The role of dlPFC laterality in the expression and regulation of anxiety

Abstract

Anxiety disorders are the most common mental health disorder. Therefore, elucidating brain mechanisms implicated in anxiety disorders is important avenue for developing novel treatments and improving care. The dorsolateral prefrontal cortex (dlPFC) is thought to be critically involved in working memory processes (i.e. maintenance, manipulation, suppression, etc.). In addition, there is evidence that this region is involved in anxiety regulation. However, it is unclear how working memory related dlPFC processes contribute to anxiety regulation. Furthermore, we know that laterality plays an important role in working memory related dlPFC processing, however there is no current model of dlPFC mediated anxiety regulation that accounts for potential laterality effects. To address this gap, we propose a potential framework where the dlPFC contributes to emotion regulation via working memory processing. According to this framework, working memory is a fundamental process executed by the dlPFC. However, the domain of content differs across the left and right dlPFC, with the left dlPFC sensitive to primarily verbal content, and the right dlPFC sensitive to primarily non-verbal (affective content). Critically, working memory processes allow for both the retention and suppression of affective information in working memory and the overall net effect of processing on mood will depend on the balance of retention and suppression, the valence of the information being processed (positive vs. negative), and the domain of the information (verbal vs. non-verbal). If accurate, the proposed framework predicts that effects of neuromodulation targeting the dlPFC may be dependent upon the context during which the stimulation is presented.

Introduction

Anxiety disorders are the most common mental health disorder¹. The frontline treatments for anxiety disorders are cognitive behavior therapy and psychoactive medications (e.g., serotonin reuptake inhibitors; SRRIs). While these treatments are effective, the rate of treatment resistance is high and prevention efforts are lacking². Elucidating brain mechanisms implicated in anxiety disorders is an important avenue for developing novel treatments and improving the current standard of care³. There are a host of brain regions that have been implicated in anxiety (e.g., amygdala, dorsal prefrontal cortex, anterior cingulate cortex) and connectivity between these regions appears to be particularly important in anxiety⁴. The dorsolateral prefrontal cortex (dlPFC), a region of the lateral prefrontal cortex in the middle frontal gyrus, generally thought to overlap especially Brodmann areas 9 and 46, appears to be integrally involved in anxiety and is a likely target for treatment and prevention efforts. The function and organization of the dlPFC can be best understood with reference to its connections with other structures, as it is a highly connected node in the frontoparietal attention network with connections to temporal, parietal, and limbic regions. The dlPFC is critically important for working memory, executive control, and regulation of emotions, particularly anxiety.

While most models of anxiety treat the dlPFC as a unitary structure, there is evidence that the left and right dlPFC contribute uniquely to the expression and regulation of anxiety. Therefore, the purpose of this review is to propose a potential mechanistic framework describing anxiety regulation that includes distinct roles for the left and right dlPFC, based on the content of items in working memory. Importantly, the goal of this review is not to summarize emotion regulation and the neural architecture underlying emotion regulation, as these topics have been covered previously^5,6. The current review will 1) Summarize the literature demonstrating a role for the dlPFC in working memory and anxiety regulation 2) Summarize the dlPFC laterality literature, with respect to anxiety regulation 3) Describe our content domain framework of dlPFC function and 4) Highlight the testable predictions and treatment implications of this framework in anxiety research.

Methods

Although this is not a systematic review, we adopted a systematic approach to retrieve and sort articles for the current work. We used a custom python script (https://github.com/balders2/pubmed_search) that queried pubmed and fetched the publication date, pubmed ID, first author, last author, title, number of citations, and article pdf (if available) of the articles retuned from the query. We queried pubmed during early to mid 2022 using the following search terms: “dlPFC anxiety”, “dlPFC connectivity”, “dlPFC depression”, “dlPFC DMN”, “dlPFC GAD”, “dlPFC inhibition”, “dlPFC manipulation”, “dlPFC MDD”, “dlPFC n-back”, “dlPFC regulation”, “dlPFC spatial”, “dlPFC Sternberg”, “frontal alpha asymmetry”. These searches returned 4451 unique results. We then screened these articles and excluded those that were either uncited or off topic (i.e., included terms like: schizophrenia, parkinsons, anorexia, addiction, autism, adhd, stroke, dementia, etc.). From there we were left with 379 articles. These articles were then reviewed by the authors (NLB and LKW) and included in the review if they contributed to the pre-planned sections of the article. Although care was taken not to exclude articles that ran counter to our proposed framework, we did not systematically define a set of inclusion/exclusion criteria for these remaining articles. Accordingly, we believe this article does not meet the criteria of a systematic review and acknowledge that the topics contained within may be biased by the perspectives of the authors.

dlPFC and Working Memory

A classic body of literature supports the role of the dlPFC in cognitive control in goal-directed behavior, including a host of cognitive processes, such as directing attention, response selection, inhibitory control, working memory, and long-term memory^7–11. The dlPFC plays a critical role in the representation, updating, and maintenance of context information that can be used to guide behavior¹²; such context information can be task instructions, internal goal states, or the results of processing previous stimuli.

In imaging work by our group and others, we show the dlPFC plays a particularly important role in working memory^13,14. Working memory – the flexible online maintenance and manipulation of information in conscious awareness – can involve multiple processes like rehearsal, item shifting, spatial rotation, information suppression, and filtering depending on task demands. For the purposes of this review, we primarily distinguish between maintenance-like processes (those that keep specific items active in working memory) and suppression-like processes (those that keep specific items from becoming active in working memory). It is known that the dlPFC engages in both, and that maintenance and suppression processes can occur concurrently in the dlPFC.

Multiple brain imaging methods support the role of dlPFC is working memory maintenance. For instance, noninvasive brain stimulation such as transcranial direct current stimulation (tDCS) and transcranial magnetic stimulation (TMS) to the dlPFC is associated with improved working memory maintenance⁷. In lesion studies, patients with dlPFC damage showed specific impairment in manipulations and maintenance of information in working memory¹⁵, a finding also supported by imaging work¹⁶. Finally, connectivity between the dlPFC and dorsal anterior cingulate cortex/pre supplemental motor area (dACC/pSMA) were related to working memory demands¹⁷. Likewise, imaging work finds the dlPFC also plays a crucial role in suppressing distractors from working memory¹⁸, possibly through top-down attentional filtering or goal-oriented processes¹⁶. Working memory suppression processes may be mediated through connectivity between the dlPFC and the inferior frontal gyrus (IFG) and may be related to cognitive control demands¹⁷.

Interestingly, not all types of information are processed similarly by the dlPFC. More specifically, the left and right dlPFC seem to be specialized for distinct domains of content. The left dlPFC appears to be primarily sensitive to verbal content; the right dlPFC appears to be primarily sensitive to non-verbal content. For instance, verbal n-back tasks evoke more BOLD activity in the left dlPFC compared to the right. In contrast, spatial n-back tasks evoke more BOLD activity in the right dlPFC compared to the left¹⁹. Accordingly, although the overall processes engaged by the left and right dlPFC are similar, their functional profile is largely dependent upon the stimulus type.

Repetitive TMS (rTMS) targeting the dlPFC during WM tasks also shows laterality effects based on stimulus content. This was found in a study in healthy subjects who received 10 Hz rTMS during the retention interval of a 2-back WM task that required maintaining either verbal or spatial information about the task stimuli. rTMS to the right dlPFC interfered with performance in the verbal 2-back condition; rTMS to the left dlPFC interfered with performance in the spatial 2-back condition. The stimuli and task were identical across domain conditions, suggesting that excitatory rTMS interferes with WM performance in a domain-specific manner, biasing WM processes toward the dominant domain of the stimulation site, such that this interference depends on the laterality of the stimulation²⁰. In another example, patients with left dlPFC damage show impairments in a WM task that required subjects to name pictures based on semantic associations, suggesting an impairment in the maintenance of verbal information. In contrast, patients with right dlPFC damage showed impairments in a procedural executive control task that required matching colors and labels, but that did not require maintenance of semantic information. Critically, both tasks required the use of task rules across trials but differed in the domain of the rules to be maintained (semantic vs non-semantic). Thus, the results demonstrated a double dissociation in the domain of information maintenance across hemispheres, even when similar working memory processes were engaged²¹.

Together the work in this section suggests that the dlPFC plays a critical role in working memory and that the laterality of processing largely depends on the content of the information to be processed.

dlPFC and Emotion Regulation

Despite being characterized as a cognitive region of the brain, the dlPFC is known to play a crucial role in emotion regulation. Emotion regulation is broadly defined as the effortful or automatic strategies employed to reduce, induce, manipulate, or modify an emotion²². Functional imaging studies have found the dlPFC activates when emotional processing needs to be inhibited or integrated with ongoing cognitive processing^23–25. Lesion works shows that patients with dlPFC damage were not able to downregulate their subjective fear to a conditioned stimulus²⁶. Work with tDCS shows that stimulation to the dlPFC is associated with increased cognitive control during emotion regulation, facilitating subjects’ ability to reduce skin conductance responses to affective pictures²⁷. The dlPFC is largely thought to regulate emotional processing in a top-down manner through its inhibitory connections to other brain regions, such as the anterior cingulate cortex, dorsomedial prefrontal cortex (dmPFC), and amygdala²⁸. The negative coupling between the dlPFC and these other regions has been well replicated^29,30, with a particularly robust negative correlation emerging between the dlPFC and amygdala^31–33.

Support for dlPFC involvement in emotion regulation also comes from observed dlPFC activation on executive control tasks that use emotion stimuli³⁴ and from disruption to emotion regulation when cognitive control and working memory demands are increased³⁵. Such top-down control is indexed by increased dlPFC recruitment when emotion-related distractors need to be inhibited (i.e., suppressed)³⁶, where the dlPFC is thought to down-regulate activity in the amygdala and other limbic regions^37,38. Interestingly, anxiety has been related to both hypo-³⁹ and hyper-activation of the dlPFC during the suppression of threat distractors⁴⁰. We have shown that increasing task difficulty during periods of unpredictable threat activates the right dlPFC more than during neutral conditions and this activation was positively correlated with task performance⁴¹. dlPFC activity is also negatively correlated with subgenual Anterior Cingulate Cortex (sgACC) activity, another emotion regulation hub of the brain, during the suppression of threat distracrtors⁴². dlPFC activity is also negatively correlated with parietal activity evoked by task-irrelevant affective face stimuli in a face-stroop task⁴³. dlPFC activation is also observed during neutral conditions of cognitive reappraisal tasks^44,45, suggesting the dlPFC may underlie more general cognitive demands of the emotion regulation process, irrespective of emotional content.

Our work using the No-shock, Predictable Shock, Unpredictable shock (i.e., NPU) task, a paradigm designed to study anxiety in humans, has shown that the right dlPFC is activated by cues presented during periods of unpredictable threat, and this increased dlPFC activation is associated with reduced anxiety⁴⁶. In other work with the NPU task, we have shown that increased activation of the dlPFC is associated with a smaller anxiety potentiated startle response^47,48. Importantly, reduction in startle potentiation is directly related to the cognitive load of the task^49,50. These results suggest that cognitive and affective information interacts in the dlPFC, possibly competing for limited resources in working memory.

Neuroimaging work showing dlPFC abnormalities during cognitive and emotional tasks also supports dlPFC-mediated emotion regulation. Greater dlPFC activity during an emotional task condition compared to a neutral condition was associated with PTSD symptoms in trauma exposed individuals with subthreshold PTSD⁵¹. Greater dlPFC activity and less amygdala activity prior to treatment is also associated with better treatment outcomes in PTSD patients⁵², suggesting better top-down regulation mechanisms may assist in treatment response. Emotion-regulation dlPFC related findings have been mixed in studies using anxious populations. For instance, GAD patients showed reduced dlPFC activation in a working memory task when the task required suppression of an emotional distractor⁵³. In contrast, GAD patients show increased dlPFC activity during the presentation of negative picture compared to SAD and control subjects, suggesting the GAD patients are possibly attempting to implement more emotion regulation while viewing negative images⁵⁴.

Together the work in this section suggests that the dlPFC plays a critical role in general emotion regulation and the regulation of anxious states; however, more research is needed to fully understand the role of dlPFC in anxious pathology.

dlPFC and Effort

Despite clear evidence that the dlPFC contributes to emotion regulation processes, it is unclear how emotion regulation fits in with the broader profile of dlPFC working memory function, especially in the context of anxiety. One possibility is that the dlPFC contributes to emotion regulation by supporting active effortful cognitive strategies. In this case, it may be argued that emotion regulation is mediated by maintenance processes of working memory, those processes required to keep certain information maintained in working memory. For instance, working memory dlPFC-related dysfunction, even under neutral conditions, is common in individuals with mood and anxiety disorders^55,56 Thus, for anxious individuals, perturbations in ability to maintain information in working memory may contribute to their dysregulated anxiety. We found hyperactivation in dlPFC during a neutral working memory task was related to anxiety^14,57, a finding replicated in other work⁵⁸. However, several studies have found anxiety is related to dlPFC hypoactivation during working memory tasks^59,60; GAD patients show reduced loadrelated dlPFC activation during the N-back WM task⁶¹. Behaviorally, we have also shown that better working memory performance was related to greater anxiety-related symptoms⁶²; however, anxiety-related WM deficits are also reported⁴⁹. Further work is needed to understand these contradictory dlPFC dysfunction findings in the context of working memory maintenance and anxiety; dlPFC laterality, different cognitive demands, or task parameters may underlie these differences.

On the other hand, it might be that the dlPFC contributes to emotion regulation by supporting more automatic emotion processes, like the down regulation of activation in emotional processing regions. In this case, it may be argued that emotion regulation is mediated by a suppression-like working memory process, those processes that inhibit certain information from entering working memory space; the dlPFC works to filter out or suppress negative emotional responses. For instance, the level of functional connectivity between the dlPFC and amygdala are related to outcomes following exposure therapy⁶³. In addition, mindfulness training, thought to strengthen top-down control of emotion, increases connectivity between the dlPFC and brain regions⁶⁴. Finally, neurofeedback aimed at increasing dlPFC activity has been found to reduce anxiety in phobic patients⁶⁵. On the other hand, stress induction can lead to a decoupling of the amygdala and dlPFC, which affects one’s ability to filter out (i.e., suppress) emotional distractors during ongoing cognitive tasks⁶⁶. Additionally, trauma exposure, independent of PTSD diagnosis, can lead to decreases in top-down connectivity between the dlPFC and the amygdala⁶⁷. Thus, poor connectivity between dlPFC and other emotion hubs, can lead to the inability to suppress or inhibit negative emotions or emotional information, thereby increasing or maintaining anxiety.

Together the work in this section suggests that the dlPFC potentially contributes to emotion regulation through both effortful and automatic processes, relying on multiple working memory processes.

dlPFC and Laterality

In addition to ambiguity related to the working memory process engaged during the regulation of anxiety, there is also ambiguity related to how the unique roles of the left and right dlPFC influence emotion regulation. Specifically, it is unclear whether the left and right dlPFC contribute to emotion regulation in similar or distinct ways. Indeed, there is mounting evidence that suggests that left and right dlPFC contribute to emotion regulation through distinct processes. For instance, early EEG research has historically offered support for a lateralized view of the dlPFC⁶⁸. Although more recent work has questioned the clinical utility of frontal alpha as a biomarker for mood disorders^69,70, there is evidence to suggest that mood states differentially affect alpha power in the left and right hemisphere. Specifically, lower left alpha power relative to right alpha power was associated with positive affect or approach (positive emotional reactivity); higher left alpha power relative to right was associated with negative affect or avoidance (i.e., negative emotional reactivity)⁷¹. Given that alpha power is generally thought to reflect cortical inhibition⁷², it is thought that greater activity or excitability in the left dlPFC (i.e., left lateralized frontal alpha asymmetry [FAA]) is associated with positive mood and approach behaviors while greater activity or excitability in the right dlPFC (i.e., right lateralized FAA) is associated with a more negative mood and avoidance or anxiety-related behaviors⁷³. Left-lateralized FAA is correlated with Behavioral Approach System (BAS)⁷⁴ scores as well as reward bias in an incentivized verbal WM paradigm⁷⁵. Individuals who have left-lateralized FAA also exhibit a positive affective bias when evaluating affectively neutral stimulus relationships⁷⁶. Similarly, FAA is left-lateralized during presentations of positive facial expressions, while right-lateralized during presentations of negative expressions⁷⁷. Several studies have shown increased right-lateralized FAA in depression⁷⁸, anxiety⁷⁹, and PTSD patients⁸⁰. PTSD patients also show increased right-lateralized parietal alpha asymmetry^81,82. More recent imaging work beyond FFA^83–85 has shown that depressed patients have lower activity in the left prefrontal cortex and relatively higher activity in the right prefrontal cortex, suggesting an imbalance between left and right prefrontal activity may be associated with the pathophysiology of depression. It remains unclear if similar findings hold true in anxiety.

The lateralization FAA work formed much of the basis for the emergence of lateralized dlPFC neuromodulation protocols for the treatment of depression and anxiety. One such technique, TMS, uses either repetitive (rTMS) or patterned (theta burst stimulation; [TBS]) stimulation to modify excitability in localized cortical circuits. Work in motor cortex suggest that some frequencies (> 5 Hz) and patterns (intermittent TBS) of stimulation are often found to increase excitability, while other frequencies (< 5 Hz) and patterns (continuous TBS) reduce excitability^86–92. Researchers designed clinical trials aimed at using rTMS to alter excitability levels in the dlPFC, most commonly with high frequency (excitatory; 10 Hz) stimulation to the left dlPFC and low frequency (inhibitory; 1 Hz) stimulation to the right dlPFC⁹³. A network meta-analysis (81 clinical trials, N = 4,233 patients) compared the effects of eight rTMS interventions for the treatment of depression and anxiety, including the high frequency left and low frequency right with sham stimulation. The study found that compared with sham, both high frequency left and low frequency right rTMS were significantly more effective with respect to remission of depression symptoms⁹⁴. Some studies have combined high frequency left and low frequency right in a sequential bilateral rTMS treatment and, while this was shown to have clear superiority over sham,⁹⁴ bilateral treatment was not superior to high frequency left treatment. There is also evidence that these asymmetric rTMS treatment protocols may be effective for comorbid anxiety symptoms as well. For instance, left HF-rTMS⁹⁵, right LF-rTMS⁹⁶, and sequential bilateral rTMS⁹⁷ can reduce anxiety symptom in depressed individuals. However, several meta-analyses have failed to find efficacy in using rTMS to treat primary GAD patients^98–101. Thus, further work is needed to understand the role of dlPFC laterality in anxiety and the current stimulation approaches need to be optimized for anxiety.

Together the work in this section suggests lateralization of dlPFC function is critical for understanding how the dlPFC contributes to emotion regulation, specifically anxiety.

Proposed Framework

Although there is evidence to suggest left and right dlPFC support distinct functions in emotion regulation, this laterality is not often addressed in anxiety research. To date, there is not a basic mechanistic framework of dlPFC function that accounts for such laterality as it relates to anxiety regulation. We believe that having such a mechanistic framework will help bridge seemingly disparate findings in the literature, elucidate the pathways leading to anxiety, and inform treatment protocols. Here we put forth a framework based on three guiding principles (See Figure 1; See also Table 1 for a list of papers supporting each of these claims). First, the dlPFC’s role in the executive function network^5,6, specifically the role of working memory in maintaining and manipulating (e.g. shifting, sorting, suppressing) information^6,102, is critical to emotion regulation (Table 1: emotion regulation). Second, although both the left and right dlPFC are engaged in common working memory processes, the different hemispheres mostly contribute to these processes within distinct content domains (i.e., verbal vs non-verbal; Table 1: content domains). Third, verbal and non-verbal emotion-related content is subject to these working memory processes and the degree to which the right and left dlPFC contribute to emotion expression and regulation is dependent upon the working memory process currently engaged (i.e. maintaining, shifting, sorting, suppressing; Table 1: maintenance vs. suppression).

Figure 1. Content domain framework.

According to the content domain framework, the dlPFC is primarily engaged in working memory related processes. However, the domain of information being processed differs across hemispheres. We posit that the primary domain of information processed by the left dlPFC is verbal, while the primary domain of information processed by the right dlPFC is non-verbal, with affect-related content making up a key sub-domain. Accordingly, both the left and right dlPFC can contribute to emotion regulation, depending on which working memory process is engaged, but their contributions will differ based on the domain of information being processed.

Table 1.

Papers related to the development of the proposed framework.

Citation	Reference Number	Principle	Primary Method	Main Finding
Nagel et al., 2013	19	content domains	fMRI	Left lateralized activity for verbal WM, right lateralized activity for spatial WM.
Sandrini et al., 2008	20	content domains	rTMS	Excitatory rTMS to right dlPFC slows verbal WM performance, while excitatory rTMS to the left dlPFC slows spatial.
Davidson et al., 1990	68	content domains	EEG	Verbal and spatial tasks differentially affect left and right EEG alpha power.
Sutton & Davidson, 2000	76	content domains	EEG	Left frontal alpha activity is associated with processing of verbal information in an affective task.
Metzger et al., 2004	82	content domains	EEG	Right parietal excitability is associated with severity of PTSD arousal symptoms.
O’Reardon et al., 2007	95	content domains	TMS	High frequency rTMS to the left dlPFC decreases depression symptoms.
Mantovani et al., 2013	96	content domains	TMS	Low frequency rTMS to the right dlPFC improves panic symptoms.
White & Tavakoli, 2015	97	content domains	TMS	Low frequency rTMS to the right dlPFC improves anxiety symptoms in comorbid GAD/MDD patients.
Gray et al., 2002	23	emotion regulation	fMRI	dlPFC is activated by emotional content when task demands are high.
Liberzon et al., 2000	24	emotion regulation	PET	Emotion regulation during affective picture viewing increases metabolic activity in the dlPFC.
Beauregard et al., 2001	25	emotion regulation	fMRI	Suppression of emotional responses during erotic picture viewing is associated with right dlPFC activity.
Kroes et al., 2019	26	emotion regulation	lesion	Patients with dlPFC are unable to use active strategies to down regulate arousal responses during fear conditioning.
Feeser et al., 2014	27	emotion regulation	tDCS	Anodal tDCS over right dlPFC decreases affective responding when subjects are instructed to down-reglate their emotions to affective images.
Simpson et al., 2000	30	emotion regulation	fMRI	Increased activation in the dlPFC and limbic regions were detected in response to negative vs neutral images
Hartling et al., 2021	34	emotion regulation	fMRI	Increased dlPFC activation was detected during an N-back task using emotional/affective stimuli
Adamczyk et al., 2022	35	emotion regulation	EEG/ERP	Increased cognitive load was associated with a decreased emotion regulation-realted ERP component
Ironside et al., 2018	37	emotion regulation	tDCS + fMRI	Stimulation of the dlPFC decreased amygdala response to threats and increased activation in cortical regions associated with attentional control during a threat attention task
Loos et al., 2020	38	emotion regulation	fMRI	Anxiety Patients showed decreased amygdala response to neutral and threat stimuli when cognitive load; this was mediated by cognitive load related increases in the dlPFC
Bishop et al., 2004	39	emotion regulation	fMRI	High anxious participants showed larger decreased left dlPFC on a threat-attention task when threat distractors were frequent
Fu et al., 2017	40	emotion regulation	fMRI	Behaviorally inhibited individuals (anxiety risk) showed increased right dlPFC during an attention-threat task
Balderston, Hsiung, et al., 2017	41	emotion regulation	fMRI	Increased dlPFC activation was seen under threat conditions compared to neutral conditions during a memory task
Mitchell et al., 2007	42	emotion regulation	fMRI	Negative coupling between dlPFC and ventromedial PFC/subgenual Anterior Cingulate Cortex (sgACC) was detected on tasks requiring the suppression of emotional stimuli
Banich et al., 2019	43	emotion regulation	fMRI	Increased dlPFC activity was associated with less processing of task irrelevant emotional information
Morawetz et al., 2020	44	emotion regulation	fMRI	Increased dlPFC activity and dlPFC-amygdala connectivity was associated with reappraisal; level of connectivity was associated with the success of reappraisal
Golkar et al., 2012	45	emotion regulation	fMRI	dlPFC activity was associated with reappraisal of both negative and neutral stimuli
Balderston, Liu, et al., 2017	46	emotion regulation	fMRI	During an anxiety manipulation, dlPFC activity was associated with increased subjective and objective anxiety responses
Vytal et al., 2013	49	emotion regulation	behvaior	High WM load was associated with decreased effects of an anxiety manipulation compared to medium and low WM loads on a verbal WM task; decreased anxiety disruptions on WM were seen across all WM loads on a spatial WM task
Vytal et al., 2012	50	emotion regulation	behavior	High WM load was associated with decreased effects of an anxiety manipulation compared to low WM load; increased WM load was also associated with decreased anxiety
Fonzo et al., 2017	52	emotion regulation	fMRI/TMS	In PTSD patients, increased dlPFC activation during an emotion task was associated with greater treatment response. TMS to right dlPFC was associated with decreased amygdala activity on the task
Moon & Jeong, 2015	53	emotion regulation	fMR	Anxiety patients showed lower dlPFC activity in the presence of anxiety-related distractors during a WM task
Buff et al., 2016	54	emotion regulation	fMRI	Anxiety patients showed increased dlPFC-amygdala connectivity when viewing threat vs neutral images
Balderston et al., 2016	61	emotion regulation	fMRI	Anxiety patients showed decreased dlPFC activity as a function of cognitive load across threat and neutral conditions on a spatial WM task
Balderston et al., 2020	14	retention vs. suppression	fMRI	Anxiety patients showed decreased dlPFC during an anxiety manipulation during a letter WM task; patients did not show hemispheric differences, but control participants showed decreased activity in right vs left dlPFC
Vasic et al., 2009	55	retention vs. suppression	fMRI	Depressed patients showed increased left dlPFC connectivity during a letter WM task
Fernández-Corcuera et al., 2013	56	retention vs. suppression	fMRI	Bipolar patients showed decreased bilateral dlPFC during a letter WM task
Basten et al., 2012	57	retention vs. suppression	fMRI	Trait anxiety was associated with increased right dlPFC and stronger right dlPFC connectivity during the manipulation phase of a letter working memory task compared to maintenance phase
Fales et al., 2008	58	retention vs. suppression	fMRI	High anxious participants showed increased transient WM-related bilateral dlPFC activation but reduced sustained right dlPFC activation during a verbal WM task
Moon et al., 2016	59	retention vs. suppression	fMRI	Anxiety patients showed decreased right dlPFC activation in response to emotional distractors on a face WM task
Moon & Jeong, 2017	60	retention vs. suppression	fMRI	Anxiety patients showed decreased left dlPFC to emotional distractors during a face WM task
Balderston et al., 2016	61	retention vs. suppression	fMRI	Anxiety patients showed decreased right dlPFC activation as a function of cognitive load during a spatial WM task across threat and neutral conditions
Luo et al., 2018	66	retention vs. suppression	fMRI	During a stress induction, decreased left dlPFC-amygdala connectivity was detected during encoding phase on a letter search task, this was especially true under conditions of low cognitive load and emotion distractors
Chen et al., 2018	67	retention vs. suppression	fMRI	PTSD patients and trauma exposed controls showed decreased dlPFC to amygdala connecting during resting state activity

Discussion

We believe that the perspective encompassed by the proposed dlPFC processing framework fits into a larger trend of recognizing commonalities between cognition and emotion. Indeed, there is ample information to suggest that cognition and emotion mutually interact and affect one another¹⁰³. According to the framework presented here, there should be little distinction between affective and non-affective information in the dlPFC as both are subject to the generalized processes of working memory carried out by this region. Critically, these working memory/executive processes allow for both the maintenance and suppression of affective information in working memory^102,104, which has implications for understanding emotion regulation strategies. For instance, maintenance-like processes engaged by the left dlPFC may regulate emotion by supporting the active, effortful engagement of verbal emotion regulation strategies, like reappraisal. In contrast, suppression-like processes engaged by the right dlPFC may regulate emotion by supporting the automatic top-down filtering of negative affective information, resulting in reduced activity in emotional expression regions. However, in real-life situations, the dlPFC is likely constantly engaged in some combination of maintenance and suppression processes across a wide variety of content types. Additionally, with the limited resources of working memory, the active maintenance of one type of information will likely lead to the suppression of other types of information, resulting in competition for resources¹⁰⁵. Accordingly, as discussed above, active TMS to the left dlPFC during spatial WM memory may bias the system toward the maintenance of verbal information and as a result lead to the suppression of non-verbal information, resulting in a site-specific decrease in performance²⁰. According to the proposed framework, the overall net effect of this WM-related dlPFC processing on mood will depend on the balance of maintenance and suppression, the valence of the information being processed (positive vs. negative), and the domain of the information (verbal vs. non-verbal).

Model Predictions

The proposed framework predicts that the dlPFC can both up and down-regulate negative emotional information depending on context. If this is the case, it should be possible to show increases in dlPFC (specifically right hemisphere) activity are associated with the active maintenance of positive mood as well as the active suppression of negative mood and these directly influence anxiety levels. Such a result would be inconsistent with the frontal asymmetry model mentioned above^68,71.These results have implications for how neuromodulation may affect mood and, importantly, anxiety.

The standard clinical practice for treating depression and anxiety symptoms with TMS is to apply excitatory TMS to the left dlPFC and inhibitory TMS to the right dlPFC^93,94. In the proposed framework, the left dlPFC primarily operates on verbal content domains, while the right dlPFC primarily operates on non-verbal content domains. Accordingly, high frequency (excitatory) stimulation of the left dlPFC should facilitate effortful, verbal cognitive emotion regulation strategies. In contrast, this framework suggests low frequency (inhibitory) rTMS to the right dlPFC would affect anxiety primarily by weakening the maintenance processes that stabilize negative affect, leading to a net decrease in negative affect (e.g., decreased anxiety). However, TMS stimulation can have a state-dependent effect, impairing or facilitating specific functions that are active at the time of stimulation. (i.e., the underlying activity/connectivity patterns corresponding to the current mood state). Thus, in healthy subjects who typically have a net positive affect, the proposed framework predicts opposing effects on mood; low frequency (inhibitory) rTMS to the right dlPFC might weaken the maintenance of positive affect, leading to a net decrease in positive affect. More interestingly, the proposed framework predicts that manipulation of mood at the time of stimulation may be leveraged to modulate the lasting effects of the stimulation. Take these examples. First, positive mood induction could be paired with excitatory stimulation to the right dlPFC to strengthen the retention processes important for maintaining positive affective tone. Alternatively, negative mood induction could be paired with inhibitory stimulation of the right dlPFC to weaken the retention processes important for maintaining negative affective tone. In both cases, the result would be a net positive effect on mood.

Indeed, we have seen similar effects when comparing the effects of within-session vs. between session stimulation of the right dlPFC on anxiety during threat of unpredictable shock. When we use 10 Hz (presumably excitatory) stimulation to modulate right dlPFC activity within-session, we get increases in anxiety potentiated startle (APS). This suggests a potential strengthening of the anxiety maintenance circuits activated during the anxiety manipulation¹⁰⁶. Accordingly, we predicted that an inhibitory stimulation pattern to the right dlPFC would lead to a reduction in APS. However, when we use cTBS (presumably inhibitory) stimulation to modulate right dlPFC activity between sessions (i.e., not during an anxiety manipulation) this inhibitory stimulation also leads to an increase in anxiety potentiated startle¹⁰⁷. According to the proposed framework, this increase in startle, an anxiety-related response, following offline cTBS may be due to a weakening of the circuits underling the maintenance of a non-anxious affective tone.

Importantly, the proposed framework and associated evidence suggests a path forward for treatment targets for anxiety. Neuromodulation has to date been less successful for treatment of anxiety disorders than it has been for depression^98–101. The current approach highlights the nuances involved with cognition-emotion interactions across the two hemispheres; understanding these nuances is especially important for informing both mechanistic and treatment studies. As highlighted in the introduction, however, important differences may emerge across different disorders and across different individuals, requiring a precision medicine approach.

Limitations

Although emerging evidence is consistent with the proposed framework, it should be noted that this explanation for our conflicting findings is post hoc and should be replicated. This is particularly true for the neuroimaging data presented above. While hemisphere-specific neuroimaging findings support the proposed framework, imaging studies often lack the power to determine whether a lack of dlPFC responding in one hemisphere represents a true null result. It is therefore critical that causal methods like TMS continue to be employed to understand how the left and right dlPFC contribute to emotion regulation. Care must also be taken when interpreting the results of TMS studies as well. For instance, although there is evidence in motor cortex for the excitatory/inhibitory rTMS patterns described above^86–92, we understand that it is premature to generalize these patterns to prefrontal cortex. It will be critical to expand these findings to other patterns of stimulation like low-frequency stimulation and iTBS, delivered during both online and offline sessions. Finally, future work should also test whether explicit positive mood induction moderates the effect of right dlPFC rTMS on anxiety. Indeed, contextual modulation of rTMS delivery is an active area of research aimed at improving the precision of these neuromodulatory techniques¹⁰⁸.

Conclusions

Here we summarize literature showing that the dlPFC contributes to both working memory processes and emotion regulation. Our primary interest is understanding how these two processes relate in the context of anxiety. Work showing that cognitive and affective processes interact, suggest that they are part of the same underlying process in the dlPFC. However, work showing that the left and right dlPFC can have opposing effects on affective tone suggest that affect can be modulated up or down, independent of ongoing cognitive processes. Until now, there was not a framework to explain these two discrepant findings. By reframing emotion regulation in terms of the type of working memory process employed (maintenance vs suppression), the proposed framework can account the discrepant dlPFC findings. Additionally, the proposed framework offers a mechanistic understanding for how dlPFC neuromodulation affects mood and anxiety and has implications for how to improve and personalize neuromodulatory treatments.

Highlights.

• We propose a model relating dlPFC-mediated emotion regulation to working memory.

• We argue that working memory is the fundamental process executed by the dlPFC.

• We argue that only the content domain differs across the left and right dlPFC.

• Left dlPFC can maintain or suppress verbal content in working memory.

• Right dlPFC can maintain or suppress non-verbal content in working memory.