A competitive model for striatal action selection

Abstract

The direct and indirect pathway striatal medium spiny neurons (dMSNs and iMSNs) have long been linked to action selection, but the precise roles of these neurons in this process remain unclear. Here, we review different models of striatal pathway function, focusing on the classic "go/no-go" model which posits that dMSNs facilitate movement while iMSNs inhibit movement, and the "complementary" model, which argues that dMSNs facilitate the selection of specific actions while iMSNs inhibit potentially conflicting actions. We discuss the merits and shortcomings of these models and propose a new "competitive" model to explain the contribution of these two pathways to behavior. The "competitive" model argues that rather than inhibiting conflicting actions, iMSNs are tuned to the same actions that dMSNs facilitate, and the two populations "compete" to determine the animal's behavioral response. This model provides a theoretical explanation for how these pathways work together to select actions. In addition, it provides a link between action selection and behavioral reinforcement, via modulating synaptic strength at inputs onto dMSNs and iMSNs. Finally, this model makes predictions about how imbalances in the activity of these pathways may underlie behavioral traits associated with psychiatric disorders. Understanding the roles of these striatal pathways in action selection may help to clarify the neuronal mechanisms of decision-making under normal and pathological conditions.

Choice suppression is achieved through opponent but not independent function of the striatal indirect pathway in mice

“…Overall, our data support existing models of basal ganglia function in which trial and error choice drives learning that is later stored or read out in the balance of activity emerging from DMS dSPNs and iSPNs (70). The fact that learned choice behavior is specifically disrupted by chemogenetic inhibition of dSPNs and activation of iSPNs (but not by inhibition of iSPNs) is consistent with these manipulations counteracting reported patterns of long term potentiation (LTP) onto dSPNs and long-term depression (LTD) onto iSPNs following goal-directed action learning (6).…”

Corticostriatal synaptic weight evolution in a two-alternative forced choice task

“…In our model, spiking activity of striatal neurons in each action channel is driven by ongoing cortical spike trains ( Figure 1), with action selection based on spike patterns at the striatal level resulting from learning-induced weight asymmetries, not from differences in cortical patterns between action channels. Our results predict that the predominant site of corticostriatal plasticity arises at synapses to dMSNs and that emergent striatal activity patterns involve significant spiking in both dMSN and iMSN populations associated with a selected action, consistent with both experimental findings [10,48,49] and with competing pathway models like the Believer-Skeptic hypothesis [3,15,31].…”

“…This co-activation of dMSN and iMSN populations has challenged the traditional model of a strict isomorphism between dMSN activity and excitation and iMSN activity and inhibition. Indeed, more recent theoretical models have proposed that, within an action channel, the dMSN and iMSNs work in a competitive manner to regulate the certainty of a given action decision [3,14,15,31]. For example, Dunovan & Verstynen (2016) proposed a Believer-Skeptic framework for understanding CBGT circuit computations [15].…”

Adiposity covaries with signatures of asymmetric feedback learning during adaptive decisions

“…Here the network represents four competing decisions, one for each deck, as individual action channels that accumulate evidence toward a decision boundary ( Figure 4A ). The certainty of each action is reflected in the drift rate of the decision process ( Dunovan et al., 2015 , 2019 ; Dunovan and Verstynen, 2016 ; Mikhael and Bogacz, 2016 ; Bariselli et al., 2019 ; Dunovan and Verstynen, 2019 ) reflecting the competition between an action-promoting decision process (i.e. Go process, direct pathway), and an action-suppressing process (i.e.…”

“…A key architectural feature of these networks is the dueling influence of the direct (behavior promoting) and indirect (behavior suppressing) striatal pathways ( Alexander et al., 1986 ). Theoretical models have proposed that decisions are encoded as a dynamic competition between these two pathways ( Dunovan et al., 2015 ; Dunovan and Verstynen, 2016 ; Mikhael and Bogacz, 2016 ; Bariselli et al., 2019 ), with the strength of evidence for a given action computed as the likelihood ratio of a hypothesis to ‘execute’ (direct pathway) vs a hypothesis to ‘suppress’ (indirect pathway). During learning, phasic dopamine responses, thought to reflect reward prediction errors ( Schultz et al., 1997 ), have opposing influences on the direct and indirect pathways, depending on the action that they represent and the nature of the feedback signal (i.e.…”