Action selection and the Basal Ganglia

What is action selection and how could the brain do it?

For my first full post I thought that rather then pick out some recent paper that’s made a big splash or something prominent in the media I would just talk about a topic in neuroscience I find interesting, it isn’t even related to what I study for my PhD, just something I find interesting and hopefully you will to. I’m going to talk about the action selection: what it is and how it's carried out in the brain. To put it simply as we engage with the world around us any number of potential actions are possible and action selection is choosing the next action. If we elaborate a bit, while any number of actions is available not all choices are made equal. Depending on your current needs and goals, certain courses of action are more or less appropriate. So, action selection refers not only to the initiation of one course of action but also to the decision processes that lead to the initiation of (hopefully) the most appropriate course of action and the inhibition of others.

One way this could be achieved in the brain is to have a dedicated action selection module. To carry out the task of action selection successfully this module would require certain fundamental features. First, it would need access to neural representations of competing action choices i.e. receive input from a wide range of brain regions involved directly in motor control. Second, its output would need to have a direct influence on the expression of the winning movement, and inhibition of losing movements. Finally, to be able to make a decision on the most appropriate course of action the module would also need access to information about goals, motivations, context, internal drives (e.g. hunger, thirst) etc. in a form that would have direct influence over movement choice.

The basal ganglia and its anatomy

A candidate for one of the brains action selection modules is the basal ganglia. Which is? The basal ganglia is a group of interconnected nuclei* nestled deep below the bumpy outer casing of the cerebral cortex, more or less at the core of the brain. The four nuclei to keep in mind when it comes to action selection are the striatum, subthalamic nuclei (STN), globus pallidus (in particular the internal portion or GPi) and substantia nigra (a particular subregion called the substantia nigra pars reticulata or SNr is most important here).

So what is it about the basal ganglia that make it a candidate action selection module? To begin with disorders of the basal ganglia in humans or lesions of the basal ganglia in animals often lead to problems in either the initiation or inhibition of movement. This has been known for decades so it has been understood for a long time that the basal ganglia played some role in movement control. More recently though researchers have started to recognize that the basal ganglia is capable of acting as an action selection module, this is largely based on anatomical studies of the connections to and from the basal ganglia and its complex internal architecture.

How then do these nuclei connect and communicate to act as an action selection module? The best place to start is the largest of the nuclei, the striatum which can be thought of as the gateway to the basal ganglia. Just about all incoming signals to enter the basal ganglia will pass through the striatum and this comprises a wealth of input from across most of the brain. Large swathes of the cerebral cortex project to the striatum, encompassing all of its major functional subdivisions: sensory areas (areas that process information directly from the senses), motor areas and association areas (areas that process some combination of sensory, motor and emotional information, cognitive or higher order functions and memory). Plus the striatum also receives input from brainstem and subcortical (below the cotex) regions involved in memory, emotion, motivation and regions controlling low level motor responses. In other words neural activity in the basal ganglia and particularly the striatum is open to influence by information concerning the internal drives and goals and external environment that might influence your choice of movement.

When we’re talking about action selection, of particular interest are the many cortical and subcortical regions which project to the striatum which are directly involved in planning and executing movement. The input these connections provide to the striatum presumably represent the range of available actions competing for selection. The relative “strength” of these representations could also be modulated by those additional inputs to the striatum I mentioned above, conveying information concerning memory, emotion, senses, goals, motivations and biological drives. This could influence which course of action is most strongly represented in the basal ganglia.

Pathways through the basal ganglia

The projections to the striatum are excitatory i.e. their net effect on the striatal neurons they contact is to increase firing. The majority of striatal neurons (over 95%) are medium spiny neurons that project to both the GPi and SNr (as the function of these two nuclei is so similar I’ll stick with GPi/ SNr for now). The medium spiny neurons are inhibitory, in other words they have the net effect of decreasing firing of their target GPi/ SNr neurons (see my picture below for a (hopefully) easy to understand illustration of these connections!)

The main output cells of the GPi/ SNr are also inhibitory and the GPi/ SNr cells that receive input from motor related regions of the striatum also project to the regions directly involved in movement performance- in a neat little loop!. The striatum GPi/ SNr neurons fire persistently, meaning that under normal circumstances the GPi/ SNr is acting like a constant brake on movement initiation.

So if the representation of a particular movement is strong enough (perhaps this might be coded by a certain rate of firing from neurons projecting from the motor region controlling this movement) it leads to activation of medium spiny neurons in the part of the striatum representing that movement. This activation leads to an inhibition of the corresponding GPi/ SNr neurons, releasing the brake on that movement (usually referred to as "disinhibition") and then the movement can be initiated. This pathway is known as the direct pathway.

There is also a second indirect pathway via the striatum that follows, as the name suggests, a less direct route through the basal ganglia. Medium spiny neurons also project to the external GP (GPe), the GPe then sends inhibitory projections to the STN which sends excitatory projections to the GPi/ SNr. So….. the medium spiny neurons of the striatum inhibit the GPe, this leads to a reduction in its inhibition of the STN so that the excitatory drive from the STN can drive up activity levels in the GPi/ SNr. In other words the indirect pathway has the opposite overall effect to the “direct” pathway, increasing the inhibitory influence the GPi/ SNr have on movement.

The classical view of the basal ganglia focused on these two pathways, but to make things a bit more complex there is a third hyperdirect pathway to consider. This pathway is the fastest pathway through the basal ganglia and bypasses the striatum entirely. The same excitatory motor inputs that project to the striatum also project directly to the STN. Like the indirect pathway the hyperdirect pathway increases the inhibitory output of the GPi/ SNr on to its target motor regions by increasing the excitatory drive it receives from the STN.

The pathways and action selection

How do these pathways combine to carry out the task of action selection? Recordings of the electrical activity** taken from the GPi/SNr in non-human primates (monkeys) after stimulation of cortical motor regions show an early excitation, presumably from the hyperdirect pathway followed by an inhibition from direct pathway and lastly an excitation from the indirect pathway.

As I said before the classical accounts of basal ganglia function, including its role in action selection only dealt with the direct and indirect pathways. The classical model suggested by Mink is known as the "centre-surround" model which I'll come back to later on. A later paper adjusted this model to account for the role of the hyperdirect pathway (Nambu et al, 2002) and this account is the one I am going to go with when I talk about how the pathways combine in action selection. It's worth bearing in mind that this is just one working model and other models might ascribe different roles to one or all of the pathways or leave some out altogether. I'm telling you about this particular model because for me it accounts best for the data we have relating to basal ganglia anatomy, connections and activity (although I'm by no means an expert on this, just an interested outsider so if you have any other ideas that is what the comments section is for!) 

In this model signals from cortical motor control regions project through the hyperdirect pathway - drive a non-specific increase in the inhibitory drive from the GPi/ SNr to motor control regions, this blocks all movements from being initiated. Then specific activity representing a “winning” movement passes through the direct pathway, disinbiting the winning movement, finally a third signal through the indirect pathway strengthens the inhibition on the losing movements by further exciting the GPi/ SNr. So, the hyperdirect and indirect pathway provide a general signal which causes a "surround" inhibition of all unwanted movement whilst the direct pathway provides a targeted signal which drives a "centre" disinhibition of only the chosen movement (the centre-surround model).

This then is one model of how the basal ganglia could carry out the function of action selection, one way in which the brain has evolved to solve a problem crucial to our everyday life. By just adjusting the balance and targeting of excitation and inhibition in the system the basal ganglia can carry out an apparently complex function with a relatively limited repertoire of processes.

This is literally my first blog post ever of any description, and I suspect I am a pretty terrible writer when it comes to making things accessible and engaging, having only ever written dry academic pieces so please….. please…… please….. if you happen to have stumbled across this blog, leave a bit of constructive criticism if you have any. Also, for the sake of clarity and concision I’ve chosen not to go on for too long about certain aspects of the basal ganglia and skip over some of the less important aspects entirely! So if you want anything described in any more detail or if there is anything a bit unclear let me know and I’ll try and give a good answer.

Hyperdirect pathway: Motor regions that project to the striatum also project to the STN via a quicker pathway. The excitatory drive on to the STN increases the excitation on the GPi/ SNr . This drives inhibition on all target motor execution regions. 

Direct pathway: Motor regions also project to the striatum through the more delayed "direct" pathway. Excitatory inputs from motor controlling regions combine with modulatory input from elsewhere in the brain. The strongest (winning) response drives inhibitory output from the sriatum to the GPi/ SNr, pausing inhibitory output and disinhibiting the winning response whilst losing responses remain inhibited.

Indirect pathway: The slowest of the three pathways. By causing inhibition of the GPe the "indirect" pathway reduces inhibition of the STN increasing the excitatory drive on the GPi/ SNr. This could maintain/ increase the inhibition of losing movements whilst the chosen movement is disinhibited.

*Just as an aside to avoid confusion, when talking about nuclei in the brain this doesn’t refer to the DNA containing part of a cells but a tightly packed cluster of neurons that are usually related by some central function or other.

** A way of directly measuring neuron firing levels but more on this in some other post probably.

FURTHER READING

Freely accessible introductory articles:

Basal ganglia Wikipedia entry. Lots of detail on the anatomy and connections of the basal ganglia.

Basal ganglia Scholarpedia entry. Also good discussion of anatomy and connections. Also some information on the activity of neurons in the different nuclei and a short discussion of the basal ganglia and action selection. Perhaps a bit out of date, doesn't make any mention of the hyperdirect pathway.

Action selection Scholarpedia entry. General discussion on the possible ways a system (e.g. the brain) might achieve action selection and a few passages on how the basal ganglia meets the requirements for a dedicated action selection module.

More advanced articles (some may be free to access but others will be behind a paywall/ require institutional access):

Alexander, G.E. et al (1986). Parallel organisation of functionally segregated circuits linking basal ganglia and cortex. A review paper about the way in which input from the certain regions of the cortex including motor regions travels through the basal ganglia sepeately, allowing the range of inputs to have different influences on the output of the basal ganglia.  

Mink, J.W. (1996). The basal ganglia: focused selection and inhibition of competing motor programs. The classical "centre-surround" model of basal ganglia function, the author introduces the idea of how the "direct" and indirect pathway might work together to initiate winning actions and inhibit losing ones.

Nambu, A. et al (2000). Excitatory cortical inputs to pallidal neurons via the subthalamic nucleus in the monkey. Research paper that recorded from the GPi in monkeys to show the influence of the three pathways on basal ganglia output. 

Nambu, A. et al (2002). Functional significance of the cortico-subthalamo-pallidal hyperdirect pathway. A modification of the "centre-surround" model incorporating the hyperdirect pathway.