-mediated signals modulate corticostriatal synapses, thus forming the central part of the actor-critic model for reinforcement learning (Joel et al., 2002; Montague et al., 2004). The signals processed in the striatum are returned to cortical areas through the striato-pallido-thalamo-cortical pathway (Alexander et al., 1986). The striatum is thought to provide instructions to the cortex (Graybiel, 2005), because learning-related changes occur earlier in the striatum than in the PFC of monkeys performing reversals in a conditional association task (Pasupathy and Miller, 2005). Reinforcement learning of how to play a musical instrument requires the integration of multisensory attention, cognitive control, and outcome expectations, which is a striking property of striatal function (Horvitz, 2002). Therefore, the architecture of a network that connects these systems with different domains and converges in the striatum should be critical in developing a musical performance capability.

Current brain network analysis techniques based on graph theory can reveal the network properties of the entire brain (Cole et al., 2010; Van Den Heuvel and Sporns, 2011). A graph is composed of nodes and edges or connections between nodes. Nodes represent brain regions or voxels. There are several graph metrics, including degree and betweenness centrality, that characterize the network properties of nodes and global network properties (Rubinov and Sporns, 2010). Graph theoretical analysis suggests that the putamen is one of the brain network nodes that has significantly high degrees (i.e., the number of connections with other nodes; Van Den Heuvel and Sporns, 2011). Other similar nodes include the SFG, SPL, and precuneus. These nodes are included in the cognitive control network or the default mode network. These networks have often been reported to be negatively correlated with each other and are characterized as having high global connectivity (Cole et al., 2010), which probably reflects the demands for collecting or integrating various types of information to accomplish their tasks. Among the subcortical structures, the amygdala, hippocampus, putamen, and thalamus have been demonstrated to have high global connectivity (Cole et al., 2010). Several nodes have high degrees, constituting a rich-club organization (Van Den Heuvel and Sporns, 2011, 2013). A network with rich-club organization contains hubs that are more densely connected among themselves than with nodes of lower degrees (Van Den Heuvel and Sporns, 2011). A similar network organization has been found by an analysis of structural networks using diffusion tensor imaging data (Van Den Heuvel and Sporns, 2011). The analysis identified 12 strongly interconnected bihemispheric hub regions, including the SFG, SPL, precuneus, hippocampus, putamen, and thalamus. These hub regions were more densely interconnected than was expected solely on the basis of their degrees; therefore, these regions were identified as a rich-club. Such hub regions receive massive inputs from various areas and process a high load of information. An advantage of a network with rich-club organization is the ability to efficiently distribute the load of highly demanding information processing across the hubs. A functional connectivity analysis to determine the voxel-wise distributions of striatum connectivity has revealed a wide distribution across cortical areas, suggesting converging connectivity (Di Martino et al., 2008; Barnes et al., 2010; Choi et al., 2012). This converging corticostriatal connectivity has been confirmed both structurally and functionally (Jarbo and Verstynen, 2015).

A study using voxel-based morphometry (VBM) has reported that, compared with nondancers, professional female ballet dancers have decreased gray matter volumes in the left PMC, SMA, putamen, and SFG as well as decreased white matter volumes in both corticospinal tracts, both internal capsules, the corpus callosum, and the left anterior cingulum (Hänggi et al., 2010). Another VBM study has found that professional pianists have smaller gray matter volumes in the bilateral striatum, compared with amateur pianists or nonmusicians (James et al., 2014). Our VBM study has shown a decreased gray matter volume in the right caudate nucleus of the musicians’ brains compared with that in the brains of nonmusicians (Sato et al., 2015). All the subjects were right-handed females. Musicians have greater volumes in other regions, for example, in the right IFG, left middle occipital gyrus (MOG), bilateral lingual gyrus (Sato et al., 2015), left IFG, right mid orbital gyrus, left intraparietal gyrus, right fusiform gyrus, and left Heschl’s gyrus (HG; James et al., 2014). Therefore, the decreased volumes in the motor cortical areas and striatum of musicians and ballet dancers contrast with the increased volumes in other cortical areas. In a study focusing on pianists, the striatum volume has been found to be highest in nonmusicians, intermediate in amateur pianists, and lowest in professional pianists (James et al., 2014). We have found the same tendency in our VBM study; the caudate volume decreases in the order of nonmusician, music hobbyist, and music expert (Sato et al., 2015). Therefore, the reduction in striatal volume is considered to be a consequence of the neuroplasticity induced by long-term training.

Musical training-related neuroplasticity induces also functional changes in the brain of musicians (Munte et al., 2002; Herholz and Zatorre, 2012; Barrett et al., 2013). For instance, pianists have been shown to exhibit weaker activations, during a bimanual sequential finger movement task, than nonmusicians within a network including the anterior cingulate cortex (ACC), right dorsal PMC, basal ganglia, and cerebellum (Haslinger et al., 2004). This result indicates that professional pianists recruit the motor network to a lesser degree than nonmusicians (Haslinger et al., 2004). In addition, musicians exhibit no significant differences in activation levels between parallel and mirror bimanual finger movements (Haslinger et al., 2004). Parallel finger movements are more challenging for nonmusicians, and therefore, increased activation has been detected in nonmusicians. Musicians may lack significantly increased activation because they would have developed a network for high-level motor skills. Therefore, the functional network for skilled movement control in musicians would differ from that in nonmusicians. There are parallel cortico-striato-pallido-thalamo-cortical (CSPTC) loops composed of the motor, oculomotor, prefrontal, orbitofrontal, and cingulate circuits (Alexander et al., 1986). Motor skills are acquired through the modification of the information flow in the motor CSPTC circuit (Graybiel, 2005). Because the dorsal striatum is a critical node of the CSPTC circuit, receiving convergent input from a large part of the cerebral cortex, and functions as the gateway to the basal ganglia, we hypothesized that long-term musical training alters the functional connectivity of the dorsal striatum. To test this hypothesis, we analyzed the functional connectivity of the striatum (caudate/putamen) in musicians and nonmusicians. All of the participants were female. This study sought to examine how the functional connectivity of the striatum differs between female musicians and nonmusicians.

Figure 1 displays the functional connectivity profiles of the putamen in both groups. For both positive and negative functional connectivities, the profile was wider in nonmusicians than in musicians. Figure 2 shows the surface maps of the functional connectivity of the left putamen in the musicians and nonmusicians. The nonmusicians had wider functional connectivity profiles in the opercular/insular regions, SMA, and cingulate gyrus. In contrast, the musicians seemed to have developed a wider functional connectivity profile in the precentral gyrus; however, the strength of the functional connectivity did not differ significantly between the groups. The differences in the functional connectivities between the two groups are summarized in Table 1. The musicians had significantly decreased functional connectivity between the left putamen and the bilateral frontal operculum (FO), followed by the cerebellum (vermis 7) and the left insular cortex (IC). The functional connectivity between the right putamen and the right FO, left IC, and ACC was decreased in the musicians.

Figure 3 illustrates the functional connectivity profiles of the caudate nucleus in both groups. In the musicians, the functional connectivity profile was less extended laterally but extended more to the frontal regions. Figure 4 displays surface maps of the functional connectivity of the left caudate nucleus in the musicians and nonmusicians. In contrast to the results for the putamen, a positive functional connectivity with the precentral gyrus was not observed in either group. However, the left caudate nucleus had negative functional connectivity with the pre- and postcentral gyri in both groups. The functional connectivity profiles in the FO/anterior insula (AI) regions, SMA, and ACC were more extended in the nonmusicians, whereas the functional connectivity profile in the ventral medial prefrontal cortex (mPFC) was wider in the musicians. There were extended negative functional connectivity profiles in and around the precuneus, posterior cingulate cortex (PCC), and retrosplenial cortex; these connectivities were not observed in the functional connectivity of the putamen. The differences in the functional connectivities between the two groups are summarized in Table 1. The musicians had significantly decreased functional connectivity between the left caudate nucleus and the cerebellar regions (6LR and 8L).

Graph theoretical analysis of the entire brain functional network revealed that musicians had significantly lower global efficiency, cost, and degree and a higher average path length in the bilateral putamen (p < 0.05, FDR). Table 2 lists the degrees of the putamen and caudate nucleus. The bilateral putamen, but not the caudate nucleus, had significantly lower degrees or fewer connections in musicians than in nonmusicians. Figure 5 shows the degrees of the left and right putamen and caudate nuclei. The degrees did not differ between the left and right putamen in musicians (p = 0.791) and nonmusicians (p = 0.672). Nonmusicians had a higher degree in the left caudate nucleus than in the right caudate nucleus (p = 0.047), whereas this tendency was weaker at an insignificant level in musicians (p = 0.190).

Musical performance requires complex information processing, including the integration of relevant information represented widely across cortical areas. The putamen, which has a high degree of connection among nodes of the entire brain, can function as a hub for the integration of such information. However, this corticostriatal network is smaller or selective in female musicians than in nonmusicians, consistently with results from previous anatomical studies reporting reduced volumes of the dorsal striatum in musicians and ballet dancers. The authors consider that long-term musical training constructs such a network by pruning unnecessary connections to optimize performance.

ST and EK planned and conducted all the experiments. ST analyzed the data and wrote the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.