Open Access

Inhibition of midbrain dopamine (DA) neurons codes for negative reward prediction errors, and causally affects conditioning learning. DA neurons located in the ventral tegmental area (VTA) display two-fold longer rebound delays from hyperpolarizing inhibition in comparison to those in the substantia nigra (SN). This difference has been linked to the slow inactivation of Kv4.3-mediated A-type currents (IA). One known suppressor of Kv4.3 inactivation is a splice variant of potassium channel interacting protein 4 (KChIP4), KChIP4a, which has a unique potassium channel inactivation suppressor domain (KISD) that is coded within exon 3 of the KChIP4 gene. Previous ex vivo experiments from our lab showed that the constitutive knockout of KChIP4 (KChIP4 KO) removes the slow inactivation of IA in VTA DA neurons, with marginal effects on SN DA neurons. KChIP4 KO also increased firing pauses in response to phasic hyperpolarization in these neurons. Here I show, using extracellular recordings combined with juxtacellular labeling in anesthetized mice, that KChIP4 KO also selectively changes the number and duration spontaneous firing pauses by VTA DA neurons in vivo. Pauses were quantified with two different statistical methods, including one developed in house. No other firing parameter was affected, including mean frequency and bursting, and the activity of SN DA neurons was untouched, suggesting that KChIP4 gene products have a highly specific effect on VTA DA neuron responses to inhibitory input. Following up on this result, I developed a new mouse line (KChIP4 Ex3d) where the KISD-coding exon 3 of KChIP4 is selectively excised by cre-recombinase expressed under the dopamine transporter (DAT) promoter, therefore disrupting the expression of KChIP4a only in midbrain DA neurons. I show that these mice have a highly selective behavioral phenotype, displaying a drastic acceleration in extinction learning, but no changes in acquisition learning, in comparison to control littermates. Computational fitting of the behavioral data with a modified Rescorla-Wagner model confirmed that this phenotype is congruent with a selective increase in learning from negative prediction errors. KChIP4 Ex3d also had normal open field exploration, novel object preference, hole board exploration and spontaneous alternation in a plus maze, indicating that exploratory drive, responses to novelty, anxiety, locomotion and working memory were not affected by the genetic manipulation. Furthermore semi-quantitative IHC revealed that KChIP4 Ex3d mice have increased Kv4.3 expression in TH+ neurons, suggesting that the absence of KChIP4a increases the binding of other KChIP variants, which known to increase surface expression of Kv4 channels. Furthermore, in the course of my experimental study I identified that the most used mouse line where cre-recombinase is expressed under the DAT promoter (DAT-cre KI) has a different behavioral phenotype during conditioning in relation to WT littermate controls. These animals displayed increased responding during the initial trials of acquisition and delayed response latency extinction, consistent with an increase in motivation, which is in line with a decrease in DAT function. I propose a working model where the disruption of KChIP4a expression in DA neurons leads to an increase in binding of other KChIP variants to Kv4.3 subunits, promoting their increased surface expression and increasing IA current density; this then increases firing pauses in response to synaptic inhibition, which in behaving animals translates to an increase in negative prediction error-based learning.

Heterozygous mice that express Cre-recombinase under the dopamine transporter promoter (DAT-Cre knock in mice, or KI) are widely used for targeting midbrain dopamine neurons, under the assumption that their constitutive physiology is not affected. We report here that these mice display striking sex-dependent behavioral and molecular differences in relation to wildtypes (WT). Male and female KI mice were constitutively hyperactive, and male KI mice showed attenuated hyperlocomotor responses to amphetamine. In contrast, female KIs displayed a marked reduction in locomotion (“calming” effect) in response to the same dose of amphetamine. Furthermore, male and female DAT-Cre KI mice showed opposing differences in reinforcement learning, with females showing faster conditioning and males showing slower extinction. Other behavioral variables, including working memory and novelty preference, were not changed compared to WT. These effects were paralleled by differences in striatal DAT expression that disproportionately affected female KI mice. Our findings reveal clear limitations of the DAT-Cre line that must be considered when using this model.

Midbrain dopamine (DA) neurons are essential for multiple behaviors. DA neurons that project to different regions also have unique biophysical properties, and it is thought that this diversity reflects specializations to particular computational functions. If this is true, there should be specific genetic determinants of this heterogeneity whose manipulation would lead to circumscribed impacts on behavior. We test this general hypothesis by homing in on one particular mechanism using a new transgenic model and a combination of molecular, electrophysiological, computational, and behavioral approaches. We demonstrate that KChIP4a, a singular Kv4 β-subunit splice variant, determines the long hyperpolarization-rebound delays observed in nucleus accumbens core-projecting DA neurons, that this biophysical switch controls the efficacy of inhibitory inputs to pause firing and, congruently, selectively regulates learning from negative prediction errors. Our results reveal a highly specialized gene-to-behavior mechanistic chain within the DA system, illuminating how cellular diversity shapes information processing in this key neuronal population.

Abstract Inhibition of midbrain dopamine neurons is thought to underlie the signaling of events that are less rewarding than expected and drive learning based on these negative prediction errors. It has recently been shown that Kv4.3 channels influence the integration of inhibitory inputs in specific subpopulations of dopamine neurons. The functional properties of Kv4.3 channels are themselves strongly determined by the binding of auxiliary β-subunits; among them KChIP4a stands-out for its unique combination of modulatory effects. These include decreasing surface membrane trafficking and slowing inactivation kinetics. Therefore, we hypothesized that KChIP4a expression in dopamine neurons could play a crucial role in behavior, in particular by affecting the computation of negative prediction errors. We developed a mouse line where the alternative exon that codes for the KChIP4a splice variant was selectively deleted in midbrain dopamine neurons. In a reward-based reinforcement learning task, we observed that dopamine neuron-specific KChIP4a deletion selectively accelerated the rate of extinction learning, without impacting the acquisition of conditioned responses. We further found that this effect was due to a faster decrease in the initiation rate of goal-directed behaviors, and not faster increases in action disengagement. Furthermore, computational fitting of the behavioral data with a Rescorla-Wagner model confirmed that the observed phenotype was attributable to a selective increase in the learning rate from negative prediction errors. Finally, KChIP4a deletion did not affect performance in other dopamine-sensitive behavioral tasks that did not involve learning from disappointing events, including an absence of effects on working memory, locomotion and novelty preference. Taken together, our results demonstrate that an exon- and midbrain dopamine neuron-specific deletion of an A-type K+ channel β-subunit leads to a selective gain of function in extinction learning. One Sentence Summary Exon- and midbrain dopamine neuron-specific deletion of the Kv4 channel β-subunit KChIP4a selectively accelerates extinction learning