Brain function and mechanisms in adolescent depression: A computational psychiatry approach

Abstract

Adolescent major depressive disorder (MDD) is associated with major impairments in the quality of life and a dramatically heightened suicidality. To improve understanding of the etiology of adolescent MDD and advance the efficacy of treatments for affected individuals, computational cognitive neuroscience has developed novel methods that find a growing number of application in psychiatric research. This dissertation project examined the neurobiology of incentive and emotion processing in adolescent MDD with functional magnetic resonance imaging (fMRI) and provides a framework for integrating methods for studying healthy and impaired neurodevelopment. The first study (Chapter 2) presented in this thesis investigated the neurodevelopment of the functional coupling within the incentive network in reward and loss contexts. In this study, participants with an age range from 11 to 35 years covering early adolescence until young adulthood performed a monetary incentive delay task, where they had to press a button to either receive a monetary reward or had to avoid a monetary loss during individual trials. The aim of the analysis was to (1) develop a computational model capturing effects of expected values and predictions errors on response vigor and (2) their association to the developmental, age-related state of corticostriatal connectivity. With increasing age, participants improved their behavioral discrimination of low and high expected values in tandem with keeping more stable value representations. This suggests that adult participants were less prone to be negatively affected by feedback when performing the task. Transient corticostriatal connectivity changes were associated with the developmental change. Particularly, a stronger influence from the prefrontal cortex over the striatum was found and suggests a shift towards a more incentive-driven, motivated behavior in adulthood. In the second study (Chapter 3), we employed the same monetary incentive delay task in two matched groups of adolescents (11-18 years), of which one was a group of adolescents with major depressive disorder. The focus of this investigation was to identify possible behavioral or neural disruptions of incentive processing in adolescent MDD within corticostriatal networks. Behaviorally, we established that healthy adolescents used a more complex dual learning model to learn cue-outcome associations throughout the task. The behavior of participants with MDD was best described with a simpler learning model, with the learning rate of depressed individuals being lower compared to healthy controls. This suggests a limited capacity to update representations of value in adolescent MDD. Analysis of neural correlates during loss processing revealed that differential encoding of errors in the orbitofrontal cortex in depression was linked to aberrant gain control of this region. Previous reports of disrupted reward processing in depression were not confirmed in this work. While the first two studies were testing incentive processing, the second focus of the thesis was to develop a task for the assessment of the emotion processing circuitry in adolescent MDD. In Chapter 4, the functional architecture of the prefrontal-amygdala network in a group of 33 healthy adults with a newly developed, ecologically more valid dynamic face- and shape-matching task was examined. The aim of this study was to identify valence-sensitive connectivity patterns within the prefrontal-amygdala network that could serve as candidate pathways underlying valence-specific dysfunctional processes in (adolescent) MDD. We identified valence-dependent coupling between the amygdala and the medial prefrontal cortex that showed sensitivity to aversive and ambiguous emotional information. In a preliminary fourth study (Chapter 5), a similar face- and shape-matching task was performed by two groups of adolescents with and without a diagnosis of MDD. We used the linear ballistic accumulator model to split the decision process into mechanistically interpretable components (e.g. processing efficiency reflected as drift rate of the model). We found strong evidence that adolescents with MDD exhibited slower evidence accumulation of ambiguously/neutrally valenced faces. For processing ambiguous faces, this less efficient information processing was associated with hypoactivity in the subgenual part of the medial prefrontal cortex. We conclude that deficient perception and evaluation of ambiguous social cues underlie adolescent MDD, providing insights into a dysfunctional emotion processing mechanism of the disorder. This thesis extends the knowledge about typical and aberrant development of functional brain circuits during adolescence in the domains of incentive and emotion processing significantly. First, we found how the emergence of corticostriatal connectivity give rise to motivated behavior across adolescence. Second, we advanced the understanding of adolescent MDD by demonstrating that it is associated with (1) a maladaptive learning mechanism from loss and (2) inefficient processing from ambiguous emotional faces. Thus, this work provides novel mechanistic insights into one of the most debilitating psychiatric disorders and exemplifies a methodological framework of how the integration of behavioral and neural models can be harnessed to study the human brain during development and in psychiatric disorders.