This is a beautiful study! Investigators in brain science always ask the mechanism-related questions. So what does the term ‘mechanism’ mean? This study well exemplified it: understanding or interpreting a phenomenon from multiple (circuit, cellular and molecular) levels. Multiple previous studies had reported that inactivation of dopamine neurons and activation of D2-SPNs could induce aversive/avoidance behaviors. D2-SPNs mediated the DA inactivation induced aversive responses, while the underling logic was rarely uncovered. First, by training mice to learn an auditory go-no go task, the authors observed a significant dopamine dip (by monitoring the GCamp6 signals of the axonal terminals of DA neurons in nucleus accumbens with fiber photometry) after CS- trials. Notice the dip occurred earlier than the real learning in behavioral performance, implying that there should be a long-term plasticity supporting afterwards behavioral learning. Then with optogenetics, they persuasively demonstrated that enhancement/removal of the DA dip could improve/impair discrimination learning (an increase/decrease of correct rejection for CS- trials). Furthermore, by combining elaborate glutamate un-caging, current patch-clamp recording and optogenetics, they found that A2A and D2 receptors in D2-SPNs oppositely modulated the structural long-term plasticity of their spines. Tonic dopamine activity and appropriate glutamate inputs (maybe from cortices) could activate A2A receptors which promoted the spine enlargement. Tonic activity would activate D2-SPNs through D2 receptors and suppress spine enlargement at physiological condition.  On the contrary, dopamine dip (a suppression of tonic activity) could decrease the DA concentration and induce the spine enlargement of D2-SPNs in a D2-receptor dependent manner.  Bursting activity of DA neurons could increase the DA concentration and induce the spine enlargement of D1-SPNs.  Importantly, disruption of the spine enlargement of D2-SPNs could effectively impair the learning of correct rejection for the CS- cues.  Depending on the carefully designed experiments, a clear outline of the working mechanisms emerged: negative outcomes induced the dopamine dip promptly and transiently, then the dip induced a long-term structural plasticity (spine enlargement) of D2-SPNs in a D2-receptor dependent manner, then the structural plasticity causally contributed to the afterwards behavioral learnings. To grab some clinical significance, the authors also found that methamphetamine (MAP), which causes psychosis in humans, could impair discrimination learning through the disruption of D2-SPNs structural plasticity. 

Taken together, the study beautifully uncovered the circuit-, cellular- and molecular- brain basis underling aversive learning and perfectly exemplified the meaning of ‘mechanisms’ in brain science.