Roles for globus pallidus externa revealed in a computational model of action selection in the basal ganglia
Introduction
The basal ganglia are an evolutionarily conserved group of subcortical nuclei, which have long been implicated in action selection Frank, 2005, Frank et al., 2004, Grillner and Robertson, 2016, Hikosaka et al., 2000, Lindahl et al., 2013, Redgrave et al., 1999, Schroll et al., 2012, Stephenson-Jones et al., 2011. Several computational models have been developed, examining their role in action selection Berthet et al., 2016, Frank et al., 2004, Gurney et al., 2001a, Gurney et al., 2001b, Hikosaka et al., 2000, Kamali Sarvestani et al., 2011, Mink, 1996, Schroll et al., 2012. They propose the basal ganglia as a ‘selection machine’ resolving conflicts between competing behaviours for common and restricted motor resources Frank, 2005, Redgrave et al., 1999, Schroll and Hamker, 2013. This notion is backed by studies showing that the stimulation of the striatum, the main input nucleus, can either trigger actions or inhibit them Freeze et al., 2013, Kravitz et al., 2010. Furthermore, loss of dopamine neurons in the substancia nigra pars compacta (SNc), result in a reduced ability to select motor responses (Wylie et al., 2009) in pathological conditions like Parkinson’s disease. In furtherance of the selection hypothesis, the basal ganglia are also implicated in learning of stimulus–response associations (Alexander, DeLong, & Strick, 1986) as well as in establishing stimulus–response–outcome associations (Redgrave & Gurney, 2006).
Existing models have dealt with a variety of aspects of basal ganglia function and anatomical context. Thus, many discuss the role of reinforcement learning Brown et al., 2004, Frank, 2006, Gurney et al., 2015, Redgrave and Gurney, 2006, Schroll et al., 2012 and have also incorporated the thalamo-cortical loops Beiser and Houk, 1998, Chersi et al., 2013, Frank et al., 2004, Humphries and Gurney, 2002, van Albada and Robinson, 2009. These models also cover a range of levels of biological description — from abstract system-level to detailed multi-compartmental neuronal models, as well as simulations of ensembles of neurons. Addressing computations at the level of the subnuclei of the basal ganglia, there have been several models of the striatal microcircuitry Damodaran et al., 2015, Humphries, Lepora et al., 2009, Humphries, Wood, and Gurney, 2009, the subthalamic nuclei (STN, Frank 2006), as well as examinations of the oscillations associated within the STN–GPe network Blenkinsop et al., 2017, Corbit et al., 2016.
Most models are based on the classical architecture of connectivity of the basal ganglia (Fig. 1 (A)), focusing on the direct pathway – the striatal D1 projections to the output nuclei, globus pallidus interna and substantia nigra pars reticulata (GPi/SNr), and the indirect pathway – the striatal D2 projections to the GPe, and the GPe projections directly to GPi/SNr and the STN–GPe/GPi loop. The GPe has been considered as homologous in structure and function in most of these models. However, recent studies have revealed a new subpopulation of GPe neurons, the arkypallidal cells (Mallet et al., 2012) that are active in anti-phase to their more common counterparts, the prototypical GPe neurons (Mallet et al. 2012, see also Methods). These two classes are also referred to as the TA and TI neurons respectively (Mallet et al., 2012). The arkypallidal cells provide a major input to the striatum (Mallet et al., 2012).
We aimed to incorporate the arkypallidal neurons into a well-tested model architecture of the basal ganglia Gurney et al., 2001a, Gurney et al., 2001b. The architecture has been validated at several levels of description: at the systems level using rate coded neural populations constrained by anatomical and physiological data (see Blenkinsop et al., 2017, Gurney et al., 2004, Humphries and Gurney, 2002); spiking neuron models challenged with physiological data Chersi et al., 2013, Humphries et al., 2006, Stewart et al., 2012; and at the behavioural level in embodied (robotic) models (Prescott, Montes González, Gurney, Humphries, & Redgrave, 2006). Most recently, it has been used to link a raft of neurobehavioural phenomena to neuronal mechanisms observed in vitro (Gurney et al., 2015). Thus, this model architecture offers a strong platform to try to understand the role and function of arkypallidal neurons and their afferent and efferent pathways in action selection. Furthermore, we also included another scheme of organisation in the GPe in terms of neuronal subpopulations — the outer and inner GPe neurons (Sadek, Magill, & Bolam, 2007). We built on the original model and used the methodologies developed therein to assess them, on extended architectures of connectivity of the GPe. The arkypallidal neurons have been accommodated in a few computational models Bahuguna et al., 2017, Bogacz et al., 2016, Lindahl and Hellgren Kotaleski, 2016, Moolchand et al., 2017 and their function in supporting optimal action selection (Bogacz et al., 2016) as well as in network dynamics underlying basal ganglia movement disorders have been investigated Bahuguna et al., 2017, Lindahl and Hellgren Kotaleski, 2016. However, their role in action selection and their influence on other basal ganglia subnuclei, needs additional investigation. Further, the outer and inner neuron dichotomy has not been included in any model so far (to our knowledge), and their role in action selection remains unknown. Our work addresses these lacunas and reveals important functions for different neuronal subpopulations within the GPe, and unites these two prevalent schemes of organisation within the GPe (GPe TI/TA and GPe outer/inner, Mallet et al. 2012 and Sadek et al. 2007) and furthermore, places the GPe in perspective as an important control centre of the basal ganglia.
Section snippets
Anatomy of the basal ganglia
The classical anatomy of the basal ganglia Bolam et al., 2000, Calabresi et al., 2014, Redgrave et al., 1999 is shown in Fig. 1 (A). It consists of the following principal nuclei: the striatum, the globus pallidus ((GPe) and internal (GPi) divisions in primates), the STN and the substantia nigra (SNr and SNc). The primary input nuclei are the striatum and the STN. The output nuclei are the GPi and the SNr. The input nuclei receive afferent signals from most of the cerebral cortex and the
Results
Recall from the methods that we make use of step-wise and combined models, investigating single and multiple pathways respectively, and that their deployment is carried out in three modelling phases. This approach is reflected here in reporting the Results.
Discussion
We have investigated the newly discovered intrinsic connectivity of GPe in considerable detail. Quantitative evaluation of selection performance in this model has revealed several new functions of GPe that may be understood within the selection framework. The prototypical neurons have been shown to be the principal subpopulation influencing action selection. The arkypallidal neurons are used by both the prototypical neurons and the STN, to modulate the activity of the striatum. These
Concluding remarks
The simulations have thrown light on the importance of the GPe in the basal ganglia, and its crucial and myriad role in action selection. It seems to be a ‘control centre’ of the basal ganglia with considerable influence on the functioning of other basal ganglia nuclei. The results show the GPe controlling the striatum, the GPi/SNr and as shown also in previous models, the STN (Gurney et al., 2001a). In particular, the prototypical GPe TI (outer/inner) neurons, seem to be the ‘controllers’,
Acknowledgements
We acknowledge the following grant sponsors: European Horizon 2020 Framework Programme (Grant 720270 (Human Brain Project SGA1), Grant 785907 (Human Brain Project SGA2)), the Swedish Research Council and the Swedish e-Science Research Center . We are also grateful for comments on the manuscript by Dr. Brita Robertson and Associate professor Dr. Peter Wallén.
References (93)
- et al.
Responses of substantia nigra pars reticulata neurons to intrastriatal D1 and D2 dopaminergic agonist injections in the rat
Neuroscience Letters
(1996) Reinforcement learning control
Current Opinion in Neurobiology
(1994)- et al.
Deep brain stimulation of the subthalamic nucleus for the treatment of Parkinson’s disease
The Lancet Neurology
(2009) - et al.
Physiological aspects of information processing in the basal ganglia of normal and Parkinsonian primates
Trends in Neurosciences
(1998) - et al.
Move to the rhythm: oscillations in the subthalamic nucleus–external globus pallidus network
Trends in Neurosciences
(2002) - et al.
How laminar frontal cortex and basal ganglia circuits interact to control planned and reactive saccades
Neural Networks
(2004) - et al.
Striatal local circuitry: A new framework for lateral inhibition
Neuron
(2017) - et al.
A spiking neuron model of the cortico-basal ganglia circuits for goal-directed and habitual action learning
Neural Networks
(2013) - et al.
Distinct developmental origins manifest in the specialized encoding of movement by adult neurons of the external globus pallidus
Neuron
(2015) Hold your horses: A dynamic computational role for the subthalamic nucleus in decision making
Neural Networks
(2006)
Regulation of firing of dopaminergic neurons and control of goal-directed behaviors
Trends in Neurosciences
Intrinsic function of a neuronal network — a vertebrate central pattern generator. Published on the World Wide Web on 8 April 1998.1
Brain Research Reviews
Mechanisms for selection of basic motor programs — roles for the striatum and pallidum
Trends in Neurosciences
The basal ganglia over 500 million years
Current Biology
Dopamine-modulated dynamic cell assemblies generated by the {GABAergic} striatal microcircuit
Neural Networks
On the neural substrates for exploratory dynamics in basal ganglia: A model
Neural Networks
Dichotomous organization of the external globus pallidus
Neuron
Arkypallidal cells send a stop signal to striatum
Neuron
The basal ganglia: Focused selection and inhibition of competing motor programs
Progress in Neurobiology
A robot model of the basal ganglia: Behavior and intrinsic processing
Neural Networks
The basal ganglia: a vertebrate solution to the selection problem?
Neuroscience
Basal ganglia oscillations and pathophysiology of movement disorders
Current Opinion in Neurobiology
Working memory and response selection: A computational account of interactions among cortico-basalganglio-thalamic loops
Neural Networks
Abnormal patterns of theta frequency oscillations during the temporal evolution of freezing of gait in Parkinson’s disease
Clinical Neurophysiology
Evolutionary conservation of the basal ganglia as a common vertebrate mechanism for action selection
Current Biology
Mean-field modeling of the basal ganglia-thalamocortical system. I: Firing rates in healthy and Parkinsonian states
Journal of Theoretical Biology
Developmental changes in dopamine neurotransmission in adolescence: Behavioral implications and issues in assessment
Brain and Cognition
Spontaneous firing patterns of identified spiny neurons in the rat neostriatum
Brain Research
The effect of Parkinson’s disease on interference control during action selection
Neuropsychologia
Frontal dopaminergic abnormality in Tourette syndrome: A postmortem analysis
Journal of the Neurological Sciences
Lesions in monkey globus pallidus externus exacerbate Parkinsonian symptoms
Experimental Neurology
Prototypic and arkypallidal neurons in the dopamine-intact external globus pallidus
Journal of Neuroscience
Parallel organization of functionally segregated circuits linking basal ganglia and cortex
Annual Review of Neuroscience
Homologous basal ganglia network models in physiological and Parkinsonian conditions
Frontiers in Computational Neuroscience
Recent advances in hierarchical reinforcement learning
Discrete Event Dynamic Systems
Model of cortical-basal ganglionic processing: Encoding the serial order of sensory events
Journal of Neurophysiology
Functional relevance of different basal ganglia pathways investigated in a spiking model with reward dependent plasticity
Frontiers in Neural Circuits
Selective innervation of neostriatal interneurons by a subclass of neuron in the globus pallidus of the rat
Journal of Neuroscience
Frequency and function in the basal ganglia: the origins of beta and gamma band activity
The Journal of Physiology
Properties of neurons in external globus pallidus can support optimal action selection
PLoS Computational Biology
Synaptic organisation of the basal ganglia
Journal of Anatomy
Direct and indirect pathways of basal ganglia: a critical reappraisal
Nat Neurosci
Basal ganglia system as an engine for exploration
What do the basal ganglia do? A modeling perspective
Biological Cybernetics
Single neostriatal efferent axons in the globus pallidus: a light and electron microscopic study
Science
Pallidostriatal projections promote beta oscillations in a dopamine-depleted biophysical network model
Journal of Neuroscience
Cited by (32)
Distributed dopaminergic signaling in the basal ganglia and its relationship to motor disability in Parkinson's disease
2023, Current Opinion in NeurobiologyPallidal and Cortical Oscillations in Freely Moving Patients With Dystonia
2023, NeuromodulationReinforcement learning in a spiking neural model of striatum plasticity
2023, NeurocomputingTonic dopamine, uncertainty and basal ganglia action selection
2021, NeuroscienceCitation Excerpt :The central idea of our model is that increasing the background level of dopamine raises the stochasticity of action selection (random exploration) as the basal ganglia’s ability to select deterministically is reduced. “Soft” selection was proposed by Suryanarayana et al. (2019) as a potential source of exploration, as higher levels of dopamine reduces basal ganglia’s capacity to filter cortical input leading to a greater number of actions being disinhibited. Animals genetically engineered to be hyper-dopaminergic, exhibit behaviour that is consistent with increased random exploration (Zhuang et al., 2001), however, this behaviour could result from influence from extra-striatal sites such as prefrontal cortex (Beeler et al., 2010).
A Simple-Units Complex-Structure Neural Network model of the Basal Ganglia to simulate reinforcement learning tasks
2021, Procedia Computer ScienceCortical and subcortical functional specificity associated with response inhibition
2020, NeuroImageCitation Excerpt :Specifically, it has been suggested that the GPe may play a role in the execution of response sequences (Chan et al., 2005; see also Nambu, 2008), which is required on signal trials in both the SST and DT. Recent work has highlighted the GPe as crucial to action selection, hypothesising that activity from the STN to SN (components of the hyperdirect pathway) might initiate a ‘pause’ so that selective cancellation of actions can occur via the GPe to STR (Mallet et al., 2016; Suryanarayana et al., 2019). Indeed, STN has been found to be the main excitatory input to the GPe (Hegeman et al., 2016).