The Role of Theta and Beta Oscillations in Cognitive Control

Post by Shireen Parimoo

What's the science?

Cognitive control is the ability to flexibly adapt behavior in the face of changing goals or demands, and it involves suppressing irrelevant information in the presence of conflict. The medial prefrontal cortex (mPFC) is thought to be involved in exerting control during conflict or during errors, after which behavior must be adjusted. Recent research also suggests that neural oscillations in the subthalamic nucleus (STN) are closely linked with those in the mPFC during conflict. However, it is unclear how changes in neural activity in the mPFC and STN are related to behavioral responses during tasks requiring varying levels of cognitive control. This week in Brain, Zavala and colleagues used intracranial electroencephalography and electrode recordings to investigate oscillatory activity in the mPFC and STN during a cognitive control task.

How did they do it?

Neural activity was recorded from 22 Parkinson’s disease patients who were undergoing deep brain stimulation surgery. Local field potentials and multiunit spiking activity were recorded from the STN using depth electrodes, and intracranial electroencephalography recordings were obtained from the superior mPFC. The participants performed a cognitive control task with three trial types: Go, Conflict, and No-go trials. In the Go trials, an arrow was presented on the screen, and participants had to move a joystick in the direction of the arrow; in the Conflict trials, participants had to move the joystick in the opposite direction to the arrow on the screen. In the No-go trials, they had to withhold their response. Feedback was provided at the end of each trial. The authors determined the average oscillatory power (or amplitude) for each trial type in the theta (2-5 Hz) and beta (8-30 Hz) frequency bands, which provided an index of the strength of oscillatory activity on a given trial. They also computed the phase coherence between the STN and the mPFC, which is a measure of the consistency of the phase of oscillatory signals. Higher phase coherence indicates that there is greater synchronization of activity between the two regions.

What did they find?

Participants made more errors and were slower on the Conflict trials. They were also slower and less accurate on Go trials that occurred after Conflict trials, compared to consecutive Go trials. In the No-go and Conflict trials, just before participants made a response, there was an increase in theta power in both the mPFC and the STN. This increase occurred several hundred milliseconds earlier in the mPFC than the STN and was greater during Conflict trials than the Go trials. During No-go trials in which a response had to be withheld, there was a large increase in theta power in the mPFC but a small change in the STN. Similarly, multiunit activity in the STN decreased during No-go trials, but greatly increased during Conflict trials that required inhibiting a prepotent response and performing the opposite movement. This indicates that the mPFC is involved in providing top-down signals to indicate that behavior must be adjusted, whereas the STN is responsible for modulating the actual behavioral response. Theta  phase coherence was higher during Conflict and No-go trials than during the Go trials. As pre-response theta power increased before participants made a response, it suggests that theta oscillations are involved in adjusting cognitive control within trials, especially when the demand for control is high.

Beta power in the STN decreased prior to the participants’ response, whereas beta power in the mPFC increased after the response. In the mPFC, post-response beta power was lower during Conflict trials and when participants made an error. Conversely, beta power in the STN on a given Go trial was modulated by Conflict and errors on the previous trial. That is, beta power was higher on Go trials that occurred after Conflict trials. A similar pattern of activity was observed during Go trials that occurred after participants made an error, compared to Go trials that occurred after participants provided a correct response. Thus, beta oscillations in the STN may regulate responses in situations requiring a change in behavior, such as after an error is made or after high conflict trials.

What's the impact?

This study is the first to show how changes in oscillatory activity in the mPFC and the STN contribute to different aspects and temporal stages of cognitive control. Theta oscillations are involved in control during a trial, while beta oscillations play a role in exerting control across trials, particularly when control demands are higher. This research has important implications for understanding conditions in which cognitive control is impaired, such as schizophrenia and dementia.

Zavala et al. Cognitive control involves theta power within and beta power across trials in the prefrontal-subthalamic network. Brain (2018). Access the original scientific publication here.

The Paraventricular Thalamus Encodes the Salience of Stimuli and Regulates Associative Learning

Post by: Amanda McFarlan

What's the science?

It is not well understood how the brain is able to filter complex sensory inputs and decide what is most important, or ‘salient’. Recent studies have shown that the thalamus, a hub for sensory information (among other functions) in the brain, may contribute to this filtering process. The paraventricular thalamus (PVT), one of the subnuclei in the thalamus, is of particular interest. This region of the thalamus is a relay station connecting the brainstem (which provides information related to internal bodily states) to limbic brain regions involved in learning in a variety of emotional contexts. The PVT is connected to several higher-level brain areas, including the frontal and insular cortices and innervates the amygdala (which is involved in processing salience in an emotional context) and therefore could play an important role in encoding the relevance of stimuli. This week in Science, Zhu and colleagues explored the role of PVT neurons in determining the salience of behaviourally relevant stimuli and their contribution to associative learning.

How did they do it?

In mice, the authors tested several behavioural paradigms using Pavlovian conditioning (see experiment described below), optogenetics and calcium imaging to determine the role of PVT neurons in encoding salient stimuli. First, they performed stereotaxic injections of a virus expressing a genetically encoded calcium indicator (AAV-GCaMP6m) in PVT neurons. After a minimum two-week recovery period from surgery, they used fiber photometry to record calcium signals in head-fixed mice across days of associative learning. The mice, either water-restricted or sated (not water-restricted), were trained to pair an odor (‘conditioned stimulus’ in Pavlovian conditioning) with an outcome (unconditioned stimulus). There were three possible outcomes: rewarding (5 µl or 15 µL of water), neutral (nothing) or aversive (puff of air or tail shock). In the associative learning paradigm, mice were presented with the odor for 1s, followed by a 2s-delay period and then the outcome. In a subsequent experiment, PVT neurons were transfected with a virus expressing a light-gated ion channel that inhibits neuronal activity (to determine the effects of PVT inactivation during associative learning). In this experiment, PVT neurons were optogenetically inhibited in water-restricted mice during associative learning at three possible time points: the odor cue + 2-s delay period, the outcome (water) or between trials (intertrial interval).

What did they find?

The authors determined that the initial exposure to an unfamiliar odor evoked a robust response in PVT neurons that rapidly diminished as the mice became habituated to that cue. In the first associative learning paradigm, water-restricted mice were trained to pair an odor with a rewarding or aversive cue of varying intensity or a neutral cue. They found that PVT neurons responded to both the odor and outcomes in rewarding and aversive trials and that the magnitude of the response was graded to reflect the intensity of the reward. For example, PVT neurons showed greater activity in response to 15  µL of water versus 5 µL, and to a tail shock versus a puff of air. Together, these findings suggest that PVT neurons can encode a variety of stimuli (i.e. novel, rewarding, aversive) and their behavioural relevance. In the second associative learning paradigm, water-restricted and sated mice were trained to pair an odor cue with a water reward. The authors determined that the odor cue evoked robust anticipatory licking in thirsty mice but not sated mice.

Consistent with these findings, the activity of PVT neurons, as measured by calcium signals, was suppressed in sated compared to thirsty mice. Notably, PVT neurons had a greater response when sated mice were presented with an aversive cue (an air puff in the eye) compared to thirsty mice. This indicates that the aversive cue (air puff) became more salient when homeostatic needs had been satisfied. These findings suggest that PVT activity reflects the dynamic nature of stimulus salience after contextual changes. Finally, optogenetically inhibiting PVT neurons while delivering the odor cue or reward cue during training greatly decreased the number of anticipatory licks in water-restricted mice. In contrast, inhibiting PVT neurons between trials had no effect on the number of licks. Taken together, these findings indicate that the PVT is important for the formation but not expression of conditioned reward-seeking behaviour.

What's the impact?

This is the first study to show that the PVT encodes information about salient stimuli, including novel, rewarding and aversive stimuli, in a context-specific manner. The PVT has an important role in determining the salience of a stimulus, although how this salience information is communicated throughout the rest of the brain remains unknown. Elucidating the neural mechanisms involved in identifying salient stimuli and its impact on associative learning may provide insight into new therapies for the treatment of disorders like addiction.

Zhu et al. Dynamic salience processing in paraventricular thalamus gates associative learning. Science (2018). Access the original scientific publication here.

The Roles of Learning and Value Comparison in Suicidal Behavior

Post by Kayla Simanek

What's the science?

While society may view suicide attempts as strategic, from a clinical standpoint, suicidal behavior is often attributed to impaired decision-making in moments of crisis. One possibility may be that suicidal individuals do not optimally incorporate transient experience with long-term values, goals, and prior knowledge. However, this hypothesis may oversimplify the complexity of decision-making during crises. Two more specific factors related to learning that may contribute to a suicide attempt are: 1) poor integration of recent experience with prior experience and learned values (i.e. disrupted learning of expected values) and 2) an impaired ability to compare the worth or value of two options and (i.e. compare learned values when faced with choice). This week in Biological Psychiatry, Dombrovski and colleagues tested these two hypotheses to distinguish between the roles of value learning and value comparison in people with histories of suicide attempts.

How did they do it?

The authors conducted a three-choice decision making task. Across 300 trials in the task, participants picked from three pictures on a screen (each of which had a different probability of being the ‘correct’ picture over a series of trials) and were rewarded a small amount of money for picking the correct picture. 260 adults participated in the study and were separated into four groups: major depression with suicide attempts, major depression with suicidal thoughts, non-suicidal depression or healthy (non-depressed). Groups were further stratified by lethality of previous suicide attempts to assess the correlation between attempt severity and task performance. Learning during the task was analyzed by a participant’s responsiveness to reinforcement (monetary compensation). The time taken to decide between three choices was measured as a reflection of the participant’s ability to learn from feedback. Participants’ ability to differentiate between values was also measured by response time: choices close in value would theoretically increase the time taken to decide. Finally, analysis of decision trends within groups was conducted. The authors looked at whether each group chose the best or second-best option (exploitative choices) or a third, obscure option (exploratory choice) most regularly to determine difference in choice preference between groups.

What did they find?

The authors found that those who had attempted suicide were less responsive to reinforcement compared to other groups. Additionally, rewarding feedback slowed the decision time of suicide attempters more than non-attempters or those with suicidal thoughts. Participants with attempts considered to be highly lethal were even less responsive to reinforcement and made slower decisions compared to those participants with less lethal suicide attempts. This data suggests that participants with suicide attempts have difficulty learning worth based on outcomes (supports hypothesis #1). Suicide attempters’ decision time was also slowed by choices with subtle differences in value (probability of being the correct choice), suggesting that these participants have trouble distinguishing between closely ranked values. This conclusion supports hypothesis 2 and could explain the over-estimation of the value of suicide in times of crisis. Finally, analysis of answer choice frequency within groups revealed that suicide attempters more often chose exploitative options whereas non-attempters more often chose the exploratory option. This suggests that the suicidal individual’s ability to search for and consider alternative options is impaired.

What's the impact?

This is the first study to compare the contribution of learning expected value and comparison of learned value in depressed suicide attempters. This study found that both learning via reinforcement and value comparison are more impaired in individuals with major depression and a history of suicidal behavior than those without. This understanding may help shape treatment of suicidal individuals and prevent lethal outcomes in times of crisis.

Dombrovski et al. Value-based choice, contingency learning and suicidal behavior in mid-life and late-life depression. Biological Psychiatry (2018). Access the original scientific publication here.