Introduction

Mood disorders and anxiety disorders are characterized by the abnormal persistence of affective symptoms and cognitive abnormalities [14] that contribute to poor psychosocial functioning and outcome in these patient populations. Mood and anxiety disorders have consistently been identified as a leading cause of overall disease burden worldwide because of their significant public health cost, morbidity, and mortality [5]. This is partially due to the lack of highly effective treatments. Despite numerous pharmacological agents available, approximately 30–40 % of patients do not achieve adequate therapeutic response [68].

Several psychologically based interventions have also been developed. Cognitive behavioral therapy (CBT) [912] and interpersonal psychotherapy (IPT) [13, 14] are two commonly utilized, evidence-based psychotherapies for treating anxiety and depression. Although promising post-treatment and follow-up remission rates have been observed in numerous randomized trials [12, 14, 15], studies investigating long-term treatment outcomes have reported notable decreases in remission rates over time [1618], with anxiety and depressive symptoms often returning to clinically significant levels. Failure to achieve sustained rates of remission in mood and anxiety disorders has prompted the development of novel, alternative interventions designed to target the key underlying deficits in neural network functioning.

Cognitive bias modification (CBM) represents a first attempt at a neuroscience-based intervention. CBM directly targets pathological cognitive content (e.g., negative interpretations of events) or cognitive processes (e.g., biased orienting of immediate attention toward negatively valenced stimuli) in mood and anxiety disorders [19]. Despite initially promising studies [20], meta-analyses of the preliminary research findings on CBM interventions suggest that treatment effects are generally small to non-significant in anxious and depressed samples [21•], especially when administered in routine clinical settings. Nonetheless, CBM has been very useful in introducing the concept of neuroscience-informed cognitive-affective training interventions that can be further refined to improve short- and long-term outcomes.

It is generally acknowledged that the efficacy gap observed in treatment reflects misalignment between the mode of action of the intervention and the mechanisms underlying disease expression. The need to address this gap has been clearly articulated as a major strategic goal of the National Institute of Mental Health (NIMH) that mandates “research that translates emerging findings on the neuroscience and behavioral science of mental disorders into novel psychosocial (e.g., cognitive strategies and innovative behavioral approaches) and other non-pharmacological interventions…that will alter dysfunctional neural circuits and psychological processes underlying mental disorders to reduce symptoms” [22].

The opportunities for the development of novel translational interventions continue to grow as we build upon our understanding of the intersection between neuroscience, cognitive science, and psychiatry. In this paper, we describe the rationale and method for developing cognitive-affective training interventions for patients suffering from mood and anxiety disorders and provide an example of how this could be applied in the treatment of major depressive disorder (MDD).

Relevant neuroscience findings informing the development of cognitive-affective training for mood and anxiety disorders

The basic premise of a neuroscience-informed strategy for cognitive-affective training is that repeated training targets psychopathology (particularly cognitive and affective processing biases) by way of enhancing top-down cognitive control and modulates neural network dysfunction through brain plasticity mechanisms to effect remedial change [23, 24]. The neural networks underlying affect processing have been explored in healthy individuals, and abnormalities in these networks have been consistently shown in patients with disorders. The overlap in neural abnormalities among these disorders highlights the opportunity for applying trans-diagnostic cognitive-affective training interventions that target these mechanisms as potential new intervention strategies.

Neural networks involved in affect processing

Upon perception, affective information is relayed to the thalamus, which projects to the amygdala/hippocampal complex [25]. The amygdala and hippocampus, proximal structures that exhibit mutual facilitation during this processing, evaluate the affective salience of the stimuli [26, 27]. The ventral striatum (also referred to as the nucleus accumbens) is another region, intricately connected to the amygdala/hippocampus complex and projecting to the thalamus, which appears to preferentially promote processing of rewarding or positive stimuli [28]. Additional bottom-up information is also processed by the subgenual cingulate cortex (SGC) [29]. Input from these regions is mostly directed to the medial prefrontal cortex (mPFC), a region involved in conscious awareness of the affective salience of information and in self-referential processing [3033]. Top-down cognitive control, on the other hand, is primarily exerted by lateral, ventral, and dorsal prefrontal cortical (PFC) regions, particularly the dorsolateral PFC (dlPFC) and dorsal anterior cingulate cortex (dACC) [26, 27]. In healthy individuals, amygdala activity increases during the processing of affective information [27] and shows an inverse relationship with dlPFC activity [26, 3436].

An expansive magnetic resonance imaging (MRI) literature has contributed to a detailed delineation of the brain regions implicated in mood and anxiety disorders, both in terms of structure (volume) and function.

Several meta-analyses have summarized the relevant literature in mood disorders [particularly MDD and bipolar disorder (BD)] and, to a lesser extent, in anxiety disorders. In terms of MDD, structural MRI (sMRI) studies have shown that the most consistent findings are volumetric decreases in the anterior cingulate cortex (ACC) extending further upwards to the ventromedial PFC (vmPFC), as well as the dlPFC, and downwards to the amygdala, hippocampus, and parts of the striatum (particularly the putamen) [3740]. In BD, there are now several reviews and meta-analyses that have synthesized the extensive primary sMRI literature [38, 4145]. Consistent volumetric reductions between patients compared to healthy individuals have been observed in the vmPFC, amygdala, hippocampus, insula, and ACC [42, 44, 45]. In anxiety disorders, the literature is more limited. Nevertheless, volumetric reductions compared to healthy controls have been reliably identified in the amygdala, hippocampus, and the vmPFC in patients with post-traumatic stress disorder (PTSD) [4648], panic disorder (PD) [49, 50], generalized anxiety disorder (GAD), and phobias [51].

The functioning (activity) of the regions implicated in affect processing, and inter-regional interactions, has also been examined using functional MRI (fMRI) both during task-related and task-independent (“resting state”) conditions. Regardless of condition, individuals with MDD demonstrate hyperactivity notably in the thalamus, the amygdala/hippocampal complex, and the ventral striatum [5255]. In contrast, concurrent hypoactivity has been observed in lateral PFC structures, particularly during the processing of affective information [56]; this is believed to contribute to the impaired cognitive control and perseverative thinking symptoms observed. Similarly in patients with BD, hyperactivation is observed in the amygdala/hippocampus complex, the putamen, and the thalamus, with hypoactivation in lateral PFC regions [54]. A similar pattern of increased limbic hyperactivation with prefrontal under-engagement has been reported across anxiety disorders [5557].

Taken together, the evidence presented above suggests that persistent affective dysregulation has a diagnosis-independent neural signature that is characterized by structural and functional abnormalities within the neural circuits involved in affect processing. Specifically, reduced lateral PFC-related cognitive control over medial cortical and subcortical regions involved in affect generation and processing result in biased and prolonged processing of negative affective information (e.g., perseverative worrying, ruminating, etc.). Figure 1 presents a schematic depiction of the neural network dysfunction underlying the abnormal cognitive-affective processing observed in mood and anxiety disorders and representing targets for cognitive-affective training interventions.

Fig. 1
figure 1

A neurobiological model for cognitive-affective processing abnormalities in mood and anxiety disorders and targets for interventions. Hyperactivation of the thalamus (THAL), amygdala (AMY), and hippocampus (HIPP) upon perception of emotionally salient stimuli is associated with increased subgenual cingulate cortex (SGC) activity. SGC activity is associated with increased medial PFC (mPFC) activity; this activity integrates limbic feedback and relays further upstream to other prefrontal cortex (PFC) regions. Concurrently, PFC regions (dlPFC, vlPFC) and the dorsal anterior cingulate cortex (dACC)—which facilitate inhibitory control over limbic regions and are associated with cognitive control and emotion regulation—are observed to be hypoactive. The functional connectivity between the PFC and other brain regions (AMY, HIPP) appears weakened, which suggests poor cognitive control over these regions. Dashed lines and asterisks indicate functional connectivity abnormalities that result in impaired cognitive control and emotion regulation, and delineate targets for cognitive-affective training interventions.

Neuroplasticity

Neuroplasticity refers to the brain’s ability to change due to experience and alter future responses to related stimuli [58]. Neuroplasticity is a multifaceted phenomenon that can involve functional changes such as the strengthening or weakening of connections between neurons and remodeling of synapses [59, 60] or structural changes such as neurogenesis. Synaptic remodeling is a core dimension of neuroplasticity. Synaptic pruning and synaptogenesis refer to the idea that neuronal connections are routinely being removed or recreated based upon how they are experientially used [61•]. A fundamental principle is that frequent neuronal co-activation leads to the creation of new synapses or the fortifying of existing ones. Alternatively, the synaptic connections between neurons that do not regularly co-activate weaken. Consequently, the quantity and strength of brain synapses can change over time, impacting neural network functioning.

Whole cell plasticity, also known as homeostatic plasticity, is the change in the excitability of neurons [62]. Neuronal excitability reflects the assimilation of excitatory postsynaptic potentials (EPSPs) and inhibitory postsynaptic potentials (IPSPs) and is influenced by the function of ion channels in the axon, dendrites, and cell body. Changes to the electrical properties of neurons can influence spike generation, synaptic integration, and other core mechanisms at the cellular level. Higher order brain functioning, such as learning and memory in particular, can be altered as a result of these individual neuronal changes.

Neurogenesis may also contribute to neuroplasticity. However, in the adult brain, this contribution is likely to be less widespread compared to synaptic changes and homeostatic plasticity. In the adult brain, neurogenesis is mostly convincingly demonstrated in the hippocampus [63, 64] and in the subventricular zone (SVZ) [65]. More recent evidence implicates further brain regions not commonly identified as sites of neurogenesis, including the neocortex [66, 67]. Taken collectively, synaptic pruning, synaptogenesis, homeostatic plasticity, and, to a lesser extent, neurogenesis begin to explicate mechanisms by which brain plasticity can unfold on structural and functional levels.

Preclinical and clinical studies provide preliminary support for the notion that cognitive interventions can influence neuroplasticity. Preclinical studies, for example, using a neonatal ventral hippocampal lesion rat model of schizophrenia, have demonstrated that cognitive training in adolescent rats conferred protection against cognitive control deficits and modified hippocampal neural activity as a mechanism of effect [68].

Clinical studies using PET and fMRI involving human subjects receiving cognitive therapy for depression and anxiety identified functional changes in neural circuits responsible for cognitive control and affect, specifically by increasing dlPFC activity and subsequently reducing limbic system activity during affective information processing [69]. Studies have also shown that cognitive priming for emotional stimuli had a persistent effect on subsequent amygdala activity [70]. A systems neuroscience rationale has also been proposed for the possible mechanisms of cognitive-affective training interventions. This model emphasizes the distributed infrastructure of neural circuits that maintain cognitive and affective processing, and relies on brain plasticity [71••]. As the human brain learns, cognitive/affective inputs and action outputs ultimately become manifested by greater and more coordinated groups of neurons across multiple brain regions and levels of processing. Taken together, cognitively based interventions that include cognitive-affective training may have the potential to effect change in brain neurobiology and functioning in ways that are consistent with improvements in clinical outcomes.

Neuroscience-informed cognitive-affective training: paradigm and proof of concept application in MDD

We have developed a new cognitive-affective training intervention for MDD. The intervention aims to enhance cognitive control of emotional information processing by specifically targeting abnormal activation patterns in dlPFC and amygdala that underlie the cognitive control and emotion regulation impairments observed in MDD. To accomplish this, the intervention combined working memory and facial affect identification tasks, which have been shown to specifically target activity in the dlPFC [72] and amygdala [73], respectively. The task performed during the intervention requires that patients identify the emotions expressed on a series of faces displayed on a computer screen while remembering the specific sequence of faces. The working memory component of the task utilizes an N-back working memory paradigm in which participants indicate if the emotion expressed on a currently viewed face matches with the emotion N faces back. Depending on the patient’s performance, the N level varies; N can increase or decrease across all blocks while patients are instructed to complete many blocks for each session. Administering the task in this way effectively calibrates to the participants’ ability level while proactively challenging them as a means of enhancing engagement of the targeted brain structures. The target engagement of this task has been validated using fMRI techniques in a single, non-progressive session in healthy volunteers; it simultaneously recruits dlPFC structures involved in working memory and limbic structures involved in affective processing (amygdala) [74].

Proof of concept for the efficacy of this intervention in MDD has been provided in a pilot study [75•]. The interventions are based on the assumption that exercising the ability to manipulate emotional information in working memory could enhance cognitive control for emotional information processing via modulation of the corresponding neural networks. Enhancement of cognitive control is expected to reduce rumination and depressive symptoms. Accordingly, we conducted a placebo-controlled clinical trial of our cognitive-affective training intervention. Twenty-one patients with MDD participated in this trial and were seen twice per week for 4 weeks. Half the sample received the experimental task while the other half was randomly assigned to an active sham task, which involved a straightforward working memory exercise using an adaptive N-back task with shapes as the stimuli, not emotional faces. This task is known to activate the dlPFC but not the amygdala. The effectiveness of the cognitive-affective training was evaluated using a battery of measures to assess: (1) changes in cognitive biases and rumination; (2) changes in neurocognition/working memory; and (3) changes in MDD symptoms. At study outcome, significant differences in reductions in negative affective bias in short-term memory and MDD symptoms were reported between the groups [75•] but both groups showed similar, small improvements in working memory. These findings ultimately suggest that cognitive-affective training holds the promise of enhancing cognitive control for affective information processing and modulating the underlying neural networks, and that these changes could have antidepressant or antianxiety effects by fortifying emotion regulation and reducing the perseverative thinking (e.g., worrying, rumination) observed in mood and anxiety disorders.

Conclusion

As the body of knowledge about the connection between cognitive and affective neuroscience and psychopathology continues to grow, opportunities to develop and refine novel intervention strategies capitalizing on brain plasticity will grow with it. Empirical support is currently the strongest for strategies like cognitive bias modification and cognitive therapies. The development of cognitive training exercises as neurobehavioral therapies, which target underlying neural components linked to the pathophysiology of specific disorders, is now also a possibility. The intention of this paper is to provide an outline for how researchers can develop, and clinicians employ, cognitive-affective training interventions to treat anxiety and mood disorders in non-research settings. Further research is still needed for understanding the full effects of these mechanisms. Finally, subsequent studies may also investigate the effects of cognitive-affective interventions as an augmentation to current evidence-based psychotherapy or pharmacotherapy treatments for anxiety and mood disorders, as well as the sustainability of training effects after the regimen is completed.