Taste is a mysterious sensation that humans experience. We perceive a wide range of tastes, ranging from the sweetness of strawberry shortcake to the bitterness of black coffee.
It’s well-known that taste buds play a part in the flavor party, but there’s much more to it than what’s on the tip of our tongue. The taste buds send signals to our brain, activating the neurons in our gustatory cortex: the part of our brain associated with perception of taste. Neuroscientists at Stony Brook University are working to increase our understanding of taste and the gustatory cortex.
An efficient way to understand human functions is to observe other animals. In this case, it’s mice. “Temporal progression along discrete coding states during decision-making in the mouse gustatory cortex” is a research paper in which neuroscientists analyze how the gustatory cortex is involved in decisions that depend on the correct associations of specific tastes to actions. This research was conducted by PhD candidate Liam Lang and his advisors, Giancarlo La Camera and Alfredo Fontanini, both professors in the Stony Brook Department of Neurobiology & Behavior.
pictured above is Liam Lang, PhD candidate
In previous experiments, mice were tasked to make taste-based decisions. The mice were given a taste sample randomly selected among four. The four taste samples were sucrose, quinine, maltose, and sucrose octaacetate. After tasting the sample, the mice had a short interval to decide whether to approach a lateral spout on their left or a lateral spout on their right. The correct decision depended on the taste sample and led to a reward.
Interestingly, sucrose and quinine (one sweet and one bitter) were associated with the left choice, while maltose and octaacetate (again, one sweet and one bitter) were associated with the right choice. This way, the animals could not simply associate a sweet taste with one decision and a bitter taste with the other. They had to pay attention to the identity of the taste. Fontanini, the main researcher behind this line of research, has shown in the past that gustatory neurons have an affinity for taste identity but also for taste ‘quality’ (e.g. sweet vs. bitter). This ensured that the researchers could look for neural signatures of taste instructing the correct decision and contrast them with signatures of taste quality.
“Alfredo’s experiments are really the driving force behind these studies,” says La Camera. “With his design, we could look into the connection between taste perception and decision-making in the neural activity of the taste cortex.”
This new paper focuses on particular aspects of the dynamics of neural activity during the decision period.
Using an unsupervised machine-learning method called the Hidden Markov Model (HMM), the data of the gustatory cortex during the decision-making experiment was analyzed. It was found that the activity of the gustatory cortex was ‘metastable,’ meaning that it would suddenly transition from one internal state to the next during decision-making. Some of these internal states encoded task variables such as taste identity, taste quality, and more.
Decoding these states revealed information in the neural activity that indicated:
-
the type of taste the mice were given, for instance, sweet vs. bitter (quality-coding)
-
the categorization of the tastes pertaining to moving left or right (cue-coding)
-
the actions taken by the mice (action-coding)
“The onset time of cue-coding states was very variable from one trial to the next, possibly reflecting the moment in which the animal makes its decision”, says La Camera. “So it’s as if this analysis gave us a way to look inside the brain to determine the moment in which a decision occurs.”
In order to further comprehend what was found, a metastable spiking network model of the gustatory cortex was developed. The model was built to mimic the same behavioral task of the mice and the neural activity that occurs during the task.
“This model is biologically plausible because its neurons and synapses are connected in a certain way which is informed by experimental measurements. We put all of the experimental information we have in the model. Then the rest of the model, which is still free to be manipulated, is built in such a way to reproduce the behavioral results of the experiment,” says La Camera.
The HMM was used again, this time on the model, to analyze if the results were similar to the actual gustatory cortex results of the experiment. After confirming that a successful model was built, it was used to make predictions.
The model predicted that when optogenetic perturbations occurred during the sampling period, the behavioral accuracy of the mice was not affected. However, when optogenetic perturbations were made during the decision-making period, it was predicted that the mice were much more likely to pick the incorrect spout. This result was in keeping with the experimental observations; however, the model allowed to look at a possible cause for this. It was found that behavioral impairment co-occurs with a ‘scrambling’ of the metastable dynamics observed under control.
This research concludes that the neural activity of the gustatory cortex in a mouse is metastable during decision-making. Additionally, it provides a computational model that mimics the neural activity of the gustatory cortex. Further work will aim to confirm the predictions of the model, in particular, whether the disruption of metastable dynamics is causally related to the behavioral impairments.
-Sara Giarnieri, Communications Assistant