A new study in Nature Metabolism sheds light on the activity of metabolic signaling molecules that mediate changes in human behavior depending on the internal state of the body.
Introduction
Human brains use sensory data to regulate physiological needs in such a way that the individual can survive in an environment that exerts selective pressure. An important aspect of this is the production of sensory associations that are learned rather than inborn.
Associative learning is a key skill for survival when faced with selection pressures. As the environment changes, sensory input data also does. This leads to the automatic grouping of sensory and behavioral signals in the brain that helps the organism adapt its behavior to the environment rather than losing its fitness. This is the central facet of associative learning.
On the other hand, the internal state of the body also requires behavioral changes by the organism. For instance, human behavior is adapted to the individual’s energy state. This also occurs via parallel metabolic and behavioral pathways in the body. These must be integrated with the external cues to provide the behavioral drive appropriate to the internal and external situation.
The ventral midbrain is full of dopamine (DA) neurons that help bring about adaptive behavior by regulatory and reinforcing actions. These control behavioral motivation but also reinforce behavioral actions. Reward-based learning is mediated by the mesoaccumbens pathway, wherein dopamine is projected from the ventral tegmental area (VTA) to the nucleus accumbens (NAc)
Dopaminergic neurons in the VTA have an important role in predicting in real-time the outcomes that are probably favorable in any situation. The result is that the individual can choose optimal behavior. These neurons encode reward prediction errors in the VTA.
These errors denote the difference between the expected outcome value and the real value. They mark the need for the organism to rethink the importance of the current sensory inputs.
However, this re-evaluation must also integrate the current state of the individual’s physiology, producing appropriate behavioral choices in keeping with the need of the moment.
The mesoaccumbens pathway is among the probable mediators of this link between metabolic cues and associative learning, in addition to the external cues.
This will lead to appropriate motivated behavior such as feeding when food is available, in response to energy deprivation.
Two such cues could be insulin and glucagon-like peptide 1 (GLP-1) receptors. These affect excitatory signaling at some synapses and may reduce both anticipation and preference formation for food cues. However, there is not much evidence linking these peptides to the regulation of associative learning.
Some research suggests that if metabolic cues are not able to impact dopaminergic neuron function, such as when there is impaired metabolism due to insulin resistance, the effect could be poorly adapted behavior including overfeeding, leading to obesity. The specific way this happens could be through poor associative learning in response to the physiological state.
In insulin resistance, GLP-1 receptor agonists increase the release of insulin in response to glucose, and may normalize motivational behavior, perhaps by restoring normal metabolism, and thus restoring this link between associative learning and sensory cues.
The current study was a crossover randomized controlled study examining the effect of GLP-1 receptor activation on associative learning in both obese and non-obese individuals.
What did the study show?
In the study, humans were given liraglutide, a GLP-1 agonist, and then tested for associative learning in conjunction with fMRI. This was repeated on another day after placebo administration.
The test of associative learning involved measuring how well participants associated a high or low auditory tone with a following picture. The associations varied within the duration of the experiment from being very predictable to unpredictable, meaning that adaptive learning would be required. The directionality and speed of associative learning between the cue and outcome were estimated for each learner.
The three key measures assessed included the sensory prediction error, the adaptive learning rate, and the resulting adaptive prediction error. The first refers to the degree of difference between the participant’s choice and the chances that it is the correct choice.
The second variable accounts for the subjective lack of assurance about the association, which affects learning. The third is the goal of the experiment, and measures how well the participant learns from previous errors by making better choices with each new trial.
The results of the experiment showed that neither insulin sensitivity nor liraglutide affected general task performance, with more predictable outcomes being associated with more accurate predictions, that is, smaller prediction errors.
The researchers used insulin sensitivity as a marker of decreased metabolic sensing. They found that obesity, which leads to insulin resistance, is associated with reduced adaptive learning.
When liraglutide was used, this impairment vanished in both men and women with obesity. This led to the restoration of impaired sensory association learning to normal levels.
However, liraglutide reduced the learning rate in the group with normal insulin sensitivity. Since the drug produced twice as large an effect on learning in the group with impaired insulin sensitivity compared to that with normal insulin sensitivity, the overall effect on learning was comparable between both groups on liraglutide.
With fMRI analysis, they found that adaptive prediction errors were associated with activity in the nucleus accumbens (NAc) and ventromedial prefrontal cortex of the brain. With liraglutide administration, this encoding was upregulated in the NAc and the subcallosal area in participants with insulin resistance but not the others. This confirms the behavioral experiment findings.
Thus, both arms of the experiment show that liraglutide restores associative learning in people with impaired insulin sensitivity and that this is achieved by increasing the DAN-mediated encoding of adaptive prediction errors in these brain areas. Aberrations of homeostasis thus play a feedback role in the encoding of prediction errors.
Thus, the central effects of obesity, acting on the mesoaccumbens pathway via changes in insulin sensitivity, affect associative learning. However, such feedback signals can be impacted by signals such as GLP-1 from the peripheral tissues.
What are the implications?
The findings indicate that liraglutide is associated with the normalization of associative learning in obese individuals with insulin resistance. In other words, “GLP-1 receptor activation modulates associative learning in people with obesity.” This may be one mechanism of liraglutide-associated weight loss in obesity.
The study thus yields proof that metabolic signaling can modulate neural pathways that mediate learned human behavioral responses according to the internal state of the body. The study makes headway in exploring the clinical effects of GLP-1 agonists.
