Human hippocampal ripples tune cortical responses based on predicted uncertainty.

Taking drugs at the expense of natural rewards (NRs) is a hallmark of drug addiction. This preference bias suggests that artificial rewards (ARs), such as addictive drugs, can hijack the reward system, eventually leading, in some individuals, to compulsive use despite negative consequences. However, studies showing that, in rats, NRs, such as sweets, can be more attractive than cocaine or heroin have challenged this notion. In these studies, only a minority of rats chose drugs at the expense of a sugar solution. The individual preference likely reflects the value assigned to rewards and this value is driven by the context. Indeed, multiple factors influence the assignment of subjective value, including the animal’s current physiological state, past experiences and costs such as effort, risk and delay. The delivery of NRs and ARs initially activates the mesolimbic DA system, but this changes with repeated exposure. When a NR becomes predictable, DA neuron activity and DA levels in the nucleus accumbens (NAc) shift from reward delivery (Rd) to predictive cues. The dynamics of DA transients during cue–reward association learning resemble the reward prediction error (RPE) signaling in reinforcement learning models. In this model, each additional trial generating a DA signal after Rd fuels the DA transient elicited by predictive cues in subsequent trials, increasing the cues’ encoded value and eventually their motivational properties. As addictive drugs invariably elevate DA levels after intake through their distinct pharmacological properties, the DA transients at cue presentation would increase with each iteration. A leading theory builds on this difference between NRs and drug rewards and posits that drug-evoked accumbal DA release represents an aberrant reinforcement learning signal. Such hijacking of the reward system would artificially inflate the value attributed to drug-related cues, subsequently biasing decision-making toward drug use, because the mismatch between expectation and outcome is maintained. Mesolimbic DA release can also be controlled using optogenetic effectors expressed in ventral tegmental area (VTA) DA neurons. Operant optogenetic VTA DA neuron self-stimulation (oDASS) produces robust and reinforcing DA transients in the NAc, eventually leading to compulsive self-stimulation in some individual mice. Thus, oDASS is considered an AR, one that enables precise control over the degree of reinforcement and leverages a fundamental mechanism common to addictive drugs. Like other rewards, oDASS drives the emergence of accumbal DA transients triggered by a predictive cue, including in the NAc core. Although compelling, the hypothesis that DA dynamics predict preference for drug rewards over NRs and determine addiction risk has not been empirically tested.