Deficits in explicit emotion regulation in bipolar disorder: a systematic review

The influential classification of Gross (1998a) distinguishes two categories, depending on whether the regulation of emotions takes place in anticipation of, or response to, emotions.

Antecedent-focused regulation strategies involve selecting and modifying the situation in which we experience emotions and modify emotions via cognition (Gross 1998a, 2013; Phillips et al. 2008). Cognitive processes can be further differentiated into subprocesses on a continuum between pure attentional processes and processes of cognitive change (Ochsner and Gross, 2005). An example of an attentional strategy is distraction, which describes the detachment of attention from a stimulus that evokes emotions to other, non-emotional aspects of the situation, away from the situation entirely, or non-emotional thoughts (Gross, 2013). In contrast to simple attentional processes, cognitive change includes higher cognitive abilities that lead to an alternative evaluation of a situation that evokes emotions. Cognitive change can either be caused by concentrating on one’s coping strategies or an alternative interpretation of the situation (i.e., reappraisal). Reappraisal can be further distinguished in self-focused and situation-focused reappraisal (Ochsner et al. 2004). While self-focused reappraisal describes the modification of the degree of personal involvement in an emotionally loaded situation, situation-focused reappraisal describes the reinterpretation of the situation itself.

Response-focused strategies include the suppression or alteration of behavioral, experiential, and physiological responses that accompany emotions. They further include processes that enable us to prolong or maintain the experience of emotions (Gross 1998b; Tugade and Fredrickson 2007). To accomplish emotion maintenance, an individual has to express the behavioral response associated with an emotion (Izard 1990), for example, yelling in case of anger. Further, an individual can maintain an emotion in a working-memory system specialized for emotions (Mikels et al. 2008). Notably, the listed ER strategies are not valence specific; both negative and positive emotions can be regulated.

Neurally, emotion regulation is accomplished through the interaction of two systems. (Ochsner and Gross 2014). One system is responsible for the experience and the contextual valuation of emotions, consisting of the amygdala, insula, ventral striatum, and medial areas of the prefrontal cortex (PFC). A control system acts on this system, consisting of the dorsal anterior cingulate cortex (dACC), dorsal posterior medial PFC, dorsolateral PFC (DLPFC), inferior parietal cortex, and ventrolateral prefrontal cortex (VLPFC). In part, specific cognitive processes are associated with specific areas within this network. For example, the DLPFC, the dorsal posterior mPFC, and the inferior parietal cortex are associated with the control of attention to relevant stimuli and the maintenance of goals during emotion regulation. The ventrolateral PFC is associated with inhibiting inappropriate behaviors, and the dACC is associated with monitoring conflicts between desirable and actual actions. For the involvement of most of the mentioned areas, meta-analytical evidence has been accumulated (Kohn et al. 2014). While the model of Ochsner and Gross (2014) is characterized by a high level of detail, including the attribution of cognitive processes to specific areas, it does not distinguish between explicit and implicit ER, concepts whose importance to ER more recently has been demonstrated and which are the focus of the next section.

In recent research, ER strategies have been classified according to how voluntarily they occur (Braunstein et al. 2017; Gross, 2013; Gyurak et al. 2011; Phillips et al. 2008). While implicit ER runs spontaneously, automatically, and unconsciously, explicit ER operates planned, consciously, and volitionally. A further distinction can be made between whether the goal of the ER is explicit or implicit and how automatic or controlled the ER process itself is (Braunstein et al. 2017). In the present review, we adhere to a strict definition of explicit ER in which both the goal of ER is explicit (which is ensured in experimental setups by instruction), and the required processes are controlled.

Explicit and implicit ER are further distinguishable by their demands on different neural networks. The neural model by Phillips et al. (2008) postulates a hierarchical organization of implicit and explicit ER (Fig. 1). On the lowest level, areas involved in emotion reactivity (i.e., ventral striatum, thalamus, and amygdala) are embedded in a network, together with areas involved in automatic ER (i.e., rostral anterior cingular cortex (rACC), subgenual anterior cingular cortex, orbitofrontal cortex (OFC), and a hippocampus-parahippocampus region). On the highest level, explicit ER is primarily realized by prefrontal areas (dorsolateral prefrontal cortex (DLPFC) and the ventrolateral prefrontal cortex (VLPFC)) but mediated by areas associated with implicit ER (such as the OFC) that, in turn, have direct connections to subcortical areas associated with emotion reactivity. According to the model, the dorsomedial prefrontal cortex (DMPFC) and dorsal anterior cingulate cortex fulfill functions in explicit and implicit ER.

In addition to the consideration of explicit and implicit ER, the model of Phillips et al. (2008) has the advantage that it includes findings on the ER of patients with bipolar disorders. Based on previous results (Elliott et al. 2004; Wessa et al. 2007), the researchers concluded that inefficient use of the DLPFC and VLPFC results in increased activity in these regions and is, thus, a probable cause for deficit regarding explicit ER in patients with BD. However, due to the model's hierarchical structure, aberrations in areas associated with implicit ER or emotion reactivity could also impair explicit ER. For this reason, an alternative explanation is that hyperactivity in DLPFC and VLPFC may represent a compensatory mechanism for aberrations in areas associated with implicit ER and emotion reactivity.

Taken together, the model of Phillips et al. (2008) provides essential insights into the neural basis of ER deficits in patients with BD. However, the model is based on only a few studies, especially for explicit ER, which leads to limitations since the results are derived from homogeneous samples and low diversity tasks.

Patients who have BD show aberrations not only at the neural level but also in the experience of emotions and associated behavior. Studies sampling experiences in everyday life found an increased frequency of attempts to regulate emotions in BD patients (Gruber et al. 2013b). However, these spontaneous emotion-regulation attempts are less successful compared to healthy subjects (Gruber et al. 2012), which is partly due to the more frequent use of maladaptive strategies (e.g., rumination) by patients with BD (Becerra et al. 2013; Dodd et al. 2019; Green et al. 2011; Gruber et al. 2012; Gruber et al. 2013b). Interestingly, in laboratory settings, differences between BD patients and healthy controls (HC) are less pronounced (Dodd et al. 2019). A possible explanation for this phenomenon could be that participants in laboratory settings are often explicitly instructed to use ER; that is, they initiate ER voluntarily, which could positively affect their success. This positive effect makes explicit ER strategies an exciting topic for research as explicit ER strategies might be a resource to build on in psychotherapeutic interventions.

Ladouceur et al. (2013) used an emotional n-back task to investigate alterations of voluntary, attentional ER and its neural correlates. In this task’s 0-back condition, the attention demand is low (a simple keystroke is requested for a specific stimulus). In the 2-back condition, the demand is high (a keystroke is required when a stimulus is the same as in the second last trial). In some of the task’s trials, task-irrelevant pictures, either emotional or neutral faces, are presented. Participants show the ER by controlling their attention to focus on the n-back task and ignoring the emotional distractors. Therefore, a heightened error rate and an extended response time (RT) in the n-back task during the presence of emotional distractors indicate failures in ER.

Interestingly, the authors chose to investigate healthy offspring of patients with BD compared to healthy low-risk controls (i.e., offspring of parents, neither of whom had been diagnosed with an axis 1 disorder). The n-back task performance in healthy offspring of BD patients and low-risk controls did not differ significantly. However, in the high demanding 2-back condition, both groups differed in neural activity evoked by emotional distractors. BD patients’ healthy offspring showed hyperactivity in the VLPFC in the presence of happy but not fearful distractors. In the presence of both types of emotional distractors, the healthy offspring of BD patients showed a reduced modulation of the amygdala by the VLPFC. Additionally, in the presence of happy distractors, healthy offspring of BD patients showed a reduced modulation of the DLPFC by the VLPFC.

Heissler et al. (2014) examined individuals with an increased risk for BD (indicated by high scores in the Hypomanic Personality Scale (HPS; (Eckblad and Chapman 1986) and compared their performance in a picture viewing task (positive, negative, and neutral stimuli) with a low-risk group. In the picture viewing task, participants look at pictures with emotional content (Heissler et al. 2014; Kanske et al. 2013, 2015). They receive either the instruction to simply view the pictures or to regulate their emotions. In this particular version of the task, participants could distract themselves from emotions by solving an arithmetical task presented in addition to the pictures.

During the processing of emotions in the passive viewing condition, the risk group exhibited increased amygdala activity in response to negative stimuli compared to the control group. Behaviorally, the risk group showed no significant impairments in ER. However, participants with increased risk for BD exhibited enhanced amygdala activity in the passive viewing condition and increased activity in the left inferior parietal cortex in the distraction condition.

Kanske et al. (2015) used the same setup. Individuals at risk for BD did not significantly differ from HC concerning behavioral and neural measures during distraction. Kanske et al. (2013) used a similar setup and partly the same sample as Kanske et al. (2015), examining two risk groups. For the first, an increased risk was indicated by high HPS scores. The second risk group consisted of unaffected first-degree relatives of patients with BD. This study investigated the activity of the neural network to emotional stimuli during the arithmetic task. No significant losses or aberrations concerning neural measures were observed in the risk groups.

In sum, there was no evidence on the behavioral level that people at risk for BD differ significantly from HC in distracting themselves from negative or positive emotions (Heissler et al. 2014; Kanske et al. 2013, 2015; Ladouceur et al. 2013). Concerning neural measures, patients at risk for BD showed increased neural activity while distracting themselves from positive emotional stimuli in the VLPFC (Ladouceur et al. 2013) and the inferior parietal cortex (Heissler et al. 2014). Despite increased activity in the VLPFC, connectivity analyses revealed a lower modulation of the amygdala (for positive and negative stimuli) and the DLPFC (for positive stimuli) by the VLPFC in the risk group (Ladouceur et al. 2013) during distraction. These findings contrast with studies that found no significant differences regarding neural measures between people at risk for BD and HC (Kanske et al. 2013, 2015).

Caseras et al. (2015) investigated differences in distraction in euthymic patients with BD-I, euthymic patients with BD-II, and HC assessed in an emotional n-back task. Particularly patients with BD-I exhibited impairments. First, this group showed inferior performance in the 2-back condition. Additionally, in the 2-back condition, BD-I patients showed slower RT due to distractors than the other groups. In contrast to the other groups, especially the distraction from happy distractors was impaired. Regarding the neural correlates of ER, patients with BD-I compared to other groups exhibited increased activity in the DLPFC, amygdala, and accumbens when happy distractors were present. In the presence of fear distractors, activity in these areas was highest in patients with BD-II. BD-II patients also showed a higher (inverse) functional connectivity between DLPFC and amygdala when fearful distractors were present. For patients with BD-I, results indicate lower structural integrity in the uncinate fasciculus, a white matter association tract that connects frontal regions with subcortical regions. In contrast, no significant structural aberrations were found for patients with BD-II.

Euthymic patients with BD-I in Kanske et al. (2015) did not significantly differ from HC regarding self-reports of emotions or neural activity during distraction. However, in Kanske et al. (2013), the same patient group solved the arithmetic task slower than HC when emotional pictures were present. In the presence of emotionally background pictures, patients with BD-I, compared to HC, exhibited increased activity in the right parietal cortex.

Regarding distraction in patients with BD, it is essential to note that all patients examined in the cited studies were in a euthymic state. While the self-report of emotions in euthymic BD patients provides no evidence of significant impairment (Kanske et al. 2015), the cognitive measures show the opposite (Kanske et al. 2013; Caseras et al. 2015). Regarding neural measures, patients with BD-I exhibited hyperactivity in DLPFC, amygdala, and nucleus accumbens during distraction from happy pictures and aberrations in structural integrity between frontal areas and the amygdala (Caseras et al. 2015). In contrast, patients with BD-II exhibited heightened activity in these areas and increased inverse functional connectivity between DLPFC and amygdala during the distraction from fearful stimuli (Caseras et al. 2015). Other results indicate no such aberrations in areas associated with emotion reactivity and regulation, but instead aberrations in areas associated with the cognitive distraction task (Kanske et al. 2013).

Heissler et al. (2014) did not find significant differences in behavioral measures of reappraisal in HC and participants at risk for BD. However, individuals at risk for BD exhibited hyperactivity in the amygdala during the reappraisal of negative emotions, indicating deficits in reappraisal, which did not manifest in behavioral measures.

Kanske et al. (2015) were able to clarify the role of frontal areas in the reappraisal of participants at risk for BD and euthymic patients with BD-I using the same setup as Heissler et al. (2014). In contrast to Heissler et al. (2014), differences in ER abilities were found for the high-risk group compared to their controls, with the high-risk group showing a decreased downregulation of positive emotions during reappraisal. Regarding neural measures, the high-risk group exhibited increased activity of the amygdala during reappraisal compared to HC. Most importantly, further analyses revealed that this was due to an impaired downregulation of the amygdala by OFC in the reappraisal condition.

Like Heissler et al. (2014), Ajaya et al. (2016) measured the risk for BD with the HPS. The authors were particularly interested in the effect of explicit and implicit instruction on reappraisal in individuals at risk for BD. Subjects in the video game task either received the instruction to concentrate on the game, were implicitly primed with the sentence-unscrambling task (Mauss et al. 2007) to apply reappraisal, or were explicitly instructed to apply reappraisal. Increased expression of anger indicated aberrant emotion reactivity in individuals with higher HPS scores. Interestingly, in Ajaya et al. (2016), HPS scores were correlated with respiratory sinus arrhythmia (RSA) deviations from baseline during ER after explicit instruction. This effect was absent when the instruction was implicitly primed or when participants received no instruction. Because RSA is an ER indicator, the authors conclude that these results suggest that explicit instruction enabled participants with higher HPS scores to successfully control their emotions. A characteristic of this analog study that has to bear in mind is the sample whose members were undergraduates and on a wide range of HPS scores. Thus, it differs from the strict definition of risk groups in the other studies reviewed in this section.

Taken together, the results of studies investigating reappraisal in participants at risk for BD revealed aberrations in emotion reactivity as well as in the ability to reappraise emotions. Regarding reappraisal, aberrations were found on behavioral (Kanske et al. 2015), as well as on neural levels (Heissler et al. 2014; Kanske et al. 2015). Neurally, aberrant emotion reactivity of negative emotions is reflected by hyperactivity in the amygdala (Heissler et al. 2014; Kanske et al. 2015). In participants at risk for BD, amygdala hyperactivity was associated with aberrant connectivity between the OFC and the amygdala during reappraisal (Kanske et al. 2015). While in HC, both areas were inversely coupled, they were positively associated with the risk group. The studies provide equivocal findings regarding the extent to which neural aberrations also affect the behavioral level. In two of three studies investigating reappraisal in people at risk for BD, no significant impairments at the behavioral level were found (Ajaya et al. 2016; Heissler et al. 2014).

Rive et al. (2015) investigated ER in a mixed sample, which included patients who had BD-I, patients who had BD-II, and HC in a picture viewing task with negative and positive stimuli in an fMRI setting. Patient groups were further divided into two subgroups depending on whether they were currently remitted (BDr) or depressed (BDd). In this section, we review the results of BD patients with a depressive episode; the results regarding remitted patients are reviewed in a corresponding section below. A notable characteristic of this study is that exclusively self-focused reappraisal was used. In healthy individuals, self-focused reappraisal has been shown to be less effective than situation-focused reappraisal (Willroth and Hilimire 2016). However, because situation-focused reappraisal is a relatively complex cognitive process, it has been argued that self-focused reappraisal for patients with affective disorders is more comfortable to apply (Rive et al. 2015). For BD-patients in a depressed state, the reappraisal of sad emotions compared to the reappraisal of happy emotions was compromised. Neurally, this patient group showed a decreased activity in rACC during the regulation of happy and hyperactivity during the regulation of sad emotions.

A distinguishing feature of Morris et al. (2012) is the BD-patient sample, which includes euthymic patients and patients who currently suffered from hypomanic episodes. Subjects were instructed to either use situation- or self-focused reappraisal, intensify, or maintain their emotions in a picture viewing task. The subjective affect rating revealed no significant differences between the patients and the group across all conditions. However, while the control group reported lower affect after the decrease condition than the maintain condition, affect ratings of patients who suffered from BD did not differ significantly between these conditions. Regarding the upregulation of emotions, both groups were able to increase their emotions compared to the maintenance of emotions. At the neural level, BD patients, as compared to HC, showed aberrant hyperactivity in the right VLPFC during up- and downregulation. Additionally, hyperactivity in rACC was found in BD patients during upregulation. Patients who suffered from BD differed from HC in the functional connectivity between areas in the frontal cortex and the amygdala. A negative correlation between activity in the left IFG and activity in the amygdala, found in HC during the downregulation of emotions, was absent in BD patients.

The studies investigating reappraisal in patients with current affective episodes show an affect-congruent impairment of reappraisal in the depressed state associated with hypoactivity in the rACC. Further, the opposite pattern was found in patients with a hypomanic episode. Here, affect-incongruent deviations in reappraisal were shown. These were associated with hyperactivity in the rACC and aberrations in fronto-limbic connectivity. A different approach to investigate reappraisal in BD was taken in the following studies. In these studies, reappraisal in patients with euthymic mood state was investigated.

In four studies we identified in our literature search, reappraisal abilities of patients with BD at a euthymic state were investigated (Corbalán et al. 2015; Kanske et al. 2015; Townsend et al. 2013; Zhang et al. 2018). Three further studies described the patient group's mood state as remitted (Gruber et al. 2014; Kjærstad et al. 2016; Rive et al. 2015).

In the study of Gruber et al. (2014), euthymic subjects that suffered from BD-I and HC saw movie clips that induced neutral, positive, or negative emotions. Instructions were to either observe these clips or to use reappraisal to modify emotions. Emotions were assessed by self-report, facial expression, and physiological measures (skin conductivity and respiratory sinus arrhythmia). Both groups were successful in reappraisal, implying intact reappraisal in euthymic patients.

Kjærstad et al. (2016) investigated reappraisal in patients with BD (a mixed group that contained BD-I patients and patients with BD-II) and HC. Patients were in a partly or fully remitted state. Two different ER paradigms were used in this study. First, in a social scenario task, participants read fictional social scenarios and corresponding self-beliefs. The scenarios could either be positive, negative, or neutral. Participants were instructed to either typically react to the scenarios or to dampen their emotions. Second, participants applied reappraisal in a picture viewing task. The authors assumed that the social scenarios task evokes stronger emotions due to patients' higher personal involvement. Indeed, in the social scenarios task, patients with BD were less able to downregulate negative emotions than HC.

In contrast, in the picture viewing task, patients with BD did not significantly differ from HC in reappraising negative emotions. It is difficult to conclude the divergent findings in both tasks since both tasks differ not only in the stimulus material but also in the instruction. The instruction to dampen emotions in the social scenarios task may be more challenging to implement than clearly defined reappraisal. Nevertheless, like Gruber et al. (2014), Kjærstad et al. (2016) did not found indications for significant deficits in reappraisal in euthymic patients with BD. However, it is not clear from either study whether euthymic patients and HC differ in their neural activity regarding reappraisal. Fortunately, the following studies addressed this question.

Euthymic patients in Rive et al. (2015) showed impaired self-focused reappraisal of happy and sad emotions. Neurally, patients with BD in a euthymic state did not significantly differ from HC during self-focused reappraisal.

Regarding neural measures, euthymic patients with BD-I in Kanske et al. (2015) showed amygdala hyperactivity during reappraisal compared to HC. Further analyses revealed aberrations in functional connectivity between the amygdala and the OFC, and the ventral ACC. While in HC, the connection between these areas was reverse, in euthymic patients, the connection was positive.

Corbalán et al. (2015) investigated differences in the reappraisal of euthymic patients with BD-I and HC during a picture viewing task. Self-report of emotions indicated that euthymic BD-I patients and HC were able to regulate their emotions successfully. Regarding self-report measurement, tests between the two groups did not reveal significant differences. However, both groups differed concerning neural activity. BD-I patients showed no reduction in amygdala activity during the reappraisal of negative emotions that healthy control participants exhibited. Further, in contrast to HC, patients with BD-I showed an involvement of the VLPFC in the neutral reappraisal condition and the negative viewing condition.

While Corbalán et al. (2015) give insights into neural activity during reappraisal, Townsend et al. (2013) primarily focused on differences in functional connectivity between the amygdala and frontal regions in euthymic patients with BD-I. This group showed reduced activity in bilateral VLPFC, insula, bilateral middle frontal gyrus (MFG), cingulate, and pre-supplementary motor area (SMA). Most importantly, reduced negative connectivity between the left amygdala and left VLPFC, left occipital gyrus, and right posterior cingulate was found in the patient group during reappraisal. In contrast, negative connectivity between the right amygdala and the right MFG was enhanced in BD-I patients. The results of Townsend et al. (2013) do not allow any conclusions about the causal direction of aberrant connectivity between the amygdala and other regions in patients with BD. This causal direction was addressed by Zhang et al. (2018).

Zhang et al. (2018) investigated causal connectivity between frontal areas (namely DLPFC and VLPFC) and the amygdala. Participants in this study were euthymic BD-I patients who had experienced past psychotic symptoms and a control group. Regarding the self-report of emotions, no evidence was found that both groups differ in their ability to regulate emotions. However, regarding neural measures, the influence of reappraisal on the connectivity from the left DLPFC to the left amygdala was weaker in patients with BD. No significant differences were found for the connectivity of the VLPFC and the amygdala.

Taken together, most studies that investigated reappraisal in euthymic patients reveal no significant behavioral differences in patients with BD and HC (Corbalán et al. 2015; Gruber et al. 2014; Kanske et al. 2015; Kjærstad et al. 2016; Townsend et al. 2013; Zhang et al. 2018). However, one study contrasts with these findings; Rive et al. (2015) did find deficits in the ability to use self-focused reappraisal in euthymic patients with BD-I. Regarding neural measures, amygdala hyperactivity of euthymic and euthymic patients suggests less effective reappraisal (Corbalán et al. 2015; Kanske et al. 2015). Analyses of functional connectivity revealed aberrations in the connectivity between the amygdala and OFC and ACC (Kanske et al. 2015), VLPFC, occipital gyrus, posterior cingulate, right middle frontal gyrus (Townsend et al. 2013), and DLPFC (Zhang et al. 2018).

In terms of behavioral measures, the reviewed studies suggest a relatively intact ER of risk groups. Regarding distraction, no impairments were found. Regarding reappraisal, two of three studies that investigated reappraisal in individuals at risk for BD revealed no significant impairments at the behavioral level (Ajaya et al. 2016; Heissler et al. 2014).

In contrast, the third study revealed reduced reappraisal effects in the risk group (Kanske et al. 2015). A crucial difference between these studies is the selection of risk groups. The studies that found no behavioral differences in reappraisal selected participants based on personality traits. In contrast, the study that examined first degree relatives of BD patients showed impaired reappraisal. Because the experimental setup of Kanske et al. (2015) and Heissler et al. (2014) was identical, the different selection of risk groups could cause contradictory findings. However, given that the difference between these risk groups was not systematically investigated, a conclusive interpretation of these findings is difficult. Previous studies have shown that both genetic risk and personality traits are highly predictive of the development of BD (Craddock and Sklar, 2013; Johnson et al. 2015; Kwapil et al. 2000; Rasic et al. 2014; Walsh et al. 2015).

Because risk groups showed sustained ER in most studies, the underlying neural activity of explicit ER is particularly interesting. Regarding distraction, this group exhibited aberrations in neural networks associated with emotion reactivity and ER (Heissler et al. 2014; Ladouceur et al. 2013). Individuals at risk exhibited VLPFC hyperactivity and reduced positive functional connectivity between VLPFC and the amygdala and the DLPFC during distraction from positive emotional stimuli (Ladouceur et al. 2013). During the distraction from positive emotional stimuli in an arithmetic task, individuals at risk showed hyperactivity in the left inferior parietal cortex, an area related to arithmetical operations (Heissler et al. 2014). Taken together with the behavioral results, these findings suggest that while individuals at risk can distract themselves from emotional stimuli at a level comparable to HC, this ability may be associated with increased cognitive effort, as indicated by hyperactivity in the corresponding areas. Please note, however, not all studies found significant neural aberrations in risk groups during distraction (Kanske et al. 2013, 2015). Regarding reappraisal, individuals at personality-based risk show hyperactivity in the amygdala (Heissler et al. 2014), which is also present in first-degree relatives, where aberrant positive functional connectivity between the amygdala and parts of the regulation network were demonstrated (Kanske et al. 2015).

These findings in risk groups provide information on whether aberrations in emotion reactivity and regulation of individuals with affective disorders are vulnerability factors that exist before the development of a disorder as claimed by the trait marker hypothesis (Rohde et al. 1990), or the consequence of affective episodes that occur during an affective disorder as claimed by the scar hypothesis (Rohde et al. 1990). Concerning these opposing hypotheses, the behavioral findings of the reviewed studies do not provide a definite conclusion. While one study revealed behavioral impairments in the explicit ER of individuals at risk for BD (Kanske et al. 2015), the majority of studies found no significant differences between risk groups and HC regarding behavioral measures (Heissler et al. 2014; Kanske et al. 2013; Ladouceur et al. 2013). However, in contrast to behavioral impairments, most studies revealed aberrations in frontostriatal areas (Heissler et al. 2014; Kanske et al. 2015; Ladouceur et al. 2013). Given that the reviewed studies' risk groups did not experience any preceding affective episodes, aberrations in explicit ER provide evidence for the trait marker hypothesis. These aberrations often cannot be detected in behavioral measures of explicit ER but are initially only visible in more sensitive neural measures. Interestingly, however, in a study that examined individuals at genetic risk, impairments in reappraisal were revealed on the behavioral level. In this sample, the additional activity alterations in areas involved in implicit ER (i.e., OFC and ventral ACC) may have led to additional behavioral impairments, which underlines the role of areas associated with implicit ER in explicit ER.

Concerning the use of explicit ER as a marker for the BD risk, neural aberrations during explicit ER seem to be promising due to their high sensitivity relative to behavioral measures. Most importantly, especially the regulation of positive emotions seems to be associated with neural aberrations (Heissler et al. 2014; Ladouceur et al. 2013). This detail is essential because valence-specific aberrations potentially distinguish bipolar disorders from other affective disorders, especially depression. Recently, it has been shown that patients with BD and UD can be distinguished based on structural and (valence-specific) functional neural aberrations associated with altered ER. (Bürger et al. 2017; Redlich et al. 2014; Repple et al. 2017). Concerning prophylactic treatment, in particular, the differentiation during the early course of the disorder could be critical. Although research on explicit ER of positive emotions in patients with major depression disorder and especially in risk groups is relatively rare, there is evidence to suggest that individuals at risk for unipolar depression (UD) differ from individuals at risk for BD regarding neural changes during explicit ER. For example, patients with UD exhibit decreased activity in the VLPFC during explicit ER of positive emotions (Canli et al. 2004; Kanske et al. 2012), which is the opposite of what was found for individuals at risk for BD during distraction from positive emotions (Ladouceur et al. 2013). To our knowledge, only one study investigated neural aberrations in individuals at risk for UD during ER of positive emotions (Simsek et al. 2017). This study revealed no significant aberrations. However, these studies' results cannot be regarded as conclusive, especially as there are, to our knowledge, no studies that have directly compared both risk groups.

Also, longitudinal studies are indispensable to investigate if, in BD risk groups, neural activity in areas associated with implicit and explicit ER is predictive and specific of developing BD later on. Nevertheless, the results in risk groups provide essential information about neural changes underlying BD since the results are not possibly confounded by medication or treatment effects. Studying medicated participants can lead to results that are challenging to interpret because it is difficult to determine whether differences between patients and HC are due to medication, the disorder, or an interaction of both. Additionally, patients—especially euthymic/remitted ones—have mostly undergone some form of psychotherapy that often specifically trains explicit emotion regulation, a factor that should be considered when interpreting the results in patient samples. However, recent tools such as the tracking software chronocord (Bauer et al. 2019; Pilhatsch et al. 2018) can be used to determine mood, sleep, and medication more specifically and can investigate their effects on emotion regulation and neural activity in patient groups.

Regarding neural measures, structural and functional aberrations in euthymic patients with BD were revealed (Caseras et al. 2015). During distraction, this group exhibited hyperactivity in DLPFC, amygdala, and nucleus accumbens. Interestingly, these findings differ between patients with BD-I and patients with BD-II. While the BD-I group exhibited hyperactivity in the areas mentioned above during distraction from happiness, the BD-II group exhibited the most pronounced aberrations during distraction from fear. Unlike BD-I patients, BD-II patients also exhibited increased inverse functional connectivity between DLPFC and amygdala during the distraction from fear. This finding is particularly interesting, combined with the behavioral findings of BD-II patients. Behaviorally, patients with BD-II showed no differences in the distraction of fear distractors compared to HC, which suggests that the strengthened connection between DLPFC and amygdala may be a compensatory mechanism that prevents behavioral losses.

Another study did not find any significant differences between euthymic BD patients and HC on the neural level (Kanske et al. 2015). However, again taking into account findings of hyperactivity in the right parietal cortex of patients with BD-I during distraction with an arithmetic task, a different picture emerges (Kanske et al. 2013). The right parietal cortex is associated with arithmetic tasks (Park et al. 2013). In line with the behavioral results, both studies suggest that BD patients can distract from emotions, but only by investing more resources in the distraction task.

Regarding reappraisal, hyperactivity in the amygdala of euthymic patients indicates impairments in ER (Corbalán et al. 2015; Kanske et al. 2015). Further, the reviewed studies revealed aberrations in frontal areas' functional connectivity and the amygdala (Kanske et al. 2015; Townsend et al. 2013; Zhang et al. 2018). However, the results are ambiguous as to which frontal areas are involved. Aberrant connectivity to the amygdala was found for the VLPFC, occipital gyrus, right posterior cingulate, right middle frontal gyrus (Townsend et al. 2013), DLFPC (Zhang et al. 2018), OFC, and ventral ACC (Kanske et al. 2015). While VLPFC and DLPFC were previously associated with explicit ER, the OFC and ventral ACC was more likely to be associated with implicit ER (Etkin et al. 2015; Phillips et al. 2008). Of particular interest here is the aberrant activity in the ventral ACC. Given that the ventral ACC and rACC together build a subsystem of the rACC (Etkin et al. 2011), as described above, aberrant activity in the rACC of euthymic patients is evidence for the scar hypothesis, which claims that the experience of affective episodes leads to permanent changes in patients with affective disorder. An interesting pending question is whether changes in activity in the ventral ACC predict the likelihood of recurrence.

A remarkable result of the studies investigating distraction in euthymic patients is the discrepancy between behavioral and neural results. While especially the studies that investigated situation-focused reappraisal did not find any significant behavioral differences between euthymic BD patients and HC (Corbalán et al. 2015; Gruber et al. 2014; Kanske et al. 2015; Kjærstad et al. 2016; Townsend et al. 2013; Zhang et al. 2018), neural aberrations in the amygdala, frontal areas, as well as in the functional connectivity between these areas were found. Regarding distraction, the studies reviewed suggest that euthymic patients with BD can successfully distract themselves from emotions, but this ability is associated with increased connectivity between areas associated with cognitive control and the amygdala (Caseras et al. 2015).

One possible explanation for the discrepancy is that it may be a statistical artifact resulting from underpowered study designs. (see the Limitations section of this article). An alternative explanation may be that neural aberrations represent compensatory mechanisms that prevent impairments on the behavioral level (Kjærstad et al. 2016). Considering that models of the neural basis of ER assume that frontal areas associated with cognitive control act to regulate striatal areas of emotion reactivity, this explanation also seems plausible. Aberrations were found in frontal areas and the connectivity between frontal and striatal areas. Further indications of a compensatory mechanism can be inferred from studies in which euthymic patients showed increased recruitment of areas associated with the distraction task but did not differ significantly from HC in terms of perceived emotions (Kanske et al. 2013, 2015). Further plausibility for the compensation hypothesis arises from the fact that most euthymic patients examined in the reviewed studies were on medication and/or psychotherapeutic treatment.

Given that medication and psychotherapy have compensatory effects on symptoms of mental disorders, it seems plausible if compensatory mechanisms are also reflected in brain activity. In any case, the interpretation of neural changes in euthymic patients must consider the effects of medication and psychotherapy. The next section outlines the implications of the findings for psychotherapy and trainability in more detail.