The neural basis of abrupt transitions from goal-directed to habitual behavior

PROJECT SUMMARY Over time, our goal-directed behavior become automatic; without deliberation, when driving for example, we will habitually push a gas pedal at a green light and the brake pedal at a red light. More specifically, an initial goal- directed response (R) after a cue (S) yields a desired outcome (O) which then evolves into a habit where the cue elicits the action (S-R) without necessarily having the goal in mind. The automatization of decisions is an efficient way to offload well-learned contingencies to free up resources for more flexible, goal-directed learning, and is thought to rely on a gradual shift in control from the dorsomedial to dorsolateral striatum (DMS and DLS). Yet, an overreliance on habit may contribute to maladaptive behaviors in substance use disorders (SUD) indicating a need to understand precisely when and how habits form. It has been challenging, however, to identify when transitions occur since established methods require discrete test sessions which cannot be performed in ‘real-time’. We recently developed a novel behavioral paradigm, ‘volitional engagement’, that reasons that animals in naturalistic environments seek nourishment intermittently and based on underlying needs (i.e. goal- directed behavior). To mimic this, we trained ‘undermotivated’ animals to respond to one auditory cue (S+) and withhold responding to a second cue (S-). Mice rapidly discriminated between the two cues but exhibited fluctuating response rates to the S+ cue for days to weeks afterwards. Surprisingly, and well after expert discrimination performance, mice abruptly transitioned to high and constant response reliability (i.e. habit). This provides one of the first ‘real-time’ measures of transitions between goal-directed to habitual control providing a unique opportunity to assess the neural dynamics and circuits of habit formation en passant. Our overarching hypothesis is that rather than sequential recruitment of DMS and then DLS, both controllers are engaged in parallel throughout learning and that pre-motor cortical signals serve as a higher-level ‘arbiter’, determining whether the DMS or DLS controls behavior at a given moment. To test this, we will (1) validate our volitional engagement approach using optogenetics, sensory-specific outcome devaluation, and contingency degradation (Aim 1), (2) use neural circuit tools to monitor and manipulate pre-motor cortical and dopaminergic inputs and medium-spiny neuron output in the DLS/DMS across learning (Aim 2), and (3) extend our volitional engagement behavioral approach to a habit-promoting task in rats to test for generalizability (Aim 3). Future work can leverage our approach to study how drugs hasten habit formation in real-time. The proposed studies using natural reward will thus pave the way to better understand and potentially treat SUD.