Introduction
Persistent stuttering is a neurodevelopmental speech fluency disorder characterized by involuntary speech blocks, sound and syllable repetitions, and sound prolongations. The onset of stuttering occurs most often between the ages of three and six, affects more than 5% of children, and manifests in 0.72% of the adult population, predominantly in males (Yairi and Ambrose
1999; Craig
2002; Howell et al.
2008; Yairi and Ambrose
2013). Stuttering phenotypes are diverse, life-span history varies across subjects, and degree of severity spans the whole spectrum from very mild to very severe. Participation in communication can be largely restricted. Resulting emotional and socio-economic consequences can seriously compromise quality of life.
Aetiology and pathogenesis of persistent developmental stuttering are still obscure (Bloodstein and Ratner
2008). Over the last few decades, a huge body of literature has provided cumulating evidence for irregular neurophysiological signs of the trait of stuttering. Several neurobiological characterizations have been described that are tightly related to each other and that shape the neurophysiological understanding of stuttering. Consistent reports of an imbalanced cortical lateralization during speech tasks manifest the idea of an aberrant hemispheric specialization (Orton and Travis
1929; Travis
1978; Foundas et al.
2001). Two recent quantitative reviews on neuroimaging studies robustly confirmed the imbalanced activation patterns associated with speech production in persistent developmental stuttering (Budde et al.
2014; Belyk et al.
2015). According to these and a previous ALE meta-analyses (Brown et al.
2005), the neural signatures of stuttering are characterized by overactivation of the cerebellum and of right frontal motor regions including the premotor cortex, inferior frontal gyrus, insula, and the operculum and underactivation of the auditory cortex. These often-replicated findings are based on group statistics. In contrast, case-study approaches represent the heterogeneous patterns that emerge due to compensation through one’s lifetime, and different types of treatment. Accordingly, case studies provide additional valuable insights into the complex neural architecture of stuttering (Ingham et al.
2012; Wymbs et al.
2013). Diffusion magnetic resonance imaging (dMRI) repeatedly provided evidence for less coherent white matter structures (Sommer et al.
2002; Chang et al.
2008,
2015; Watkins et al.
2008; Kell et al.
2009; Cykowski et al.
2010; Kronfeld-Duenias et al.
2014,
2016). Affected connections might impede the signal transfer between language-related and speech-related left fronto-parieto-temporal brain regions as summarized in a recent quantitative review (Neef et al.
2015a). Fluent speech production evolves from dynamic network organizations, but functional connectivity within these networks is aberrant in those who stutter (Lu et al.
2009,
2010a,
b; Chang et al.
2011; Chang and Zhu
2013). The spatio-temporal patterning, and particularly the timing of neuronal signals guiding fluent speech production, is not sufficiently tuned (Kent
2000; Salmelin et al.
2000; Ludlow and Loucks
2003; Alm
2004; Etchell et al.
2014). Inhibitory and excitatory intracortical circuits of the ventral primary motor cortex exhibit a reduced dynamic range possibly restricting the proper encoding of ongoing, competing speech motor programs (Neef et al.
2011b,
2015b; Neef
2013). A dysfunction of the basal ganglia circuits or a dysregulation of the dopamine system (Wu et al.
1995,
1997; Braun et al.
1997; Alm
2004; Giraud et al.
2008) might be related to an imprecise cortical input to the striatum and result in an inappropriate excitation of the motor cortex or left inferior frontal/ventral premotor regions. Neurofunctional signs of persistent stuttering are not only restricted to speech movements, but also affect the non-speech motor system (Chang et al.
2009; Neef et al.
2011a; Markett et al.
2016), indicating a broad implication of sensorimotor brain circuits.
The gradient order directions into velocities of articulators (GODIVA) model is a neurocomputational model that provides a mechanistic understanding of speech motor control (Guenther
1995; Guenther et al.
1998; Bohland et al.
2009). The model utilizes a feedforward and a feedback control system to simulate activity across connected brain regions, which results in learning and producing words. An extended version of the GODIVA model showed that both a disconnection of cortico-striatal pathways as well as a dysregulation of the dopamine system resulted in stuttering (Civier et al.
2013). Simulated neural irregularities caused a delayed readout of the motor program via affected basal ganglia thalamo–cortical circuits. This integrated framework of speech production suggested that in the context of stuttering an aberrant timing of neural signalling closely relates to a dysfunction of the basal ganglia, which comprises an erratic excitation and inhibition of engaged neuronal populations.
Numerous clinical studies support the notion that a dysfunction of the basal ganglia is involved in persistent developmental stuttering. Direct evidence for basal ganglia involvement comes from studies with deep brain stimulation in clinical populations. In patients with Parkinson’s disease or primary dystonia, for example, stimulation of the subthalamic nucleus, globus pallidus internus, or ventral intermediate nucleus of the thalamus induces stuttering or modulates preoperative comorbid stuttering: either aggravating or ameliorating it (Burghaus et al.
2005; Nebel et al.
2009; Walker et al.
2009; Allert et al.
2010; Toft and Dietrichs
2011; Risch et al.
2015). The occurrence of basal ganglia disorders, such as Parkinson’s disease, often leads to a re-emergence of recovered developmental stuttering (Shahed and Jankovic
2001). Acquired stuttering after brain injury is associated with lesions in the thalamus or striatum (Lundgren et al.
2010). Furthermore, functional neuroimaging in adults with persistent developmental stuttering repeatedly showed altered activity of the basal ganglia during speech tasks. In adults who stutter, stuttered reading is associated with increased activity in the left globus pallidus and left lateral thalamus as compared to fluent reading (Fox et al.
1996). Speaking under normal or altered auditory feedback is associated with an overactivation of a broad cluster in the midbrain of adults who stutter compared to control participants; strongest activity was found in the substantia nigra encompassing also the pedunculopontine nucleus, subthalamic nucleus, and red nucleus, as well as the left and right posterior lobes of the cerebellum (Watkins et al.
2008). Stuttering therapy caused an increase of activity in the red nucleus (Neumann et al.
2003). During a reading task, activity of the caudate nucleus correlated positively with stuttering severity, while a negative correlation between activity in the substantia nigra and degree of stuttering severity is reported for both pre- and post-treatment (Giraud et al.
2008). Altered functional connectivity between basal ganglia and cortical regions has been observed (Lu et al.
2009,
2010a,
b; Chang and Zhu
2013), and disturbed structural connectivity between cortical and subcortical regions has also been reported (Watkins et al.
2008; Connally et al.
2014; Chang et al.
2015).
The idea of a dysregulation of the dopamine system in persistent developmental stuttering finds support in studies using positron emission tomography (PET). Thereby, the distribution of dopamine receptors can be visualized for the specific engagement of brain regions in certain tasks. Speech production caused an increased uptake of 6-FDOPA in the left caudate tail, left pulvinar, right hypothalamus, medial prefrontal cortex, deep orbital cortex, insular cortex, and auditory cortex, suggesting excessive dopaminergic activity in involved brain regions in adults with persistent developmental stuttering (Wu et al.
1997). Additional support for a hyperdopaminergic state in developmental stuttering comes from the effect of dopamine and dopamine receptor effectors. While levodopa, converted to dopamine, worsens speech fluency (Anderson et al.
1999), dopamine antagonists, such as haloperidol, risperidone, or olanzapine, typically improve speech fluency (Lavid et al.
1999; Maguire et al.
2004). However, the use of pharmacological agents for the treatment of stuttering is currently under debate (Bothe et al.
2008; Boyd et al.
2011) because of provoked adverse side effects (Maguire et al.
2004). Despite the described dopaminergic directionality of the effect in persistent developmental stuttering, in Parkinson’s disease stuttering-like dysfluencies can be related to both increased and decreased dopamine levels (Goberman and Blomgren
2003). Thus, basal ganglia dysfunction in persistent developmental stuttering remains to be established more directly, and the nature of a possible dysregulation in the cortico–striato–cortical loop is yet to be characterized (Giraud et al.
2008).
The circuitry connecting the cortex and the basal ganglia comprises multiple parallel cortico-striatal input and striatonigral output systems (Gerfen
1984). The substantia nigra pars compacta (SNc) is one of the core basal ganglia substrates of dopamine synthesis containing a massive accumulation of dopaminergic neurons densely modulating striatal activity (Dahlstroem and Fuxe
1964; Felten and Sladek
1983). The substantia nigra pars reticularis (SNr) mostly consists of inhibitory GABAergic neurons (Tepper and Lee
2007). In models of cortico-basal ganglia circuits, SNc/SNr constitute a complex nonlinearly operating linchpin, conveying direct, indirect, and hyperdirect inputs (Alexander and Crutcher
1990; Mink
1996; Swanson
2000; Nambu et al.
2002). Generated output enables the selection of appropriate motor action such as speaking. Dopaminergic neurons as well as GABAergic neurons of the SNc/SNr receive (1) direct inhibitory input from cortico-striatal fibres; (2) indirect excitatory input via cortico–striato–pallidal synaptic transmissions through the subthalamic nucleus; (3) hyperdirect excitatory transsynaptic input via cortico–subthalamic nucleus projections, and (4) cortical input from the somatosensory cortex and the motor cortex (Watabe-Uchida et al.
2012). Together with the internal segment of the globus pallidus (GPi), SNr is a main output nucleus of the basal ganglia. Projections target thalamic and brainstem nuclei that further project to a broad range of cortical areas (Deniau et al.
2007).
Given the complexity and massive connectivity of basal ganglia circuits and its nonlinear dynamics on the organization of functional network activity, it is difficult to infer mechanistic principles by means of functional magnetic resonance imaging (fMRI). It is important to consider that multiple factors constrain the interpretation of blood oxygenation level-dependent (BOLD) responses (Düzel et al.
2009). BOLD responses indicate changes in the concentration of deoxyhaemoglobin in the vicinity of red blood cells and vessels, a physiological process caused by increased blood flow, volume, and oxygenation, and which accompanies neural activity (Bandettini et al.
1994). Neurophysiological investigations associate BOLD responses with input and intracortical operations rather than output processing (Logothetis et al.
2001). Neuromodulatory transmitters, such as dopamine regulate processes in neural circuits, but their effect on BOLD is still under investigation (Zaldivar et al.
2014).
To improve the understanding of how neurotransmitter producing substrates contribute to the formation and organization of functional neural networks across the whole human neocortex, it is necessary to draw inferences from noninvasive neuroimaging studies in humans. One feasible way is to study the functional connectivity of dopaminergic nuclei. Resting-state network architecture and task-state architecture are closely matched to each other, indicating that such analyses reflect an intrinsic standard architecture of functional brain organization (Cole et al.
2014). For the substantia nigra, connectivity analyses of fMRI data are rather scarce and no study exists on the functional connectivity of the substantia nigra in stuttering. Only one study has reported psychophysiological interactions (PPI) of the SN/VTA (ventral tegmental area) in healthy adults. This study was related to cognitive control demands in a Stroop task (Köhler et al.
2016). The authors associated the functional connectivity between the SN and dorsal striatum, thalamus, supplementary motor area (SMA), and dorsal anterior cingulate cortex (ACC) with resolving the task-related motor conflict; functional connectivity between VTA and the ventral striatum and perigenual ACC was associated with goal-directed motivational processes. In addition, there is strong resting-state functional connectivity (RSFC) between the SN and cortical and subcortical regions (Tomasi and Volkow
2014; Murty et al.
2014; Zhang et al.
2016; Bianciardi et al.
2016; Bär et al.
2016). Cortical regions involve the dorsomedial frontal, somatomotor, superior temporal, inferior parietal, and occipital cortices (Murty et al.
2014), as well as regions of the default mode network such as the dorsal ACC and the posterior cingulate gyrus/precuneus (Bär et al.
2016). Subcortical connectivity involves the hippocampal complex, nucleus accumbens, putamen, globus pallidus, mediodorsal nucleus of the thalamus, lower brainstem, and several cerebellar nuclei. The current study examined whole-brain functional connectivity dynamics of the substantia nigra to better understand how this central hub coordinates and reorganizes sensorimotor brain networks in persistent developmental stuttering. To achieve this we employed an fMRI paradigm that reliably induces activity in the substantia nigra as shown by a previous study (Lütcke et al.
2008). Accordingly, we used a continuous performance test (CPT, van Leeuwen et al.
1998; Heinrich et al.
2004). During this task the presentation of a particular stimulus leads to the simultaneous anticipation and preparation of a Go/No-Go response within a predictable time interval. Electrophysiological studies suggest that the underlying anticipation process is related to a characteristic slow cortical potential termed contingent negative variation (Walter et al.
1964). It has been suggested that contingent negative variation engages an ensemble of basal ganglia–thalamo–cortical structures (Birbaumer et al.
1990), and electrophysiological studies in stuttering report its irregularities (Prescott and Andrews
1984; Prescott
1988; Walla et al.
2004; Vanhoutte et al.
2015,
2016). Lütcke et al. (
2008) studied healthy participants and elucidated distinct brain networks of the early and the late component of the contingent negative variation (CNV), which typically evolve when carrying out the CPT with long inter-stimulus intervals (Loveless and Sanford
1974; Birbaumer et al.
1990). The early component is assumed to reflect an orientation reaction, while the late component might indicate the motor preparation (Rohrbaugh et al.
1976). The early component was associated with increased BOLD activity in the striatum, SMA, left motor cortex, and right premotor cortex (Lütcke et al.
2008), which might reflect the coordination of input information to the basal ganglia. The late component was associated with increased BOLD activity in the anterior cingulate cortex (ACC), the right fronto-polar cortex, bilateral insula, putamen, thalamus, and the substantia nigra (Lütcke et al.
2008) and thus might reflect the coordination of basal ganglia output. In this study, we aimed to explore whether the coordination of cortico–striato–nigral circuits characterizes the trait of stuttering. We decided to focus on the SN because it is the main source of dopamine synthesis for the motor and non-motor system and because stuttering is associated with a hyperdopaminergic state. Therefore, we planned a functional connectivity analysis, a psychophysiological interaction (PPI), which relies on a correlation analysis of a continuous regression variable subsuming the physiological time varying signal change of the substantia nigra and the time variance of the critical task condition. To gain robust PPI results, we decided not to distinguish between the early and the late component, but considered the whole time span of response anticipation, which included three subsequent brain volumes and thus time varying signals covering 6 s. For the random-effects analysis across PPI maps, we expected an activation map reflecting both cortico-striatal input and nigrostriatal output operations. Ultimately, group contrast maps were calculated to uncover possible deficiencies associated with stuttering. Results allowed us to discuss current data in a framework that accounts for a deficient motor preparation in stuttering, which is possibly related to an insufficient frontoparietal coupling mediated by corticostriatal–striatonigral brain networks.