'Direct Pathway' Connections to Globus Pallidus in a Computational Model of the Basal Ganglia
R.Wood, K.Gurney and P.Redgrave
This poster was presented at the BNA2001 Conference in Harrogate, April 8th-11th 2001.
REFERENCE: Wood, R., Gurney, K., and Redgrave, P. (2001) 'Direct pathway' connections to globus pallidus in a computational model of the basal ganglia. British Neuroscience Association Abstracts, 16, p90.
Action selection and the basal ganglia
Recently Redgrave Prescott and Gurney (1999) have suggested that the basal ganglia (BG) plays a crucial role in solving the action selection problem in vertebrates. Based on this premise Gurney, Prescott and Redgrave (2001a,b) have proposed a quantitative model of intrinsic BG processing. Presently little is know of the functional architecture of the basal ganglia. The current prevailing model (Albin et al 1989), based on a direct and an indirect pathway, has been shown to have several short comings. First it fails to account for several anatomically important pathways within the BG. Second, it has yet to be developed as a full computational model so that its claims remain to be substantiated. Third it fails to accommodate recent clinical and anatomical data. Given this situation many people believe that the functional architecture of the BG is ready for review (Albin et al, 1995). The new model proposed by Gurney et al (2001a,b) reinterprets BG anatomy as a set of neural mechanisms for selection. It explains quantitatively the selection operations of the BG and describes the role of dopamine (DA) within the context of action selection.
Imposing the computational hypothesis of action selection on the BG anatomy results in a new, dual-pathway functional architecture. A selection pathway proper, which operates through disinhibition of the output nuclei (GPi/SNr), and a control pathway which modulates the selection process in the first pathway via innervation from GPe. In addition, dopaminergic innervation has been incorporated into the model and been shown to affect the 'promiscuity' of selection.
New anatomical data
Gerfen (1992) and more recently Wu et al (2000), have shown that a subgroup of striatal medium spiny neurons project to the entopeduncular nucleus (EP), the globus pallidus (GP) and the substantia nigra pars reticulata (SNr) . As such this suggests a relatively weak pathway between the striatal medium spiny neuron, with predominantly D1 dopamine receptors, and the GPe. The current study aims to incorporate this anatomical data into the selection-control BG model.
This new data represents a challenge for our model and so we sought to discover its implications for the new functional architecture.
- To show that the selection-control pathway model of the functional architecture of the BG can accommodate the new anatomical data
described above. That is:
- The 'selection' and 'control' pathways retain their functional integrity.
- The major features of the selection behaviour of the model remain intact.
- To uncover the functional role of the striato-pallidal pathway in the context of action selection.
The Prevailing Direct-Indirect Pathway Model 
- Direct pathway from the striatum to the GPi and SNr.
- Indirect pathway from the striatum to the Gpi and SNr Via the STN.
- Only some of the anatomical pathways are emphasised.
- It is non-quantitative, what is the role of each pathway?
- The model emphasises decreased activity in the GPe as the main cause of hypokinetic disorders.
The New Selection-Control Pathway Model 
- Within the model actions are defined as the activity of cell populations, or channels, in the BG.
- The overall activity in a channel defines its salience which is extracted by the medium spiny neurons of the striatum.
- High salience channels effect a release from inhibition in the BG target structures, through inhibition of the BG output nuclei.
- The selection process is implemented through a feedforward off-center on-surround architecture. Diffuse surround excitation is provided by the STN while the source of focused inhibition is the striatum.
- Wu et al (2000) has divided striatal medium spiny neurons into three subtypes, based on anatomical data.
- Type I medium spiny neurons that project solely to the GPe.
- Type II medium spiny neurons project to both the GPe and SNr.
- Type III medium spiny neurons project to the GPe, GPi and SNr.
- Gerfen (1992) found cells projecting exclusively to the GPe produce enkephalin while those projecting to the directly to the output nuclei of the BG express substance P.
Modelling the New Anatomical Data 
- At the relatively high level of modeling employed, the functional architecture on the left is equivalent to that on the right.
- In the resulting model (henceforth the augmented model) the new anatomical data is treated as a relatively weak pathway between the striatal medium spiny neurons with D1 type dopamine receptors, and the GPe (above right).
- The division between 'selection' and 'control' is maintained; the GPe is still a focus of control activity (albeit with an extended afferent set).
- In addition, in order to see if any new features are attributable merely to an increased innervation of GPe, a 'control' model was constructed which consists of the original model, with a stronger striato-pallidal pathway.
Results a: Basic Selection Properties 
- Channel 1 and 2 are stimulated to produce moderate saliencies, channel 3 receives no input.
- In the original model, selection is promoted if only one channel is active or if its salience is significantly greater than that of the others.
- In the new model, increasing the level of simulated dopamine allows this property to be preserved.
- In all models a complete lack of dopamine results in a failure to select any channel.
- High levels of dopamine produce progressively more indiscriminate selection of the channels ('both channels selected').
Two features of the augmented model which are apparent from the bubble plots:
- At intermediate levels of dopamine, if both channels are competing strongly with similar high values of salience, the augmented model was unable to make a selection. Behaviourally this corresponds to the inability to do either of two equally urgent actions.
- For a given level of dopamine, the augmented model shows less propensity for selection. Thus, at low levels of dopamine, selection of any channel is less likely to occur; at high levels, multiple channel selection is less favored. To help quantify this we define a selection metric.
- The graphs highlight the fact the inclusion of the striato-pallidal pathway increases single channel selection. That is, over the range of dopamine for which there is selection, there is greater propensity for single channel selection.
- The range of dopamine for which there is selection is reduced.
- The 'selection' and 'control' functional architecture is maintained with the addition of the striato-pallidal pathway.
- The major features of the selection behaviour of the model remain intact.
- The additional pathway reduces the sensitivity of the selection mechanism at any given dopamine level and, in particular, induces a greater propensity for the selection of single actions, at the expense of multiple action selection.
Gerfen (1992); The Neostriatal mosaic: multiple levels of compartmental organisation in the Basal Ganglia, Annual Rev Neuroscience, 15, 123-124.
Gurney, Prescott and Redgrave (2001a); A computational model of action selection in the basal ganglia. I. A new functional anatomy, Biological Cybernetics, (in press).
Gurney, Prescott and Redgrave (2001b); A computational model of action selection in the basal ganglia. II. Analysis and simulation of behaviour, Biological Cybernetics, (in press).
Redgrave, Prescott, and Gurney (1999); The basal ganglia: a vertebrate solution to the selection problem?, Neuroscience, 89, 1009-1023.
Albin, Young and Penney (1989), The functional anatomy of basal ganglia disorders, Trends in Neuroscience, 12, 366-375.
Albin, Young and Penney (1995) The functional anatomy of disorders of the basal ganglia, Trends in Neuroscience, 18, 63-64.
Wu, Richard and Parent, (2000); Neuroscience Research, 83, 46-62.