Action Potential

Light dissection of reward

Image courtesy of Jeremiah Cohen

Out online in Nature today: a paper from Naoshige Uchida and colleagues about cell-type specific reward and punishment signals in the ventral tegmental area (VTA) of mice. The VTA is a midbrain region heavily implicated in reward and addiction, and its outputs are thought to provide reward-related signals to other brain areas. One subpopulation of cells with the VTA, the dopaminergic neurons, have been the topic of intense study for their potential computational role in reward learning. Over a decade ago Wolfram Schultz and colleagues found that in monkeys, dopaminergic neurons fired for unexpected rewards, but were also suppressed if expected rewards were not received. Schultz and colleagues proposed that the neurons were representing the difference between expected and actual outcome, and also noted that such reward prediction error has been theoretically posited to drive reinforcement learning. Although reward prediction is by no means the only proposed role for dopamine, the idea that dopaminergic neurons carry reward signals has figured prominently into theories of VTA function and what goes wrong in disease.

But only around half of VTA neurons are dopaminergic; GABAergic neurons, which make inhibitory projections onto dopaminergic neurons, make up a big chunk of the remainder.  In the current paper, Uchida and colleagues asked how the two populations encode learned rewards and punishments. They recorded from VTA neurons in mice learning to associate odors with rewards and punishments and sorted the neurons post-hoc by their firing properties. Some neurons had brief phasic responses to rewards and reward-predicting cues. Others had sustained increases in firing during the delay between cues and rewards, and yet others sustained decreases. The authors then used optogenetic stimulation to establish dopaminergic or GABAergic identity in a subset of the cells. Dopaminergic neurons all belonged to the first class of cells with phasic reward and reward-predicting responses, and GABAergic neurons the second class with tonic increases. Most but not all dopaminergic neurons were inhibited by aversive stimuli,  most GABAergic neurons were excited.

Why, one might ask, would Nature publish a paper in which many of the conclusions (i.e. the properties of dopaminergic neurons) are largely confirmatory of a slew of studies in both primate and rodent? A number of reasons: for one, the conclusions of previous work have hinged upon reliable neuron identification. Typically this has been done in vivo by measuring waveform characteristics of the action potentials and firing rates (tricky and considered unreliable by some) or, less frequently, post-mortem immunohistochemistry of labeled neurons (even trickier). Here, the authors used a (relatively) new technique to identify neurons in vivo to obtain a more definitive characterization of dopaminergic neurons. But beyond the technical elegance and rigor (which the reviewers all agreed upon), there are novel conclusions: the paper tells us about the reward responses of the GABA neurons, something we know very little about.  The authors point out that theoretical learning models have postulated the existence of exactly such a sustained inhibitory signal to dopaminergic neurons in order to create the reward prediction error signal. Could these responses be part of the mechanism of reward prediction?

So here the technical advance certainly made the paper more interesting, but it was how it was applied and the discovery it enabled that wowed the reviewers and editors.  Recently we’ve been hearing laments/complaints about how optogenetics has become ubiquitous in high profile publication in some areas of neuroscience, sometimes apparently at the expense of actual conceptual advance. For sure, optogenetics has been an exciting technological development and when new technologies first come along, papers describing proofs of principle of what they can do are valuable for driving the field along. For ChR2, that first wave has passed, so at this point we are most certainly looking for new discoveries. Optogenetics doesn’t always add value – depending on where and how it’s applied, it isn’t necessarily more informative than electrical stimulation. But in this case, as a method for quick in vivo identification, it was the right tool for the job. We are curious to see what other questions this kind of identification can and cannot answer in the future. As for the VTA, it remains to be seen whether GABA neurons actually modulate dopaminergic neurons as hypothesized and whether this has any functional significance for the downstream targets of VTA projections.  But for now, we think this is plenty here for one paper.

 Edited 1/23/12: Photo caption revised to attribute photographer (and co-first author of the paper) Jeremiah Cohen, who presumably also donated the cookies.

Comments

  1. Report this comment

    Anne Churchland said:

    This is an exciting paper: one of the frustrations of traditional electrophysiological recording is that it is difficult to interpret response heterogeneity. The idea that the diversity of response properties reflects diversity in the sampled neurons is appealing and, at least in this case, turns out to be sometimes true. As the end of the blog points out, it will be exciting to see how manipulating the GABAergic cells affects reward related behaviors. Here, optogenetic stimulation could represent a major improvement over traditional electrical stimulation which would nonselectively activate both GABAergic and dopaminergic neurons together. I look forward to seeing what this group does next!