The Role of Locus Coeruleus Noradrenergic Modulation in
Decision-Making and its Implications for ADHD Etiology
Monika Graumann Student ID: 10197214
Supervisor: Dr. Tobias Donner
Bachelor Thesis Clinical Neuropsychology University of Amsterdam
Words abstract: 175 Words review: 8594
Contents
Abstract 3
The Role of Locus Coeruleus Noradrenergic Modulation in Decision-Making and its
Implications for ADHD Etiology 4 Theoretical Accounts of Noradrenergic Modulation in Decision-Making 7
The Role of Noradrenergic Modulation in Decision-Making:
Evidence for Pupillometry as Index for LC Activity 11
Findings from Pupillometry Studies 15
Implications of Noradrenergic Modulation for ADHD Etiology 22
Conclusions and Discussion 32
References 38
Abstract
The locus coeruleus is a brainstem nucleus and forms the origin of norepinephrine neurons in the brain. Research has revealed its role in arousal, attention and decision-making. Theories describe the putative functions of the tonic and the phasic modes that its neurons can exhibit. A high tonic mode is associated with inattention, a low tonic mode combined with a high phasic activity is associated with task-focused attention induced by neural gain. Pupillometry has been proven to be a useful measure to explore noradrenergic modulation of decision-making in humans. Studies show that the LC is active in the time interval of decision-making and predicts the content of a decision in conservative subjects and tonic gain was found to shift attention towards an individual’s bias. Gain was reflected in the fMRI in high functional connectivity and clustering. Pharmacological studies investigating the role of this modulatory system in ADHD indicate that NE levels are aberrant in patients and that medication restores LC functioning towards high phasic, low tonic levels. However, future research is needed to confirm this view.
The Role of Locus Coeruleus Noradrenergic Modulation in Decision-Making and its Implications for ADHD Etiology
The neuromodulator norepinephrine (NE) has been implicated in various psychiatric conditions such as depression and attention deficit/hyperactivity disorder (ADHD) (Robbins & Arnsten, 2009; Aston-Jones & Cohen, 2005). Medication
targeting the NE neuromodulatory system is successfully used to treat these conditions (Chamberlain et al., 2007). However, the
mechanisms by which NE modulates attention and decision-making as well as the mechanisms by which medication remediates symptoms of ADHD are not well understood.
The locus coeruleus (LC) is a small brain stem nucleus situated in the pons and is the source of the brain’s
noradrenergic neurons. It sends projections to all cortical regions, thalamus, amygdala and to other neuromodulatory nuclei and is the sole source of NE in the forebrain (Sara &
Bouret, 2012). It has traditionally been linked to the sleep-wake cycle and arousal (Sara, 2009) but research of the last years has revealed its modulating role in various cognitive functions like in perception, learning, memory and attention (Bouret & Sara, 2005). For example, LC neurons in monkeys and rodents reliably respond to salient or conditioned sensory stimuli in vigilance tasks with phasic bursts of activity (Bouret & Sara, 2005; Sara & Bouret, 2012). This response is more tightly linked to the behavioral response than to the sensory features of the stimulus, but not related to
Norepinephrine: Catecholamine
neurotransmitter with a modulating role (modulates glutamate and GABA neurotransmission). Main function is the regulation of arousal (Aston-Jones & Cohen, 2005) or the sleep-wake cycle (Sara, 2009).
the motor act, indicating a role in decision-making (Bouret & Sara, 2005). Additionally, accuracy in sensory discrimination tasks is promoted by phasic LC activity, as correct trials are preceded by larger LC responses than incorrect trials (Aston-Jones & Cohen, 2005). LC activity was also found to increase the signal-to-noise ratio in neurons (Sara, 2009). Furthermore, activation of the LC before cognitive tasks that depend on
prefrontal cortex (PFC) has been observed, accounting for the diffuse innervation of these cortical areas by LC neurons. For example, the LC has been found to mediate shifting of behavioral task dependent responses where a change of strategy is required (Sara, 2009; Bouret & Sara, 2005). These observations are thought to demonstrate the LC's role in attentional and cognitive shifts and adaptation to environmental imperatives (Sara, 2009). Low tonic LC activity in contrast, is seen during drowsiness and states of low attention (Sara, 2009; Aston-Jones & Cohen, 2005). The enhancement of cognitive functions by LC activity was found to be dose-dependent and follow an inverted-u relationship, where both excessively low and excessively high levels of NE lead to deterioration of cognitive performance (Aston-Jones & Cohen, 2005).
In line with these findings, it has been suggested that the LC optimizes decision-making by facilitating adaptive behavioral responses and attention (Aston-Jones and Cohen, 2005). However, the specific role of LC in decision-making is not clear. Which effects does LC modulation exhibit on decision-making and what
Phasic mode: Burst of temporally
accumulated spikes, often alternating with phases of reduced spiking.
Tonic mode: Continuous baseline
activity which can have variable frequencies.
consequences does malfunctioning of this modulatory system have? It has been suggested, that LC phasic bursts after stimulus presentation
and before onset of the behavioral response reflect the finalization of the decision process (Aston-Jones & Cohen, 2005). A recent study, however indicates that the LC might be involved during the process of decision-making (de Gee, Knapen & Donner, 2014), therefore clarification of the role of LC modulation in decision processes is needed as well as consequences of the deterioration of this system.
Studies have shown that medication which targets the NE system can mimic ADHD symptoms in monkeys (Ma, Arnsten & Li, 2005) and in humans (Swann, Birnbaum, Jagar,
Dougherty & Moeller, 2005) and pharmacological elevation of NE levels can be used as a treatment for this disorder. ADHD subjects have been found to perform worse on simple decision-making tasks (Marzinizik, Wahl, Krüger, Gentschow, Colla & Klostermann, 2012; Tamm, Menon & Reiss, 2006), suggesting a link between
decision-making, ADHD symptoms and NE modulation. Given the well-established link between LC function and attention in previous studies (Sara, 2009), this review will also attempt to clarify the role of noradrenergic modulation in the etiology of ADHD.
The first section describes theoretical accounts of the NE modulation of cognitive processes, in particular decision-making. The second section discusses
Neuromodulation: Modulation
of signal transmission by tuning of neural excitatory and inhibitory signals (Sara, 2009). Neuromodulation can alter the efficacy of signal transmission by enhancing the signal-to-noise ratio or by enhancing neural gain. In gain modulation, the responses of glutamate or GABA neurons to the input of other neurons is enhanced.
some recent studies on the noradrenergic modulatory system in decision-making. The last section reviews some findings on medication for ADHD targeting the LC-NE modulatory system and inferences which can be drawn from them for ADHD
etiology.
Theoretical Accounts of LC Noradrenergic Modulation in Decision-Making
Aston-Jones and Cohen (2005) reviewed some of the evidence available at the time on noradrenergic modulation of cognition. They used different findings as a fundament for their theory on NE modulation of cognition, such as the inverted-u relationship between tonic LC activity and task performance, phasic activation of monkey LC neurons as response to salient stimuli and the P3 in humans. Their theory essentially posits that the two LC activity modes (tonic and
phasic) are triggered by environmental characteristics and induce neural gain to fulfill adaptive, task-related functions. The phasic mode is triggered by well defined, predictable environmental characteristics where correct decisions are rewarded. Phasic bursts of LC activity follow the consolidation of a decision. The authors refer to this mode as exploitation of situational resources. The tonic mode is triggered by less predictable situations in which exploration is more rewarding in the long run. The theory suggests that exploitation corresponds with high task-focused attention
Gain: Increased responsivity of a
cell to input from another cell. This can be induced by a neuromodulator. Excitatory signals become more excited and inhibitory signals become more inhibited (Eldar, Cohen & Niv, 2013). The higher the gain, the more marked is the contrast between excitation and inhibition (Aston-Jones & Cohen, 2005).
and higher accuracy whereas exploration is associated with a wider, less focused mode of attention. They speculate that the adaptive switching between these two modes could be mediated via connections between the LC and the anterior cingulate cortex (ACC) and the orbitofrontal cortex (OFC). The ACC is part of the ventral attention network (Walz et al., 2013) and is involved in the detection of behavioral errors (Gazzaniga, Ivry & Mangun, 2009) and the OFC has been linked to impulsive behavior (Gazzaniga et al., 2009). Reciprocal modulation of activity between these structures and the LC is suggested as a candidate mechanism for adaptive switching between LC modes leading to optimization of behavior, according to Aston-Jones and Cohen (2005).
One core assumption of this theory is that adaptive gain is the core feature via which optimal decision-making is accomplished. Thus, adaptive gain facilitates exploitation during the phasic mode and exploration during the tonic mode,
according to the theory. This assumption has been tested in a computational model, which used a two-layer network to simulate the influence of gain on
decision-making, showing a clear advantage of adaptive gain compared to fixed gain (Shea-Brown, Gilzenrat & Cohen, 2008). The model consisted of a sensory and a response layer, both connected via modulatory connections to a "gain node". In the fixed gain simulation, gain was held constant throughout the trials. In the adaptive gain
simulation, gain increased after crossing a gain threshold which in turn facilitated crossing of the response threshold during correct trials, leading to a higher rate of correct responses in a 2AFC task. The model thus predicts more accurate
making with adaptive gain, confirming assumptions made by Aston-Jones and Cohen's theory of modulation of cognition (2005).
Another assumption made by this theory is that tonic LC activity follows an inverted-u relationship with task performance, exhibiting low performance at low and high levels, but high performance over the middle range of LC activity.
Eckhoff, Wong-Lin and Holmes (2009) tested this assumption in a biophysical network model of norepinephrine modulation of decision-making. Simulating NE modulation on a cellular level, they provided theoretical evidence that behavioral performance can be optimal over a broad middle range under NE modulation,
replicating an inverted-u relationship between NE modulation and task performance. Their model comprised two selective glutamatergic pyramidal cell populations, each one being selective for one type of visual stimuli of the visual detection task and one nonselective glutamatergic pyramidal cell population. All of these were connected via excitatory nodes to each other and to a population of GABAergic interneurons, which in turn had inhibitory connections to the populations of pyramidal cells. All cells received noisy input and the pyramidal cells additionally received input from the task stimuli.
The simulations yielded a number of results. First, their model produced impulsive and unmotivated behavior as a function of tonic NE modulation at GABA synapses. Less inhibition resulted in impulsive behavior which occurred when the response threshold was crossed prior to stimulus onset. Too much inhibition yielded unmotivated behavior, as in trails where no choice was made after stimulus
presentation. Impulsive behavior was a result of high tonic modulation and
unmotivated behavior was a result of low tonic modulation. Phasic NE modulation preceded threshold crossing and enabled faster responses. Second, changes in robustness of the optimal range as a function of NE modulation corresponded with spike counts, with a rising number of spikes in the medium range and a decaying number of spikes as NE levels rose.
This model indicates that NE levels modulate the affinity of a neural network to stimulus detection during decision-making. As proposed by Aston-Jones and Cohen, the optimality of the modulation follows an inverted-u relationship, with high performance in the middle range of NE levels and worse performance as NE levels reach high or low levels.
These two computational models yield theoretical evidence for the aforementioned theory of noradrenergic modulation of cognition. Two of the assumptions made by this theory were implemented in the models and results confirmed predictions made by the theory. First, the facilitation of performance observed in cognitive tasks as in decision-making tasks is likely the result of adaptive gain induced by NE. Second, this facilitation of cognitive performance follows an inverted-u relationship, with high performance being possible over a broad middle range of LC activity, but marked decreases in performance at the low and high ends of NE levels.
This theoretical evidence for NE modulation represents a framework for further empirical studies which can test predictions made by them in vivo using
animal models or human subjects. The model by Shea-Brown et al. (2008) represents a simple model of the complex multilayered structure of the brain. If fundamental ideas can be replicated in empirical studies, it will provide a useful model for understanding the mechanisms via which gain facilitates decision-making.
Furthermore, the model by Eckhoff et al. yields some interesting results concerning the relevance of noradrenergic modulation for ADHD. In their model, low NE levels resulted in unmotivated behavior and high NE levels resulted in impulsive behavior. If unmotivated behavior is used as a synonym for inattentive behavior, then both types of behavior can be seen in ADHD patients (Franken, Muris & Denys, 2013) and medications targeting the noradrenergic modulatory system can remediate these symptoms (Chamberlain et al., 2007). This model makes specific predictions about the way in which NE modulation might be dysfunctional in these patients. These predictions in turn can be tested and can provide further information about the etiology of ADHD.
The Role of Locus Coeruleus Noradrenergic Modulation in
Decision-Making:
Evidence for Pupillometry as Index for LC Activity
One idea brought forward by Aston-Jones and Cohen (2005) is that phasic LC responses follow correct decisions and facilitate adaptive behavioral responses. In line with this idea, studies found accurate responses in decision-making tasks to be accompanied by phasic LC activity (Clayton, Rajkowski, Cohen & Aston-Jones, 2004;
Gilzenrat, Nieuwenhuis, Jepma & Cohen, 2010) and one fMRI study found larger BOLD responses for target stimuli and larger responses were correlated with lower reaction time (Murphy, O’Connell, O’Sullivan, Robertson and Basters, 2014).
However, these findings failed to be replicated in other studies, where LC phasic responses did not co-occur more often with correct compared to incorrect decisions (Eldar, Cohen & Niv, 2013; de Gee, Knapen & Donner, 2014). The mixed results indicate that the role of noradrenergic modulation might not be as simple as
facilitating accurate decisions and the question that arises is, what the exact influence is and how noradrenergic modulation exhibits its influence on decision-making.
The LC has traditionally been linked to the sleep-wake cycle and to arousal Sara, 2009). One feature of sympathetic arousal is pupil dilation. LC activity in monkeys correlates remarkably well with pupil dilation (Rajkowski, Kubiak & Aston-Jones, 1993, as cited in Aston-Jones& Cohen, 2005). As a result, measures of pupil diameter and pupil dilation have been suggested as a proxy variable to measure LC activity in humans and a number of studies provide evidence for its validity for this purpose (for example Jepma & Nieuwenhuis, 2011; Murphy,
Robertson, Balsters & O’Connell, 2011; Murphy, O’Connell, O’Sullivan, Robertson & Balsters, 2014).
For example, pupil diameter in humans shows the same task-related changes as tonic LC activity in the monkey (Gilzenrat et al., 2003, as cited in Aston-Jones & Cohen, 2005) and pupil dilation shows comparable task-related reactions as seen during phasic LC activity (Beatty, 1982a,b, Richer & Beatty, 1987, as cited by
Jones & Cohen, 2005). Also, pupil diameter correlated with task strategy in a gambling task, with high diameter occurring during exploratory phases (Jepma & Nieuwenhuis, 2011) which the authors interpreted as tonic LC responses. Another study found that the relationship between pupil diameter and task performance exhibits the same inverted-u relationship as found between LC activity and task performance (Murphy, Robertson, Balsters & O’Connell, 2011) and this is the same relationship that the computational model by Eckhoff et al. (2009) predicts for LC activity and task performance.
In an fMRI study, Murphy, O’Connell, O’Sullivan, Robertson and Balsters (2014) provided further compelling evidence for pupillometry as measure for LC activity. They compared the correlation between pupil diameter and BOLD response in the LC during rest and during the oddball task. The oddball task is a simple sensory discrimination task often used to measure LC task-related activity and
reliably provokes phasic LC responses after target presentation (e.g. Rajkowski et al., 2004, as cited by Aston-Jones and Cohen, 2005). The correlations between pupil diameter and BOLD response were overall higher during the task compared to the resting state, likely due to the fact that BOLD responses reflect changes in blood flow and there might be less variability in blood flow during rest. Target stimuli evoked larger responses than non-target stimuli in LC BOLD response, replicating earlier findings of LC task-related responses. This study indicates the utility of pupillometry as a measure for LC activity in humans.
One limitation of this study is that the pupil measure used here, cannot directly distinguish between phasic and tonic responses. However, another study (Gilzenrat et al., 2010) provides evidence that the oddball task evokes phasic LC responses in humans just as in monkeys, and that these phasic responses are reflected in pupil dilations. Gilzenrat, Nieuwenhuis, Jepma and Cohen (2010)
replicated in a human sample pupil dilations which had previously been observed in monkeys in the oddball task . In monkeys, these pupil dilations occurred
simultaneously with phasic LC activation, suggesting that pupil dilation in humans also reflects phasic activation of LC neurons.
Walz, Goldman, Carapezza, Muraskin, Brown and Sajda (2013) found further evidence, that the human LC responds with phasic activation to target stimuli during decision-making. Using the oddball task, they found activity in the brainstem in a time window thought to correspond to the P3. The P3 is an event-related potential in the EEG and has been linked to phasic LC activity (Nieuwenhuis, Aston-Jones & Cohen, 2005). Walz et al. (2013) conducted a simultaneous EEG-fMRI study which revealed activity in the brainstem short after the P3, indicating that human LC, just like in monkeys, responds to target stimuli with phasic bursts of activity. Other activated regions compromised the ACC, OFC and the right inferior frontal gyrus (rIFG) among others.
Taken together, these studies provide strong converging evidence that pupil diameter reflects tonic LC activity and that task-related LC activity is phasic.
Although no direct measurements in humans are possible, the evidence described 14
above builds a strong link between LC activation and pupillometry as it replicates findings with a variety of methods such as animal models, computational models, EEG and fMRI.
Findings from Pupillometry Studies
One of the well-replicated findings in animal models is that the LC responds with phasic activation to salient stimuli (Sara, 2009) and high tonic activity is associated with less focus (Aston-Jones and Cohen, 2005) and more impulsive behavior (Eckhoff et al., 2009).
Gilzenrat et al. (2010) provided supporting evidence for these findings in humans using pupillometry. Their data suggest that the phasic mode and high tonic activity exhibit something that resembles a mutually exclusive relationship. They found that, during an oddball task, small baseline diameter predicted more task evoked phasic dilations and that large baseline diameter predicted less amplified task-related
dilations. The study shows that during a high tonic LC mode, fewer phasic bursts are seen and that phasic bursts are promoted by low baseline tonic activity.
These results are supported by an fMRI study by Eldar, Cohen and Niv (2013) which showed that task-relevant stimuli were accompanied by low baseline pupil diameter and large pupil dilations and by stronger BOLD responses, indicating that task-relevant stimuli evoke phasic LC bursts alternating with low tonic activity.
Aston-Jones and Cohen (2005) speculate in their theory of NE modulation of cognition that these alternations between the two LC modes are adaptive and that
each mode serves a function, being exploitation in the phasic and exploration in the tonic mode.
In an additional experiment, Gilzenrat et al. (2010) tested this idea by
manipulating the predictability of the environment to test whether this would trigger the corresponding LC modes. They used a pitch discrimination task and they
manipulated task difficulty to induce low predictability as trials became more difficult and high predictability as trials became easier. A more predictable
environment was associated with low baseline diameter, high phasic dilations and an unpredictable environment was associated with high baseline diameter and fewer phasic dilations, corroborating the theory by Aston-Jones and Cohen. A problem with this interpretation however is, in how far the exploratory mode reflects
exploration and not just waning of interest in the task and deterioration of attention. The theory does not make any testable predictions which could distinguish
exploratory behavior from waning attention and distractability, so the only firm conclusion that can be drawn from this study is that difficult tasks can trigger the tonic mode and that tasks that the subject is able to handle trigger the phasic mode, as expressed in low tonic activation and phasic target related responses.
Another assumption that is part of this theory is that phasic responses are driven by the outcome of a decision, reflecting the finalization of a decision process in which the LC is not directly involved. This assumption is a product of the interpretation of ambiguous data, as of a study by Clayton, Rajkowski, Cohen and Aston-Jones (2004). The study showed that LC activity was more closely related to the behavioral
response than to sensory features of the target. This finding was replicated in the computational model by Shea-Brown et al. (2008) who found that the simulated LC discharge was more tightly linked to the response than to the stimulus. Clayton et al. (2004) interpreted their results as evidence that the LC signals the finalization of a decision process and is not involved in decision-making. However, the opposite conclusion, that the LC does play a role in decision-making, is not ruled out by their data and more recent evidence shows that the LC is in fact active during the process of decision-making, indicating its direct involvement in this process.
Also, Walz et al. (2013) concluded from their EEG-fMRI data that the LC is activated by the outcome of a decision. However, they only included data into their analysis starting from 175 milliseconds after stimulus onset. The following study provides evidence that LC activity starts as soon as stimulus onset and exhibits a sustained response until after the choice is made, suggesting that LC contributes to decision-making.
De Gee, Knapen and Donner (2014) measured pupil dilation during a visual detection task. Their results suggest that the LC is part of the decision-making process as pupil dilation started to increase directly after stimulus onset and
exhibited a sustained dilation, peaking around the onset of the behavioral response. This sustained response made a more substantial contribution to the overall response than the phasic component which accompanied and followed the behavioral
response, indicating activity of LC neurons during the process of decision-making.
Furthermore, their results showed that pupil dilations predicted the content of a decision depending on individual response bias. Gilzenrat et al. (2010) found that Hits evoke larger dilations than Correct Rejections, suggesting that Yes answers evoke larger LC responses than No answers. De Gee et al. (2014) also found that Yes answers evoked larger responses than No answers. Interestingly, further analysis showed that this finding was highly dependent on individual response bias of the subject: conservative subjects, responding Yes against their bias displayed larger pupil dilations than during No responses. In liberal subjects, this difference in pupil dilations between Yes and No answers was not significant. This demonstrates that the difference in pupil dilation between yes and No answers were solely due to the increased dilations during Yes answers of conservative subjects. More importantly, these results provide evidence that LC activation reflects the content of decisions and the subject’s individual bias. The results from this study combined, suggest that the LC is involved in the decision process. It exhibits a sustained response in the decision interval and the magnitude of dilation predicts the content of the decision in
conservative subjects.
LC activity can also influence the individual learning style and have effects on the whole brain as shown in a study by Eldar, Cohen and Niv (2013). Their study indicates that neural gain induced by tonic LC activity, shifts attention during learning towards idiosyncratic predispositions. They based their predictions on a neural network model which predicted that the higher the gain, the more the
learning performance would depend on individual bias. To investigate this question 18
they let their subjects do a trial-and-error learning task. They used a 2AFC task in which the subjects were presented with two objects on each trial which could be distinguished based either on a semantic feature or on a visual feature. The correct answer could either be the object with the semantic feature or the object with the visual feature. For example, in one trial the correct answer could be the office related object as opposed to a food-related object (semantic feature) and on another trial the correct answer could be the grayscale object as opposed to a color object (visual feature). Feedback was provided after each trial so that trial-and-error learning was rendered possible. Additionally, each subject filled out a questionnaire to assess their individual predisposition to process either visual or semantic features, the Index of Learning Styles questionnaire (ILS). During the task, pupil dilations were measured as an inverse measure of pupil diameter. The results corroborated their network model’s predictions: the magnitude of the correlation between ILS score and
performance in the preferred domain correlated with baseline pupil diameter. Thus, pupil diameter predicted how much an individual’s learning was biased towards his idiosyncratic preference. This result demonstrates that during learning, neural gain can increase an individual’s bias to attend to features to which the individual is predisposed to attend.
The researchers were further interested in the question whether the enhanced signal strength induced by gain would be reflected in a brain-wide network. To explore this question, they combined various fMRI measures. First, they showed that task-related phasic dilations correlated with larger overall BOLD responses,
indicating that gain as reflected in pupil diameter, evokes brain-wide activation. The second measure was based on another simulation that predicted that higher gain would be associated with higher correlations (and anticorrelations) in functional connectivity. Functional connectivity was measured by analyzing the fluctuations in time signals among voxel pairs in predefined, arbitrary segmented parts of the brain. The results showed indeed that high diameter was more prevalent when functional connectivity was strong and low diameter was more prevalent when functional connectivity was weak. This suggests that tonic gain is reflected in stronger brain-wide connectivity. Third, their computational model predicted that neural gain should induce more specialized neural networks, with one signal affecting fewer nodes so that neural signals are more tightly clustered with high gain. Conversely, with low gain, neural networks were expected to be more widely distributed. As expected, they found a correlation between clustering and pupil diameter,
confirming the model predictions. Also, clustering was correlated with learning bias which they earlier found to be the result of neural gain.
The study by Eldar et al. shows that during trial and error learning, gain shifts attention towards the features the individual is predisposed to attend. It also shows that gain, as predicted by computational models, induces more functional
connectivity and tighter clustering of this connectivity, reflecting the amplifying effects gain is thought to have on neural networks. Both findings were brain-wide findings, demonstrating that gain affects the whole brain and is not, as previously thought locally bound to task-related structures.
As mentioned above, clustering was associated with shifts in learning towards idiosyncratic predisposition. These shifts in learning were induced by gain which indicates that clustering might be one of the neural correlates of gain.
The results suggest that clustering and functional connectivity could be measurable correlates of gain. However this idea needs confirmation in future
studies. What we can learn from Eldar et al.’s study about noradrenergic modulation of decision-making, is that high tonic NE levels can shift attention of an individual towards his predisposition during learning.
Taken together, pupillometry provides a useful measure to study noradrenergic modulation of decision-making in humans. First, it was used to replicate well-established findings of animal models, which demonstrated that LC displays phasic responses to target stimuli, alternating with low tonic activity and that high tonic activation blocks phasic activation. High tonic activation accompanies waning attention to the task. Second, LC is involved in the process of
decision-making and predicts the content of a decision in conservative subjects responding against their individual bias. Third, neural gain can be induced by tonic LC activity and shifts attention during learning towards the idiosyncratic predisposition of an individual. Amplified neural communication through gain is reflected in increased functional connectivity and tighter clustering of functional connectivity and can be measured with fMRI.
Implications of Noradrenergic Modulation for ADHD Etiology
Medication used for ADHD, such as methylphenidate targets the dopamine (DA) and NE systems. But also selective NE reuptake inhibitors (NRI) are prescribed successfully for the treatment of ADHD, such as atomoxetine. It is the only such approved agent for ADHD (Stahl, 2008), however other medication targeting the noradrenergic system are also prescribed for this disorder, for example reboxetine and clonidine (Stahl, 2008; Franken et al., 2013)
The computational model by Eckhoff et al. (2009) can be used to derive specific predictions about dysfunctional NE modulation in ADHD patients. First of all, NE levels in ADHD should be aberrant from the middle range in which optimal behavior occurs (Eckhoff et al., 2009). Second, depending on which maladaptive behavior is observed in a patient, NE levels might be either too low and cause unmotivated behavior or they might be too high causing impulsive behavior. If translated into DSM criteria, unmotivated behavior in the model might reflect the attention-deficit subtype and impulsivity might reflect the hyperactive subtype in this model, although high tonic levels could account for both subtypes. The Aston-Jones and Cohen model can be used to explain ADHD behavior, too, as inattentive and hyperactive behavior might be interpreted as a “chronic exploratory” mode caused by chronic high tonic NE levels.
Further, as the phasic mode has been associated with accuracy, low reaction time (RT) and focused attention, there might be a problem with inducing phasic
responses in ADHD. The problems of inducing the phasic mode could be a direct consequence of a chronic, high tonic mode (Gilzenrat et al., 2010).
Finally, another candidate for ADHD etiology might be a deficit in gain modulation. This idea is based on the model by Shea-Brown et al. (2008) which demonstrated that adaptive gain in a multilayered network gave rise to better performance in a decision-making task. Some of the abovementioned model's predictions can be combined when speculating that ADHD pathology is caused by deficient gain modulation due to deficient phasic activation, which is “blocked” by high tonic NE levels.
Frank, Santamaria, O’Reilly and Willcutt (2007) tested a computational model of NE dysfunction in an ADHD sample and showed that erratic switching behavior and RT variability are likely due to NE dysfunction. Their subjects, consisting of controls, ADHD patients and ADHD patients on stimulant medication, did a probabilistic selection task. On each trial, they were presented with two symbols of which one was correct and the other one was incorrect on 80% and 20% of the trials, respectively. Subjects learned by reinforcement to choose the correct symbol. The results showed an overall impaired performance in ADHD subjects compared to controls and medicated subjects. Erratic trial-to-trial switching and RT variability were increased in ADHD subjects.
These results could be interpreted as increased “exploratory behavior” due to high tonic NE levels in ADHD subjects, although this remains pure speculative, as the experiment does not assess LC modes. This could have been done by measuring
pupil diameter and dilations. A problem for the high tonic/exploration interpretation is the fact that medication lowered RT, decreased RT variability and increased
accuracy. If impaired functioning in ADHD subjects was due to high tonic NE levels, then how could medication that further increases NE levels improve performance? Lower RT and higher accuracy are associated with the phasic LC mode, so the
medication could have induced more phasic responses. Also, stimulant medication is not selective to the NE transporter (NET).
Another possible explanation is that tonic NE levels were not high in the first place, but too low, contradicting the Aston-Jones and Cohen predictions. In line with this idea, one study found the concentration of a NE metabolite in urinary excretion to be correlated with a measure of sustained attention in children with ADHD (Llorente et al., 2006). This correlation indicates that NE levels might indeed be too low, an idea which can also be derived from the simple fact that medication targeting the NE system blocks the NET and thus increases NE levels at the synapse. This idea is represented as unmotivated behavior in the Eckhoff et al. model and rising the NE levels in this model would move them towards the optimal middle range levels and could account for improved performance.
Impulsivity is a symptom which is more commonly associated with ADHD than unmotivated behavior. Impulsivity is caused, according to the Eckhoff et al. model, by high tonic levels of NE. Elevating NE levels with NRIs can also remediate impulsivity as shown in a study by Chamberlain et al. (2007).
Using a stop-signal paradigm to test response inhibition, they demonstrated that atomoxetine rendered the performance between medicated ADHD patients and healthy controls indistinguishable by lowering the RT compared to non-medicated patients. Also, it lowered the number of commission errors on a sustained attention task.
The effect of medication on impulsivity is contradictive to predictions made by Eckhoff et al. as it would mean treating high NE levels by further increasing them. A restriction in this study is, like in the Frank et al. (2007) study, that it is not
informative concerning the modulation of LC modes by medication. Another interesting questions is, where NE levels are elevated. Atomoxetine might not equally affect different structures, for example the prefrontal cortex (PFC) and the LC.
In a later study Chamberlain et al., 2009 showed that the beneficial effect of medication on impulsivity is mediated via the rIFG. This gyrus is known to be involved in response inhibition (Chamberlain et al., 2007). They compared healthy individuals on medication and on a placebo during a stop-signal task in the fMRI scanner. The results replicated earlier findings that atomoxetine shortens RT and improves response inhibition and the fMRI data additionally showed that this effect was mediated via the rIFG. The rIFG became active when participants attempted to inhibit their responses and this activation was stronger in medicated subjects. Also, plasma levels of atomoxetine correlated with the magnitude of activation during successful inhibition.
These results indicate that NRIs might exhibit their remediating effects on impulsivity via modulation of frontal brain areas such as, the rIFG. Consistent with this idea, NET density is high in prefrontal cortex (PFC), but low in subcortical structures as the striatum (Frank et al., 2007). Thus, the underlying mechanism of NRIs might be elevating too low NE levels in PFC. This conclusion contradicts the predictions made by the Eckhoff et al. model, which predicted high NE levels to cause impulsivity. A possible explanation might be that medication may have differential effects on PFC and LC NE levels. As predicted by Eckhoff et al., LC NE levels might be actually too high, but PFC levels might be low, causing high tonic activity and impulsive behavior. Studies are needed which image the LC and the PFC simultaneously under the influence of medication. Preferentially comparing ADHD subjects and healthy controls. Also, it would be interesting to measure NE levels in both structures under the influence of medication. As the study did not use ADHD subjects, it remains to be verified, if atomoxetine facilitates inhibition via the same mechanism as in healthy subjects and whether this is the only effect it has on PFC.
Interestingly, Walz et al. (2013) found the rIFG to be activated in temporal overlap with the brainstem during the oddball task. As this activation occurred in temporal vicinity of the P3, this brainstem activation is thought to reflect phasic LC activity. This could indicate that the rIFG is actually modulated via phasic LC activation. As mentioned above, a problem in ADHD might be deficient phasic LC activation, and this might in turn cause deficient activation in the rIFG. Medication
might remediate the “problem at its origin”, by restoring LC activity, into a low tonic, high phasic mode which in turn would facilitate inhibition of impulsive behavior via the rIFG. Restored phasic activity might also improve task-focused attention, as suggested by the Aston-Jones and Cohen theory and this might account for effects in the attention deficit subtype of ADHD.
To test this idea, first, the question has to be answered whether LC activity is indeed aberrant in ADHD and if so, in which way. As the LC is the main source of NE in the brain and NRIs treat ADHD symptoms, the conclusion that the LC is actually dysfunctional in ADHD patients is compelling, though not necessarily confirmed by studies of NRIs. Studies comparing LC activation in healthy subjects and patients, e.g. by using pupillometry are missing for this purpose. Aston-Jones and Cohen’s theory predicts that the LC might display too few phasic responses due to constant high tonic NE levels, as in a “chronic” exploratory mode.
Studies with another drug, modafinil, indicate that ADHD medication might exhibit beneficial effects by modulating LC activity, as predicted by Aston-Jones and Cohen's theory. Modafinil is a stimulant, not released on the market as ADHD treatment but successfully used for this purpose ( Stahl, 2008; Minzenberg, Watrous, Yoon, Ursu & Carter, 2008). Like methylphenidate it binds both to the NET and the DA transporter (DAT). Modafinil has been shown to improve motor inhibition and other cognitive measures in ADHD patients (Turner, Clark, Dowson, Robbins & Sahakin, 2004).
These effects might be mediated via task-related switches between the phasic and the tonic mode, as suggested by Minzenberg, Watrous, Yoon, Ursu and Carter (2008). In an event-related fMRI study, they compared task-dependent and –
independent BOLD responses of healthy subjects on modafinil and placebo. They found decreased tonic task-independent LC activity in the medication group. This decrease in baseline activity was accompanied by increased task-related phasic activation. The authors suggest that modafinil might lower tonic activity by
augmenting NE levels at the LC cell bodies, stimulating autoreceptors which causes decreased excitability of LC neurons. This mechanism could explain how medication which, overall heightens NE levels, could decrease LC activity.
This study demonstrates that medication targeting the NET can decrease tonic NE levels and increase phasic activations and an earlier study (Turner, Clark,
Dowson, Robbins & Sahakin, 2004) suggests that this improves cognitive
performance and inhibition in ADHD subjects. Minzenberg et al. (2008) also found a negative correlation between LC and PFC activation, which supports the idea that the low tonic mode might improve cognitive performance via phasic bursts. Phasic burst might exhibit their beneficial effects on cognition by promoting neural gain, as suggested by Shea-Brown et al.’s model (2008)..
Although the fMRI data by Minzenberg et al. seem compelling, some constraints should be considered. Doubts have been expressed regarding the activation pattern identified as LC (Astafiev, Snyder, Shulman & Corbetta, 2010). These critics doubt that the activation found by Minzenberg et al. reflects the
anatomical location of the LC. Also, activation of surrounding areas is difficult to rule out.
Furthermore, additional evidence is needed to confirm that high
task-dependent BOLD responses found here are indeed phasic in nature. fMRI has a low temporal resolution and is not specific regarding the duration of a neural response. Pupillometry might be useful to replicate these findings in humans.
A problem with the interpretation of these data as pure NE effect, is the fact that modafinil, unlike atomoxetine, is not selective to the NET but also inhibits the DAT. In general, NRI’s always partially affect DA levels, because the NET also takes up DA in PFC. Selective NRI’s thus elevate both NE and DA in PFC. However, in striatum, NET density is low and DAT is high, so any confounding effects via increased striatal DA levels cannot be ruled out, such as motor inhibition or reward-dependent effects. The same is true for methylphenidate, so research on
noradrenergic effects on ADHD symptoms should focus on atomoxetine as noradrenergic agent.
Another study used a selective NRI and also found a modulating effect. Kocsis, Li and Hajos (2007) observed rat EEG during sleep, awake in a familiar environment and during exploration of a novel environment. They compared the EEG of rats in which either a saline injection or a reboxetine injection had been administered. Reboxetine is a selective NRI, similar to atomoxetine and approved as an antidepressant, but also prescribed for the treatment of ADHD (Stahl, 2008). Theta waves in rats are associated with explorative behavior and sensory integration
(Kocsis et al., 2007). Their results revealed that reboxetine attenuated theta waves in a familiar environment and enhanced them in a novel environment compared to saline injections. The authors speculate that this reflects increased synchronization of
hippocampal neurons.
The studies by Minzenberg et al. and Kocsis et al. suggest that medication targeting the noradrenergic modulatory system can modulate this system’s functioning itself by altering task-dependent activation in an adaptive way.
Modafinil might enhance task-related attention and reboxetine might relief ADHD symptoms by lowering hyperactivity in situations which do not require high arousal and as in a familiar environment. Conversely, it might enhance attention by
increasing arousal in more cognitively demanding situations, as in a novel environment. Further confirmation of this hypothesis is needed in studies with humans, for example using EEG.
The results of these studies fit well with the Aston-Jones and Cohen model. In the light of this model and these studies, ADHD might be interpreted as a disorder of deficient adaptive behavior.
Unfortunately, the Kocsis et al. study did not investigate the effects of reboxetine in a rat model of ADHD. It would be interesting to see whether a rat model of ADHD would display deficient arousal in a novel environment and heightened arousal in the familiar environment compared to normal rats. Also, it remains to be seen whether reboxetine displays the same adaptive effects in ADHD patients.
In conclusion, the experiments discussed here indicate that ADHD might be caused by either too high or low NE levels or by dysfunctional LC modes. In the hyperactive subtype, LC tonic activity is most likely high (Eckhoff et al., 2009, Aston-Jones & Cohen, 2005), causing hyperactive behavior and distractibility. NRIs might lower tonic LC activity by acting at autoreceptors. Additionally, low tonic activity can promote the phasic LC mode (Gilzenrat et al., 2010) which enhances task-focused attention via modulation of the PFC. The inattentive subtype could be caused both by high or low tonic LC activity (Eckhoff et al, 2009; Aston-Jones and Cohen, 2005) causing over- or underarousal of PFC, although this relationship has to be confirmed in future studies. Medication might modulate the activity towards more adaptive levels (Eckhoff et al., 2009; Minzenberg et al., 2008; Kocsis et al., 2007).
ADHD may also be a disorder of dysfunctional adaptation (Aston-Jones & Cohen, 2005). It might be caused by too much arousal in familiar environments and too little in novel environments. Here, medication could modulate arousal via NE levels differently within one individual according to the situational demands (Kocsis et al., 2007; Aston-Jones & Cohen, 2005). In either of the scenarios, enhanced
performance during noradrenergic modulation is induced by neural gain.
A major drawback in these studies is that none of these directly attempted to measure LC activity in an ADHD population. This could be done with
high-resolution fMRI or using pupillometry. First of all, it would be interesting to observe task-related activation and activation during rest and compare these between ADHD patients and healthy individuals. This would yield information about LC function in
ADHD compared to normal LC function. Second, the same measures should be taken after administration of a selective NRI like atomoxetine or reboxetine to test the predicted modulatory effects in ADHD. Measuring PFC activity simultaneously can provide information about the relationship between LC and PFC activity.
Additionally, it might be interesting to look at differences between ADHD and healthy controls in neural gain, using fMRI measures proposed by Eldar et al. (2013) combined with pupillometry. This could provide information whether the phasic mode exhibits its beneficial effects on cognition via gain as predicted (Aston-Jones & Cohen, 2005; Shea-Brown et al., 2008).
Conclusions and Discussion
This review attempted to clarify the role of the LC NE modulatory system in decision-making and ADHD etiology. Computational models yield specific, testable predictions and can account for underlying mechanisms. Using pupillometry, new insights were gained concerning noradrenergic modulation of decision-making in humans. Pharmacological studies with NRIs provide evidence that the LC NE system is affected in ADHD, causing impulsive behavior and distraction as predicted by computational models (Eckhoff et al., 2009) and as a result impair decision-making.
Theories of NE modulation of cognition focus on the LC modes and their functions for cognition. Overall, both too high and too low levels of tonic NE are associated with impaired cognitive performance, while medium levels are thought to be optimal. The low tonic mode promotes phasic LC bursts which are associated with
high cognitive performance and the high tonic mode is associated with less focus, impulsivity and exploratory behavior. Improved cognitive performance is induced by neural gain.
Pupillometry has been proven to be useful in exploring the behavioral
correlates of the LC modulation. Pupil diameter reflects the tonic LC mode and pupil dilations reflect phasic activations. A study of pupillometry has shown that the LC is active during decision-making and predicts the content of a response in conservative subjects. Another study showed that tonic gain shifts attention towards idiosyncratic predispositions during learning. The same study demonstrated that gain is reflected in increased brain-wide functional connectivity and brain-wide clustering. These studies indicate that the LC influences decision-making on a higher cognitive level than previously thought.
Pharmacological studies indicate that tonic LC activity might be too high in the hyperactive subtype of ADHD or too low or high in the inattentive subtype. These indications are in line with predictions of computational models (Eckhoff et al., 2009). NRIs such as atomoxetine and reboxetine and stimulants like modafinil most likely treat ADHD symptoms by lowering tonic LC activity and promoting adaptive phasic responses, while simultaneously increasing NE levels in PFC. Phasic
responses might provide their beneficial effects on cognition by inducing neural gain (Shea-Brown et al., 2008).
Based on the evidence gathered so far, however, these results have to be regarded as preliminary, since direct evidence is still missing. The approach used here was
exclusively inferring what might be dysfunctional in ADHD based on the effects of medication. Future studies should also try to directly test above mentioned theories of ADHD etiology by comparing healthy individuals and untreated patients. The predictions which could be tested based on theoretical models are 1. Whether tonic LC activity deviates from a medium range 2. Whether hyperactive behavior
correlates with high tonic levels and unmotivated or inattentive behavior with low tonic levels 3. Whether exploratory behavior correlates with a high tonic levels and 4. Whether there is a deficiency in task-related phasic activations. Future research on the LC neuromodulatory system using medication should focus on NRIs because stimulants like modafinil also target the DAT and do not exclusively exhibit noradrenergic effects.
At the same time, focusing only on the noradrenergic system when investigating the etiology of ADHD might be a problem since it might not yield a comprehensive picture in the long run. For example, about 90% of children who are treated with medication effectively, are treated with stimulants targeting both the DAT and the NET (Franken et al., 2013). So research on ADHD etiology should investigate the role of DA in the disorder, too. This point is also true for the more general research on neuromodulators and psychiatry. There are indications that DA and NE or serotonin and NE act in concert, as both have been implicated in the same disorders like
ADHD, depression and anxiety (Stahl, 2008; Aston-Jones & Cohen, 2005) . A way of investigating differential effects may be to compare the effects of selective NRIs with
the effects of stimulants like modafinil. Effects they share might be noradrenergic and additional effects might be synergetic or dopaminergic.
The theories used here to make predictions about ADHD etiology all address the LC modes directly or indirectly. An alternative view on the problem can be derived from theories on etiology of depression, which posit that it is caused by a decreased number of serotonin receptors (Stahl, 2008). The same might be the case in the high or low deviant NE levels of the Eckhoff et al. model. The excessively high or low NE levels might be caused by an deviant low or high number of NE receptors in PFC, or by a deviant number of NET in LC neurons. This is however difficult to investigate since animal models of ADHD might not have the same etiology as humans with ADHD. Postmortem studies of ADHD subjects might be interesting in this context.
Finally, further complications for research on ADHD etiology arises from inherent problems in the diagnosis itself. Because it is diagnosed on the basis of symptoms like impulsivity and distractibility, ADHD seems like a disorder of
cognitive dysfunction of attention and inhibition and some of the research described here aims at investigating the dysfunction of these cognitive domains. However, research seems to have failed to identify any neuropsychological tests which could be used to reliably diagnose the disorder. This poses a challenge for the validity of impulsivity and inattention as diagnostic criteria. A reason for the inconsistency in findings might be that the disorder is fairly heterogeneous. Also, many patients’ diagnosis switches from one subtype to another over the years (Franken et al., 2013) and etiological theories should be able to explain those phenomena as well.
Regarding decision-making, it seems as if the currently established view on the role of the LC NE neuromodulatory system has to be refined based on the studies discussed here. That evidence can serve as a fundament for future research to
investigate the role of the LC in decisions on a higher cognitive level. First of all, replications are needed that confirm the LC' s involvement during decision
processes. Then, the relationship between LC modulation and individual bias during decision processes seems promising as two studies indicated an association. Finally, it would be interesting to see whether modulating LC actions, e.g. with medication, can change the content of decisions that people make.
Future research might investigate these questions by combining pupillometry, NRIs and decision-making paradigms. It would especially be interesting to look now at a more sophisticated type of decisions-making than that tested in simple 2AFC tasks or the oddball paradigm. For example, one could investigate the influence of media, which is often intended to alter decisions, on LC activity and the decisions that follow. Also, noradrenergic modulation during priming experiments might be interesting, as they also aim at altering decision-making. Furthermore, the Eldar et al. study demonstrated an influence of tonic gain on learning, so the effects of gain on decision-making might also be interesting to study, for example using the fMRI measures for gain used by Eldar et al.
Concerning ADHD, further research is needed to elucidate various issues concerning etiology and the effect of medication. First of all, as described earlier, LC activity should be compared between ADHD and control subjects to observe any
aberrant functions. This could be done by combining fMRI and pupillometry. Then it is important to understand the relationship between PFC and LC. What is the
direction of this modulatory relationship? Does the LC modulate the PFC or does the PFC also influence LC, as proposed by Aston-Jones (2005) and Walz et al. (2013)? Both options might be true, as there seem to be monosynaptic projections from PFC to the LC (Sara & Bouret, 2012). In this context, the hypothesis of a negative
relationship between these two structures (Minzenberg et al., 2008) can be investigated, e.g. by using high resolution fMRI to image both PFC and LC or pupillometry combined with fMRI during tasks that activate the PFC.
Finally, more controlled studies are needed to explore the effects of NRIs on LC in ADHD subjects. Studies so far support the view that medication induces a high phasic, low tonic mode, but this view has to be confirmed in pupil studies and
studies with an ADHD population.
The view on the LC neuromodulatory system has evolved over the decades from a simple arousal system to an important system in decision-making. This view will now further have to be refined as the LC seems to exhibit influence on decisions on a higher cognitive level than previously thought. Future research will have to clarify which aspects of decision-making are under its influence and which contents can be altered under noradrenergic modulation. Future research will also have to show how medication can alter this system relieving ADHD symptoms.
References
Astafiev, S. V., Snyder, A. Z., Shulman, g. L. & Corbetta, M. (2010). Comment on “Modafinil shifts human locus coeruleus to low tonic, high phasic activity during functional MRI” and “Homeostatic sleep pressure and responses to sustained attention in the suprachiasmatic area”. Science, 328,309.
Aston-Jones, G. & Cohen, J. D. (2005). An integrative theory of locus coeruleus-norepinephrine function: adaptive gain and optimal performance. Annual
Review Neuroscience 28, 403-450.
Chamberlain, S. R., del Campo, N., Dowson, J., Müller, U., Clark, L., Robbins, T. W. & Sahakin, B. J.(2007). Atomoxetine improved response inhibition in adults with attention deficit /hyperactivity disorder. Biol Psychiatry, 62, 977-984.
Chamberlain, S. R., Hampshire, A., Müller, U., Rubia, K., del Campo, N., Craig, K. et al. (2009). Atomoxetine modulates right inferior frontal activation during inhibitory control : a pharmacological functional magnetic resonance imaging study. Biol Psychiatry, 65, 550-555.
Clayton, E. C., Rajkowski J., Cohen, J. D. & Aston-Jones, G. (2004). Phasic Activation of Monkey Locus Ceruleus Neurons by Simple Decisions in a Forced-Choice Task. The Journal of Neuroscience, 44, 9914-9920.
de Gee, J. W., Knapen, T. & Donner, T. H. (2014). Decision-related pupil dilation reflects upcoming choice and individual bias. Proceedings of the National
Academy of Science USA,11, 5, E618-E625.
Eckhoff, P., Wong-Lin, K. F. & Holmes, P. (2009). Optimality and robustness of a biophysical decision-making model under norepinephrine modulation. The Journal of Neuroscience, 29(13), 4301-4311.
Eldar, E., Cohen, J. & Niv, Y. (2013). The effects of neural gain on attention and learning. Nature Neuroscience, (16), 1146-1153.
Frank, M. J., Santamaria, A., O’Reilly, R. C. & Willcutt, E. (2007). Testing
computational models of dopamine and noradrenalin dysfunction in attention deficit/hyperactivity disorder. Neuropsychopharmacology, 32, 1538-1599. Franken, I., Muris, P., & Denys, D. (2013). Basisboek Psychopathologie. Uitgeverij: de
Tijdstroom
Gazzaniga, M. S., Ivry, R. B., & Mangun, G. R. (2009). Cognitive neuroscience, the
biology of the mind (3rd ed.). New York: Norton
Gilzenrat, M. S., Nieuwenhuis, S., Jepma, M. & Cohen, J.D. (2010). Pupil diameter tracks changes in control state predicted by the adaptive gain theory of locus coeruleus function. Cognitive, Affective & Behavioral Neuroscience, 10, 252-269. Jepma, M. & Nieuwenhuis. S. (2011). Pupil Diameter Predicts Changes in the
Exploration–Exploitation Trade-off: Evidence for the Adaptive Gain Theory.
Journal of Cognitive Neuroscience, 23, 1587-1596.
Kocsis, B., Li, S. & Hajos, M. (2007). Behavior-Dependent Modulation of Hippocampal EEG Activity by the Selective Norepinephrine Reuptake Inhibitor Reboxetine in Rats. Hippocampus, 17, 627-633.
Llorente, A. M., Voigt, R. G., Jensen, C. L., Berretta, M. C., Fraley, J. K. & Heird, W. C. (2006). Performance on a visual sustained attention and discrimination task is associated with urinary excretion of norepinephrine metabolite in children with attention-deficit/hyperactivity disorder (AD/HD). The Clinical
Neuropsychologist, 20, 133-144.
Ma, C.-L., Arnsten, A. F. T. & Li, B.-M. (2005). Locomotor hyperactivity induced by blockade of prefrontal cortical α2-adrenoceptors in monkeys. Biological
Psychiatry, 57, 192-195.
Marzinzik, F., Wahl, M., Krüger, D., Gentschow, L., Colla, M., et al. (2012). Abnormal Distracter Processing in Adults with Attention-Deficit-Hyperactivity Disorder. PLoS ONE, 7(3), e33691.
Minzenberg, M. J., Watrous, A. J., Yoon, J. H., Ursu, S. & Carter, S. (2008). Modafinil shifts human locus coeruleus to low tonic, high phasic activity during
functional MRI. Science, 322, 1700-1702.
Murphy, P. R., O’Connell, R. G., O’Sullivan, M, Roberston, I. H., & Balsters, J. H. (2014). Pupil diameter covaries with BOLD activity in human locus coeruleus.
Human Brain Mapping, 0, 1-15.
Murphy, P. R., Robertson, I. H., Balsters, J. H. & O’Connell, R. G. (2011). Pupillometry and P3 index the locus coeruleus– noradrenergic arousal function in humans.
Psychophysiology, 48, 1531-1542.
Nieuwenhuis, S., Aston-Jones, G. & Cohen, J. D. (2005). Decision making, the P3 and the locus coeruleus-norepinephrine system. Psychological Bulletin, 313, 4, 510-532.
Robbins, T. W. & Arnsten, A. F. T. (2009). The neuropsychopharmacology of fronto-executive function: monoaminergic modulation. Annual Review of Neuroscience, 32, 267-287.
Stahl, S. M. (2008). Essential psychopharmacology (3rd ed.). Cambridge: Cambridge University Press.
Sara, S. (2009). The locus coeruleus and noradrenergic modulation of cognition.
Nature Reviews Neuroscience, 10, 211-223.
Bouret, S. & Sara, S. (2005). Network reset: a simplified overarching theory of locus coeruleus noradrenaline function. Trends in Neuroscience, 28,11, 574-582. Sara, S. & Bouret, S. (2012). Orienting and reorienting: the locus coeruleus mediates
attention through arousal. Neuron, 76, 130-141.
Shea-Brown, E., Gilzenrat, M. S. & Cohen, J. D. (2008). Optimization of Decision Making in Multilayer Networks the role of the locus coeruleus. Neural Computation, 20, 2863-2894.
Swann, A. C., Birnbaum, D., Jager, A. A., Dougherty, D. M., Moeller, F. G. (2005). Acute yohimbine increases laboratory-measured impulsivity in normal subjects. Biological Psychiatry, 57, 1209-1211.
Tamm, L., Menon, V. & Reiss, A.L. (2006). Parietal attentional system aberrations during target detection in adolescents with attention deficit hyperactivity disorder: event-related fMRI evidence. Am J Psychiatry, 163, 1033-1043.
Turner, D. C., Clark, L., Dowson, J., Robbins, T. W. & Sahakia, B. J. (2004). Modafinil improves cognition and response inhibition in adult
attention-deficit/hyperactivity disorder. Biol Psychiatry, 55, 1031-1040.
Walz, J. M., Goldman, R. I., Carapezza, M., Muraskin, J., Brown, T. R. & Sajda, P. (2013). Simultaneous EEG-fMRI Reveals Temporal Evolution of Coupling between Supramodal cortical Attention Networks and the Brainstem. The
Journal of Neuroscience, 49, 1912-1922.
