A neurocomputational view of the effects of Parkinson's disease on speech production.

The purpose of this article is to review the scientific literature concerning speech in Parkinson's disease (PD) with reference to the DIVA/GODIVA neurocomputational modeling framework. Within this theoretical view, the basal ganglia (BG) contribute to several different aspects of speech motor learning and execution. First, the BG are posited to play a role in the initiation and scaling of speech movements. Within the DIVA/GODIVA framework, initiation and scaling are carried out by initiation map nodes in the supplementary motor area acting in concert with the BG. Reduced support of the initiation map from the BG in PD would result in reduced movement intensity as well as susceptibility to early termination of movement. A second proposed role concerns the learning of common speech sequences, such as phoneme sequences comprising words; this view receives support from the animal literature as well as studies identifying speech sequence learning deficits in PD. Third, the BG may play a role in the temporary buffering and sequencing of longer speech utterances such as phrases during conversational speech. Although the literature does not support a critical role for the BG in representing sequence order (since incorrectly ordered speech is not characteristic of PD), the BG are posited to contribute to the scaling of individual movements in the sequence, including increasing movement intensity for emphatic stress on key words. Therapeutic interventions for PD have inconsistent effects on speech. In contrast to dopaminergic treatments, which typically either leave speech unchanged or lead to minor improvements, deep brain stimulation (DBS) can degrade speech in some cases and improve it in others. However, cases of degradation may be due to unintended stimulation of efferent motor projections to the speech articulators. Findings of spared speech after bilateral pallidotomy appear to indicate that any role played by the BG in adult speech must be supplementary rather than mandatory, with the sequential order of well-learned sequences apparently represented elsewhere (e.g., in cortico-cortical projections).

Modular horizontal network within mouse primary visual cortex.

Interactions between feedback connections from higher cortical areas and local horizontal connections within primary visual cortex (V1) were shown to play a role in contextual processing in different behavioral states. Layer 1 (L1) is an important part of the underlying network. This cell-sparse layer is a target of feedback and local inputs, and nexus for contacts onto apical dendrites of projection neurons in the layers below. Importantly, L1 is a site for coupling inputs from the outside world with internal information. To determine whether all of these circuit elements overlap in L1, we labeled the horizontal network within mouse V1 with anterograde and retrograde viral tracers. We found two types of local horizontal connections: short ones that were tangentially limited to the representation of the point image, and long ones which reached beyond the receptive field center, deep into its surround. The long connections were patchy and terminated preferentially in M2 muscarinic acetylcholine receptor-negative (M2-) interpatches. Anterogradely labeled inputs overlapped in M2-interpatches with apical dendrites of retrogradely labeled L2/3 and L5 cells, forming module-selective loops between topographically distant locations. Previous work showed that L1 of M2-interpatches receive inputs from the lateral posterior thalamic nucleus (LP) and from a feedback network from areas of the medial dorsal stream, including the secondary motor cortex. Together, these findings suggest that interactions in M2-interpatches play a role in processing visual inputs produced by object-and self-motion.

Neurocomputational modeling of speech motor development.

This review describes a computational approach for modeling the development of speech motor control in infants. We address the development of two levels of control: articulation of individual speech sounds (defined here as phonemes, syllables, or words for which there is an optimized motor program) and production of sound sequences such as phrases or sentences. We describe the DIVA model of speech motor control and its application to the problem of learning individual sounds in the infant's native language. Then we describe the GODIVA model, an extension of DIVA, and how chunking of frequently produced phoneme sequences is implemented within it.

Integration of Feedforward and Feedback Information Streams in the Modular Architecture of Mouse Visual Cortex.

Radial cell columns are a hallmark feature of cortical architecture in many mammalian species. It has long been held, based on the lack of orientation columns, that such functional units are absent in rodent primary visual cortex (V1). These observations led to the view that rodent visual cortex has a fundamentally different network architecture than that of carnivores and primates. While columns may be lacking in rodent V1, we describe in this review that modular clusters of inputs to layer 1 and projection neurons in the layers below are prominent features of the mouse visual cortex. We propose that modules organize thalamocortical inputs, intracortical processing streams, and transthalamic communications that underlie distinct sensory and sensorimotor functions.

Hierarchical and nonhierarchical features of the mouse visual cortical network.

Neocortical computations underlying vision are performed by a distributed network of functionally specialized areas. Mouse visual cortex, a dense interareal network that exhibits hierarchical properties, comprises subnetworks interconnecting distinct processing streams. To determine the layout of the mouse visual hierarchy, we have evaluated the laminar patterns formed by interareal axonal projections originating in each of ten areas. Reciprocally connected pairs of areas exhibit feedforward/feedback relationships consistent with a hierarchical organization. Beta regression analyses, which estimate a continuous hierarchical distance measure, indicate that the network comprises multiple nonhierarchical circuits embedded in a hierarchical organization of overlapping levels. Single-unit recordings in anaesthetized mice show that receptive field sizes are generally consistent with the hierarchy, with the ventral stream exhibiting a stricter hierarchy than the dorsal stream. Together, the results provide an anatomical metric for hierarchical distance, and reveal both hierarchical and nonhierarchical motifs in mouse visual cortex.

Modular Network between Postrhinal Visual Cortex, Amygdala, and Entorhinal Cortex.

The postrhinal area (POR) is a known center for integrating spatial with nonspatial visual information and a possible hub for influencing landmark navigation by affective input from the amygdala. This may involve specific circuits within muscarinic acetylcholine receptor 2 (M2)-positive (M2+) or M2- modules of POR that associate inputs from the thalamus, cortex, and amygdala, and send outputs to the entorhinal cortex. Using anterograde and retrograde labeling with conventional and viral tracers in male and female mice, we found that all higher visual areas of the ventral cortical stream project to the amygdala, while such inputs are absent from primary visual cortex and dorsal stream areas. Unexpectedly for the presumed salt-and-pepper organization of mouse extrastriate cortex, tracing results show that inputs from the dorsal lateral geniculate nucleus and lateral posterior nucleus were spatially clustered in layer 1 (L1) and overlapped with M2+ patches of POR. In contrast, input from the amygdala to L1 of POR terminated in M2- interpatches. Importantly, the amygdalocortical input to M2- interpatches in L1 overlapped preferentially with spatially clustered apical dendrites of POR neurons projecting to amygdala and entorhinal area lateral, medial (ENTm). The results suggest that subnetworks in POR, used to build spatial maps for navigation, do not receive direct thalamocortical M2+ patch-targeting inputs. Instead, they involve local networks of M2- interpatches, which are influenced by affective information from the amygdala and project to ENTm, whose cells respond to visual landmark cues for navigation.SIGNIFICANCE STATEMENT A central purpose of visual object recognition is identifying the salience of objects and approaching or avoiding them. However, it is not currently known how the visual cortex integrates the multiple streams of information, including affective and navigational cues, which are required to accomplish this task. We find that in a higher visual area, the postrhinal cortex, the cortical sheet is divided into interdigitating modules receiving distinct inputs from visual and emotion-related sources. One of these modules is preferentially connected with the amygdala and provides outputs to entorhinal cortex, constituting a processing stream that may assign emotional salience to objects and landmarks for the guidance of goal-directed navigation.

Spatial Clustering of Inhibition in Mouse Primary Visual Cortex.

Whether mouse visual cortex contains orderly feature maps is debated. The overlapping pattern of geniculocortical inputs with M2 muscarinic acetylcholine receptor-rich patches in layer 1 (L1) suggests a non-random architecture. Here, we found that L1 inputs from the lateral posterior thalamus (LP) avoid patches and target interpatches. Channelrhodopsin-2-assisted mapping of excitatory postsynaptic currents (EPSCs) in L2/3 shows that the relative excitation of parvalbumin-expressing interneurons (PVs) and pyramidal neurons (PNs) by dLGN, LP, and cortical feedback is distinct and depends on whether the neurons reside in clusters aligned with patches or interpatches. Paired recordings from PVs and PNs show that unitary inhibitory postsynaptic currents (uIPSCs) are larger in interpatches than in patches. The spatial clustering of inhibition is matched by dense clustering of PV terminals in interpatches. The results show that the excitation/inhibition balance across V1 is organized into patch and interpatch subnetworks, which receive distinct long-range inputs and are specialized for the processing of distinct spatiotemporal features.