A causal link between prediction errors, dopamine neurons and learning.

2 years ago

Summaries

1/0

This paper seeks to validate the hypothesis that DA cell bursting in response to a positive reward prediction error is causally linked with the learning to associate the predictive cue to the reward, and that the pause in DA cell firing in response to a negative reward prediction error is causally linked to the forgetting of a cue-reward association. The authors spend some time discussing why the methods previously used to probe this question have been insufficient because the time course of drugs and genetic manipulations is too long for testing of the effect of DA cell firing in response to specific cues. They also mention that optogenetics has been used before to stimulate DA neurons which has been found to have a reinforcing effect, but has not been used in a behavioral task specifically to link cue-reward learning. The authors accomplish this by using a blocking procedure in which a second cue is introduced after an initial cue-reward association has been made, which has been shown to prevent the formation of an association between the second redundant cue with the reward. They use TH-Cre rats with a Cre dependent ChR2 virus to achieve ChR2 expression in DA cells in the VTA. In their first experiment, they show that when they stimulate DA cells (or ChR2 expressing cells whatever they may be) during the presentation of the second blocked cue they find that the animals that received this stimulation approach the reward well after just the second cue is presented much more frequently than the control animals who received the stimulations at random times. In their second experiment, they show that when they stimulate DA neurons during the assumed pause that occurs when they omit the reward after cue presentation, the animals are more likely to continue approaching the reward well then the control animals who underwent the same extinction training.

Praise

0/0

> Our results clearly establish that artificially activating VTA dopamine neurons at the time that a natural reward is delivered (or expected) supports cue-elicited responding. From the discussion. This is absolutely true and interesting. I wish the paper's title would have just been this instead of overreaching.

Re: Novelty

1/0

I found this to study to be fairly convincing, and their summary of the current state of the theory relating DA bursts and pauses in response to RPE was simple but useful.

Re: Rigor

Criticism

2/0

In the main figures, they show only a very small inset with co-labeling of TH and ChR, and I don’t know what to make of this figure because there seems to be a large region of green ChR labeling that is not yellow (i.e. not co-labeled with TH). This suggests that many of their infected cells were not actually TH positive, and may have been GABAergic or neurons of surrounding brain regions. I have also recently heard that the Cre expression in TH-Cre rats is not specific to DA neurons.

Re: Rigor

1/0

I think if their goal is to really flesh out the mechanisms by which learning by DA RPE signaling occurs I would find it much more illuminating if they would have recorded from the DA neurons and other parts of the learning circuit while the animals did the task to see how the DA neuron bursting and pausing affected other parts of the circuit. Additionally, while they did the stimulation I’m not sure that the effect of their stimulation was as simple as they assume resulting in the perfect cancelling out of the DA pause in the extinction case and a simple burst in the blocking task, so recording during this time would be useful.

Re: Coolness

0/0

![(e) Visual cue test performance for the first trial and the average of all three trials. The blocking group showed reduced cue responding for the three-trial measure (**P = 0.003), but were not different on the first trial (P = 0.095).](http://i.imgur.com/oHmt4bW.jpg) hashtag AcceptingTheNull :/

Re: Novelty

0/0

> However, it is unknown whether dopamine neuron prediction error signaling and cue-reward learning are causally linked. The more interesting question is whether 'prediction error signalling' is a good description for the role of this spiking pattern in the first place. It's not surprising that behavior is causally linked to DA neuron manipulation. The hole in the literature that the authors refer to in the intro is still a tacit assumption in this study, not something that got experimentally evaluated.

Re: Coolness