Introduction

The symptomatology associated with major depressive disorder (MDD) can be roughly divided into affective, vegetative and cognitive dimensions [1]. The cognitive domain comprises deficits in attention [2], visual learning and memory [3], processing speed [4] and executive functioning [5], such as working-memory (WM) impairments [6]. WM-deficits can be observed even after clinical remission [7, 8] and may have substantial impact on (psychosocial) functioning [9, 10]. Furthermore, WM deficits have been found to negatively predict MDD-treatment outcome [11,12,13], which underlines the relevance of WM-function as a diagnostic biomarker and potential therapeutic target for individuals with acute or remitted MDD. These findings emphasize the importance of better understanding altered WM-processes in MDD and their underlying neurobiological mechanisms.

Functional magnetic resonance imaging (fMRI) studies of healthy individuals have revealed numerous networks activated during WM, highlighting the crucial involvement of prefrontal and parietal regions, which constitute important nodes of the central executive network (CEN; [14,15,16]). Furthermore, in healthy controls (HC) significant deactivation of regions within the default mode network (DMN) during WM performance were reported [17, 18]. This is in support of the theory that suppression of these regions, associated with internally-directed and self-referential cognition during periods of task absence [19], is necessary for effective execution of cognitive tasks [20]. A recent meta-analysis of fMRI findings in MDD revealed stronger activation of DMN regions in patients during WM performance [21], which in line with the above reported findings in HC can be interpreted as the failure to adaptively suppress internally-directed cognition for the effective processing of external information [22]. Another meta-analysis of MDD studies [23] reported stronger activation specifically in the left dorsolateral prefrontal cortex (DLPFC) of MDD subjects compared to performance-matched HC, which is in support of the hypothesis that frontal hyperactivation represents a compensatory mechanism to counteract dysfunctional neural activation in other regions to preserve WM performance. Despite these trends, results from studies aimed at delineating differences in WM-related (de-)activation patterns between MDD-patients and healthy control subjects show considerable heterogeneity with various reported regional differences suggesting that our understanding of the neural mechanisms underlying WM deficits in MDD remains incomplete.

In another, hitherto largely unrelated stream of research resting-state fMRI is increasingly employed for the identification of altered neural mechanisms in MDD. Implying involvement of similar regions as in task-based studies, a recent meta-analysis reported large disruptions of resting-state functional connectivity (FC) within and between nodes of the DMN and CEN in MDD patients [24]. Other resting state studies have investigated the amplitude of low frequency fluctuations (ALFF), which quantifies changes of the BOLD signal as a marker of spontaneous neural activity [25]. In MDD increased spontaneous neural activity can be found in the medial prefrontal cortex (mPFC), a core hub of the DMN, and in the insula, which is associated with a coordinative role of switching between the CEN and DMN [26]. Basic research on WM-processes has repeatedly demonstrated associations between WM-performance and functional connectivity between [27, 28] and within these networks [29, 30]. In a machine-learning-based investigation, **. Mol Psychiatry. 2019;24:888–900." href="#ref-CR92" id="ref-link-section-d6654420e2743_1">92,93]. The fact that we allowed patients with concurrent antidepressant medication into the study may have further increased between-subject variability. It is possible that increased heterogeneity may have contributed to the lack of replication of previous findings, such as the increased frontal activity during WM-performance in MDD subjects [23, 71, 72]. Application of regression models with resting-state parameters and clinical- and demographical covariates as predictors and the task-activation parameters as the dependent variable within a sample of MDD subjects could further advance our presented approach and counteract problems associated with heterogeneity of MDD samples. Adding symptom-based clustering methods of MDD patients might help to reveal symptom-specific alterations of the rest-task relationship. Third, due to our decision to use a whole-brain voxel-wise approach, we were only able to evaluate regional coherence between resting-state fluctuations and task-evoked activity. The question of whether the mechanisms underlying the rest-task relationship (and their alterations) are region-inherent or driven by large-scale network dynamics and top-down processes therefore had to remain unanswered. Since significant rest-task correlations and neural alterations in MDD mainly emerged in regions associated with the CEN or DMN, modulatory effects within or between different networks, as factors causing the rest-task relationship, seem highly likely. For example, future research may want to investigate the coordinative role of the anterior insula in DMN and CEN (de)activation [94] and between-network dynamics by exploring connectivity-based resting-state indices and their relation to task-evoked activation.

Conclusion

Taken together, these findings suggest that resting-state activity reflects important properties of WM processes and their neural representations. The fact that a consistent pattern of correlations was found across HC and MDD-patients underlines the applicability and relevance of resting-state data for the understanding of brain functionality. Most importantly, analysis of rest-task relationships identified meaningful MDD-associated differences involving main hubs of the CEN and DMN that would have remained unnoticed in analyses of separate parameters. In conclusion, the integration of rest- and task data with parameters of their relationship offers an avenue to gain a more comprehensive understanding of the processes underlying cognitive deficits and network mechanisms altered in MDD.