Putative γ-aminobutyric acid neurons in the ventral tegmental area have a similar pattern of plasticity as dopamine neurons during appetitive and aversive learning

Abstract

Dopamine influences affective, motor and cognitive processing, and multiple forms of learning and memory. This multifaceted functionality, which operates across long temporal windows, is broader than the narrow and temporally constrained role often ascribed to dopamine neurons as reward prediction error detectors. Given the modulatory nature of dopamine neurotransmission, that dopamine release is activated by both aversive and appetitive stimuli, and that dopamine receptors are often localized extrasynaptically, a role for dopamine in transmitting precise error signals has been questioned. Here we recorded from ventral tegmental area (VTA) neurons, while exposing rats to novel stimuli that were predictive of an appetitive or aversive outcome in the same behavioral session. The VTA contains dopamine and -aminobutyric acid (GABA) neurons that project to striatal and cortical regions and are strongly implicated in learning and affective processing. The response of VTA neurons, regardless of whether they had putative dopamine or GABA waveforms, transformed flexibly as animals learned to associate novel stimuli from different sensory modalities to appetitive or aversive outcomes. Learning the appetitive association led to larger excitatory VTA responses, whereas acquiring the aversive association led to a biphasic response of brief excitation followed by sustained inhibition. These responses shifted rapidly as outcome contingencies changed. These data suggest that VTA neurons interface sensory information with representational memory of aversive and appetitive events. This pattern of plasticity was not selective for putative dopamine neurons and generalized to other cells, suggesting that the temporally precise information transfer from the VTA may be mediated by faster acting GABA neurons.

Fig. 1

Behavioral task and performance. (A) Behavioral task. For eight consecutive daily sessions, a double conditioning paradigm was used in which two novel stimuli (tone or flashing light) were followed by the delivery of an appetitive (sugar pellet) or aversive (mild electric foot shock; 180 ms, 0.2 mA) US. Tone or light stimuli were delivered for 10 s and were randomly assigned to be paired with either sugar pellet or foot shock for the first eight sessions. For sessions 9–16 this association was reversed. The CS that preceded sugar pellet delivery is referred to as CSap and the CS predictive of shock is referred to as CSav. Rats (n = 7) underwent 60 trials (30 trials of each contingency) delivered pseudo-randomly. (B) Behavioral performance. As animals learned the associations across multiple sessions, the R_ap (behavioral association index; see Materials and methods) was increased and R_av was depressed. After CS associations were reversed, R_ap and R_av were developed to the new association contingencies. Error bars represent SEM.

Fig. 2

The changing response of VTA neurons to the CSs. (A) Representative examples of phasic responses in one animal to the same CS in session 1 vs. session 8. Raster plots represent neural activity, with each row corresponding to a single trial. All 30 trials are shown in each plot with the corresponding peristimulus time histogram. Blue ticks represent the time that the animal nose pokes into the food trough. (B) The population response of VTA neurons to the CSap and CSav during sessions 1–16. There was a progressive increase in the magnitude of excitatory responses to the CSap across sessions 1–8. The magnitude of the population response was strongly correlated with the session number (r = 0.918, P < 0.01). The initial excitatory response to the CSav changed to a delayed inhibitory response as task learning progressed. The population response of neurons to the CSav decreased as learning progressed and the magnitude of the population response was inversely correlated with the session number (r = −0.960, P < 0.01). Again, the population responses shifted in response to contingency reversal with increasingly excitatory responses to the new CSap (right lower panel) and inhibitory responses to the new CSav (right upper panel) as learning progressed (r = 0.925 and r = −0.973, respectively, P < 0.01 in all cases). Shadows represent SEM. (C) Changes in the proportion of responsive neurons to either CS during sessions 1–16. In session 1–2, a small proportion of neurons displayed excitatory phasic responses to the CSap (16.5%; 14/85) and CSav (17.6%; 15/85). In subsequent sessions, a larger proportion of neurons displayedexcitatory phasic responses to the CSap and inhibitory responses to the CSav (r = 0.809 and r = 0.818, respectively, P < 0.05 in all cases). On session 7–8, 43.8% (46/105) of VTA neurons were activated by the CSap and 12.4% (13/105) were inhibited by the CSav. The percentage of excitatory or inhibitory response to each CS shifted as contingencies changed. After the association contingencies were shifted, more neurons showed excited responses to the new CSap and inhibited responses to the new CSav as learning progressed (r = 0.872 and r = −0.932, respectively, P < 0.01 in all cases). (D) Trial-by-trial responses of an individual neuron during session 9 to a CS that previously predicted foot shock (on sessions 1–8) but now predicts sugar pellet delivery (left) and to a CS that previously predicted sugar pellet delivery and now predicts foot shock (right). The neuron began exhibiting activation and inhibition to the new CSap and CSav, respectively, in the later trials of this session. Color indicates the firing rates per second. (E) The normalized response of VTA neurons recorded in session 9 reflects learning of the new CS contingencies (n = 53 in session 9). Thirty trials of CS type were divided into five bins of six consecutive trials. The responses of each neuron to a stimulus were normalized in these bins. The response to the new CSap was increased (r = 0.988, P < 0.01) and the response to the new CSav was decreased (r = −0.988, P < 0.01) during the course of the session. Significant correlations were not found for either CSap or CSav in session 8 (r = 0.579 and r = 0.146, respectively, P > 0.05 in all cases, n = 54). Circles and triangles represent neural activity and behavior separately. Error bars represent SEM.

Fig. 3

Comparison of the neural responses of VTA subpopulations. (A–C) Representative waveforms for neurons classified as Type I (n = 529 from seven animals during 16 recording sessions, median n = 34 per session), Type II (n = 113, median n = 8 per session) and Type III (n = 128, median n = 8 per session). Scale: 0.2 ms and 0.05 mv. Subtyped population responses to CSap and CSav of VTA neurons during sessions 1–2 (D), 7–8 (E), 9–10(F) and 15–16 (G). There was a significant effect of time with no interaction [group × time] or group effects (P < 0.01 for time, P > 0.05 for group and interaction [group × time] effect, repeated two-way anova). (H and I) Changes in the proportion of subtyped VTA CSap- and CSav-responsive neurons. More Type I neurons were activated to CSap and depressed to CSav on session 7–8 compared with sessions 1–2 (P < 0.05 for both cases, Chi-square test). Because of the limited sample sizes of Type II and Type III neurons, these comparisons were not executed for these groups (although we observed qualitatively similar patterns of responding between these and Type I neurons).

Fig. 4

Time-course of the response to the CSap and CSav. (A) Type I neurons responsive to the CSav exhibited a biphasic pattern, first exhibiting excitation followed by inhibition. Histograms represent sliding 100 ms time bins with each time bin advanced by 20 ms from the beginning of the previous time bin. This analysis demonstrates a biphasic response to the CSav during sessions 1–8 (left, n = 248) and sessions 9–16 (right, n = 281), with an early period of excitation followed by a more sustained period of inhibition. (B) Time-course of the signal change in the neural response to the CSap and CSav. ROC analyses revealed that the signal change in response to the presentation of either CS developed as learning progressed. In sessions 1–2, VTA neurons did not significantly alter their responding to the presentation of either CS (filled circle). In sessions 5–8, VTA neurons developed significantly different responses (activation or inhibition) to the CSap and CSav (filled circle). By sessions 13–16 (after contingency reversal), VTA neurons exhibited strong differential responding to the presentation of the CSap or CSav in a non-monotonic fashion (filled circle). Examination of the signal change from baseline to the presentation of either the CSap (empty circle) or CSav (triangle) indicates that the early peak in differential responding is best explained by the increase in activity to the CSap (empty circle), which then declines rapidly. Differential responding to the CSav was maintained for a longer period of time, best explaining the second peak in VTA neuron differential activity (second peak of filled circles).