SFN: Hikosaka on Motivational Circuitry

Presidential Special Lecture: Motivational Neuronal Circuits for Value, Salience and Information – Okihide Hikosaka

After an introduction from Mickey Goldburg*, Hikosaka takes the stage!

Hikosaka begins his talk with ruminations on the meaning of motivation, stating that motivation is the internal drive to accomplish goals. He presents a conceptual scheme describing two networks: action and motivation that work together to produce goal directed activity. Function of these networks is as follows: action network produces a motor action that produces an outcome, which is then evaluated by the motivation network, which either promotes or inhibits the action network. A key feature of the motivation network is its ability to predict the outcome of the action network. In hard neurophysiological terms, the motivational signal is thought be involve dopamine, but the exact signal, and its information content are not well described.

To examine motivational signals, Hikosaka uses the reward-based saccade task for monkeys, which requires a monkey to make a saccade in return for varying amounts of rewards, depending on the cued direction of the saccade. This biased reward paradigm allows the researchers to evaluate the saccade latency when large versus small rewards are expected. Indeed, saccade latencies are significantly faster when a larger reward is predicted. Using this paradigm, HIkosaka has found reward selective neurons in many areas. One such area is the substania nigra/VTA, which contains a population of dopaminergic neurons that project to multiple areas in the striatum, palidum, and cortical areas. The role of dopaminergic neurons in reward has been reported by multiple groups, but HIkosaka’s group has recorded single dopamine neurons, showing that these neurons are activated by reward, and inhibited by the lack of a reward, predicting the future outcome of the motor activity required by the behavioral task.

But what is driving this activity? Hikosaka notes that evidence for direct functional connectivity onto dopamine neurons has been slim. One exception was research suggesting that the habenula is a major input onto these dopaminergic neurons. The habenula is involved in responses to stress and pain, avoidance learning and error monitoring, all of which have been implicated in the etiology of major depression, schizophrenia, and drug-induced psychosis. These phenomenon are quite distinct from those associated with dopaminergic dysfunction. Recording directly from habenula neurons (in particular lateral habenula), they found neurons that responded to reward prediction errors, but with the opposite sign to responses in dopamine neurons.

What drives the lateral habenula? Many brain areas, but the globus pallidus appears likely to be an input important for encoding reward prediction errors. Recordings from the globus pallidus demonstrated neurons that are inhibited following stimuli predicting reward, and excited by stimuli predicting no reward. Further results suggested that negative reward signals are passed from globus pallidus to lateral habenula. Then, additional expeiments showed that the lateral habenula acts on dopaminergic neurons in the substania nigra/VTA is via rostromeidal tegmental nucleus (see the poster by S. Hong from Hikosaka’s lab later on in this week).

Switching subjects slightly, Hikosaka notes that serotonergic neurons in the dorsal raphe nucleus appear to encode current reward state, where as dopamine neurons encode changes in reward value.

Hikosaka states that motivation is often thought to be driven by reward. But Hikosaka suggests that motivation for research is also a valid type of motivation. To examine this potentially distinct expression of motivation, a new paradigm was created, one that presents the monkey with an information cue or a random cue – the information cue gives the monkey advanced information regarding the size of the reward it will get. They use this paradigm to ask which cue the monkey prefers – to know what reward they will get or not. After a few days of training, the monkeys preferred to know if advance whether they would get a large or small reward. What neural mechanisms underlie this desire/preference for advanced knowledge of reward?

Hikosaka’s group recorded from dopamine neurons, confirming that dopamine neurons encode reward prediction error, but also showing that dopamine neurons are excited by the presentation of the information target, but inhibited by the non-informative target. They also recorded from habenular neurons, and found similar responses. This suggests that the habenula/dopamine circuit contribute to the monkey’s desire for knowledge.

Another type of motivation: Motivation for Salience. A reward may come with risk – what are the neural mechanisms underlying the decision to take a risk for the possibility for reward. To test this, another paradigm was constructed wherein a particular image was shown right before a juice reward, with another image displayed before no reward. Building on this, a new picture was associated with an airpuff to the monkey’s face, and another picture with no airpuff. Similar to the reward condition, the latency to looking at the picture predicting the punishing airpuff was faster than the latency to looking at the picture predicting no punishment. The reward and the punishment have opposite valiances, but are both salient – how does the brain encode both? Research into this question was done by Matsumoto, and was published in 2009 (Bloggers note: this paper – Matsumoto and Hikosaka – is well worth a read), showing that lateral habenula neurons encode motivation value in the negative range. Recording from dopamine neurons in VTA, they found two populations, one of which were encoding value for positive valence, another of which encoded motivational salience, not motivational value. So in summary, they found two populations, one motivational value encoding neurons, and another that encode motivational salience. Matsumoto and Hikosaks worked to localize these two populations. They found that motivational value neurons were localized to ventromedial VTA, which is though to project to ventral striatum. Motivation salience neurons were located primarily in dorsolateral VTA, which may be projecting to dorsal striatum. Again, for a more detailed description of this research, see the published research, Matsumoto and Hikosaka, 2009.

Hikosaka returns to his model of action and motivation network, and brings up the question of how the reward system alters the action network, specifically the connection between the dopamine system and the posterior striatum, which eventually feeds forward onto the superior colliculus, which is involved in saccadic eye movements. Hikosaka highlights a portion of the posterior striatum, the tail of the caudal and posterior putamen- this area is known to receive inputs from visual association areas. They recorded neurons from posterior striatum, finding to their surprise that a majority of neurons responded to visual images with a high degree of object and spatial selectivity. They wondered whether this selectivity was dependent on the experience of the monkeys – could previous reward associations influence the selectivity of posterior striatal neurons? They trained the monkeys on a task that associated a set of specific images with reward, and another set of images with no reward. They then recorded the neuronal activity during presentation of a random selection of these images. After several days of training, they started to see a clear preferential bias in monkey behavior and striatal neuron activity in response to the reward-associated images.

Posterior straitum is known to project to SNr – these neurons also bias, but in the opposite sign (excited by no reward-associated images, not activated by reward-associated images). Neurons in the SNr projected to the superior colliculus, where they are presumably influencing generation of the saccade.

Hikosaka concludes his talk by returning again to his model of action and motivation network, stating that the motivational network can control the action network in multiple ways, via multiple pathways. In summary, Hikosaka states that the heterogeneity of the motivation and action network may “allow the brain to adapt to a complex environment, efficiently and robustly”.

*Winner of the Stanford Neuroblog’s Award for Most Epic Bow Tie

[Updated: additional coverage of the Hikosaka lecture, here.]