Default mode network activity in bipolar disorder

Since its discovery in 1997, the default mode network (DMN) and its components have been extensively studied in both healthy individuals and psychiatric patients. Several studies have investigated possible DMN alterations in specific mental conditions such as bipolar disorder (BD). In this review, we describe current evidence from resting-state functional magnetic resonance imaging studies with the aim to understand possible changes in the functioning of the DMN in BD. Overall, several types of analyses including seed-based and independent component have been conducted on heterogeneous groups of patients highlighting different results. Despite the differences, findings seem to indicate that BD is associated with alterations in both frontal and posterior DMN structures, mainly in the prefrontal, posterior cingulate and inferior parietal cortices. We conclude this review by suggesting possible future research directions.


Introduction
Bipolar disorder (BD) is a complex mental disorder characterised by severe mood fluctuations associated with functional and cognitive disabilities that often persist throughout the entire life of patients (Solé et al., 2017). Specifically, BD is expressed with alternating phases of (i) mania, including abnormally elevated mood and increased energy levels and (ii) depression, consisting of severely depressed mood, anhedonia and insomnia/hypersomnia. Moreover, patients with BD show various degrees of neuro-cognitive impairments affecting several domains such as working memory, attention and executive control, all impacting on socio-occupational functioning (Robinson et al., 2006).
Considering the neurobiological nature of BD, it is crucial to study the structural and functional brain alterations associated with the disease, in order to provide novel insights for future neurobiological models and treatments. To date, however, neurobiological underpinnings of the disorder are still discussed, hindering the development of new pharmacological therapies and targeted rehabilitative interventions. Findings on brain structure in BD suggest ventricular, prefrontal and temporal abnormalities but remain largely inconsistent (Delvecchio et al., 2020). The most consistent findings on brain function in BD concern brain activation patterns during emotion processing, emotion regulation and reward processing tasks. Overall, BD seems to be marked by a generalised dysfunction of the dorsal and ventral systems, composed respectively of (i) the prefrontal cortex (PFC), anterior cingulate cortex (ACC) and the hippocampus and (ii) the insula, amygdala and ventral striatum (Chen et al., 2011;Phillips and Swartz, 2014). However, it is not clear whether the aforementioned networks are altered only during the execution of specific tasks or also at rest .
Recently, greater attention has been given to the investigation of brain activity and connectivity in BD at rest, mostly by means of resting-state functional magnetic resonance imaging (rs-fMRI). Particularly, rs-fMRI is used to evaluate the functional interactions among remote neural systems that occur in a resting or task-negative condition, i.e. when an explicit task is not being performed. This technique provides information about intrinsic brain activity, not dependent on specific tasks or experimental settings. Most commonly, the brain functional connectivity (FC) at rest is studied by analysing the statistical dependencies among the blood oxygenation level-dependent (BOLD) signals from multiple brain regions, which reflect the spontaneous fluctuations in the oxygenation state of blood (determined by blood flow, blood volume and oxygen consumption) within them (Smitha et al., 2017).
The physiological substrates of resting-state activity are currently the object of speculation (Rosazza and Minati, 2011). By contrast, functional connectivity (FC) at rest (i.e. the statistical dependence among spatially remote neuronal systems (Friston et al., 1993)) is suggested to be not a simple 'background' activity but a complex dynamic involved in cognition and memory consolidation (Rosazza and Minati, 2011) consuming more than 20% of body's energy supporting neuronal signalling and functioning (Raichle and Mintun, 2006).
Rs-fMRI networks can be studied employing different analysis pipelines that primarily fall in two main categories, (1) seed-based analyses (SBA), which require a-priori definition of the brain network nodes, (2) blind-source separation techniques, which adopt an unbiased approach by grouping brain voxels/regions based on their latent time series (independent component analysis, ICA) (Joel et al., 2011). For details on these approaches see Smitha et al. (2017).
These techniques have enabled the discovery of multiple brain networks characterised by highly correlated spontaneous BOLD fluctuations. The most commonly studied networks include (i) the Salience Network (SN), composed of the ACC and the insula; (ii) the central-executive network (CEN) consisting of frontoparietal regions and the posterior parietal cortex; (iii) the frontoparietal network (FPN) that includes the cortex surrounding the intraparietal sulcus, the inferior parietal lobe, the dorsal motor cortex and the inferior frontal cortex; and (iv) the default-mode network (DMN) extending from the prefrontal medial cortex (mPFC) to the precuneus, posterior cingulate cortex (PCC) and inferior parietal cortex. The DMN is recognised as a key component of the brain functional architecture. The vmPFC composes the anterior part of the DMN and has been shown to modulate social behaviour, mood regulation, executive functioning and control processes (Damasio et al., 1994). The PCC, the precuneus and the lateral parietal cortices, which form the posterior DMN, are involved respectively in attentional regulation, consciousness, mental imagery and episodic memory processes (Cavanna, 2007;Leech and Sharp, 2013). All these functions have been found altered (at different extents) in psychotic disorders and BD (Clark et al., 2002;Torres et al., 2007). Moreover, the precuneus has been recently suggested to be part of a hippocampalparietal network involved in learning and consolidation of everyday experiences. The global DMN is involved in multiple cognitive and affective functions such as emotional processing, self-referential mental activity, mind wandering, recollection of experiences and possibly exerts a modulatory role during attentional demanding tasks (Raichle, 2015). Since attentional deficits are a core feature of BD (Solé et al., 2017), studying the DMN in patients with BD might prove particularly beneficial to help to understand the neurobiological underpinnings of the disease.
Considering the potential role of the DMN in BD, the aim of this review is to describe current evidence of DMN connectivity in patients with BD.
The data search was conducted on the Pubmed, Scopus and Google Scholar databases. The following keywords were used for the search: 'default mode network' AND 'bipolar disorder'. The inclusion criteria were: (i) original publication in a peerreviewed journal between 2010 and 2020, (ii) English language, (iii) a diagnosis of BD for the patient groups and (iv) the analysis of the DMN activity patterns through rs-fMRI. After an initial screening 71 studies were identified of which 23 studies were included in the review. Table 1 provides a description of methods and results of the included studies. In detail, 11 studies (47%) analysed the rs-fMRI activity through ICA, providing thus an unbiased overview of brain activity at rest as a whole (Joel et al., 2011), while ten studies (43%) adopted an SBA approach, which requires a-priori hypotheses on the regions of interest within the brain (Smitha et al., 2017). Finally, the remaining two studies analysed the rs-fMRI activity through a network analysis (Wang et al., 2016) and the Data Processing Assistant for Resting-State (DPARSF) analysis pipeline, a MATLAB toolbox based on the SBA principles which gives information about regional FC and homogeneity (Brady et al., 2017). Only six out of 23 studies enrolled unmedicated BD patients. Yet, only four out of 23 studies recruited mixed samples of BD patients based on the subtypes (i.e. BD-1, BD-2) or the clinical phases of the disorder (e.g. depression, mania, euthymia) (Martino et al., 2016;Rive et al., 2016;Brady et al., 2017;Zhong et al., 2019).
Overall, findings from the reviewed studies suggest that BD patients compared to Healthy Controls (HC) are characterised by marked functional alterations of both anterior and posterior hubs of the DMN, including the PFC, PCC, precuneus and inferior parietal cortex. These alterations are mostly mixed across studies, consisting of both hypo and hyper-connectivity of the anterior DMN at rest (Table 1).
Specifically, 14 studies (60%) showed aberrant functional activity in the posterior hubs of the DMN, of which five in the precuneus (21%) and 12 in the PCC (52%). These alterations are mostly hypo-activations of BD patients when compared with HC. Specifically, most of the studies showed FC reductions in BD compared to HC within these regions (Khadka et al., 2013;Meda et al., 2014;Teng et al., 2014;Magioncalda et al., 2015;Rey et al., 2016;Wang et al., 2016Wang et al., , 2017Luo et al., 2018;Qiu et al., 2019). However, few recent studies found also hyper-connectivity between PCC, the mPFC and the insula (Liu et al., 2015). A single study by Zhong and colleagues differentiated between psychotic and nonpsychotic BD and showed an increase in the FC of the PCC only in psychotic BD v. HC, possibly indicating specificity of the PCC alterations for the psychotic conditions . Moreover, a study by Ford and colleagues (2013) found a significant negative correlation between FC in the PCC and a so-called 'bipolarity index' based on five key domains of the disorder: signs and symptoms, age of onset, course of illness, response to treatment and family history (Ford et al., 2013). Even though the authors do not speculate about this specific finding, the PCC has been previously found to be a key region in several cognitive processes such as monitoring of awareness, arousal, internal thought and attention control (Leech and Sharp, 2013). Of note, a single study performed a machine learning discrimination analysis based on the DMN FC between depressed BD patients and patients suffering from major depression (MDD), reaching a 69% accuracy (Rive et al., 2016). This result suggests that specific alterations of the DMN could be considered an endophenotype of BD. Partially in line with this hypothesis, two studies investigated FC of the DMN in both BD and schizophrenic (SCZ) patients and their close relatives (Khadka et al., 2013;Meda et al., 2014). Notably, only the study by Khadka and colleagues found alterations in the frontal hub of the DMN in both BD patients and their relatives (Khadka et al., 2013).
Some studies show the presence of altered connectivity in BD patients, specifically between the posterior part of the DMN and the anterior cingulate cortex (ACC) suggesting a possible relationship between the two regions (Magioncalda et al., 2015;Rey et al., 2016). Specifically, Magioncalda and colleagues (2015) found that in BD patients, alterations in the FC between the frontal and posterior hubs of the cingulate cortex can lead to excessive focusing on external contents and, ultimately, manic phases (Magioncalda et al., 2015).
Only two studies found no alterations in the DMN of BD patients (  colleagues (2017) failed to find DMN alterations of BD patients, however, they found a reduction in the variability of the dynamic connectivity between the medial PFC and PCC associated with impaired processing speed (Nguyen et al., 2017). Unlike the standard rs-fMRI analysis pipelines, which estimate the average brain connectivity over time, dynamic connectivity methods allow to analyse how the FC between two or more brain regions changes during the entire scan duration obtaining useful information about the dynamic variability of the brain regional synchronization (Rashid et al., 2014). This result suggests the importance of considering not only the average FC over time but also the dynamic interplay between brain regions. Goya-Maldonado and colleagues (2016) found alterations in the DMN of patients suffering from major depression but not in the BD group. However, they found alterations in the FPN, a network partially overlapping with the DMN itself (Goya-Maldonado et al., 2016;Bellani et al., 2020). Lastly, also Yip et al. (2014)did not find any alteration of the FC of the DMN in BD(. However, the authors recruited only young BD-II patients (mean age 23.07 ± 3.73) without any history of medication. The authors suggest that the absence of alterations within the DMN may be due to specificity of the recruited group (e.g. diagnosis, last clinical phase) or on the absence of medication.
To conclude, current evidence from rs-fMRI studies suggests that BD patients show FC alterations (i.e. hypo and hyperconnectivity) compared to HC in both frontal and posterior hubs of the DMN. In particular, the evidence to date suggests that BD patients compared to HC have altered FC of the DMN in a number of regions including the PFC, the ACC and PCC and the precuneus. However, only a paucity of studies has been conducted and evidence is still sparse. The inconsistency of findings may be due (in part) to methodological issues such as heterogeneous analysis pipelines (e.g. ICA v. SBA v. regional homogeneity analysis), different BD populations (e.g. euthymic, BD-I, BD-II) or the lack of control for confounders (e.g. pharmacological treatment) that may have influenced the results (Vargas et al., 2013;Martino et al., 2016). Moreover, a recent study by Bellani et al. (2020), examining the resting state activations of a group of euthymic BD patients and HC, found alterations in the DMN only when examining between network connectivity, reporting no differences between the two groups in withinnetwork functioning. The result suggests the importance of examining both within and between-network connectivity to achieve a global understanding of the BD euthymic condition.
Detailed neurophysiological correlates of resting-state activity are currently the object of speculation and only specific analysis techniques (i.e. dynamic functional connectivity) can account for the complex dynamics in the interactions between brain regions as suggested by some studies (Brady et al., 2017;Wang et al., 2020). Furthermore, since the BOLD contrast reflects hemodynamic/metabolic processes in the brain, the interpretation of rs-fMRI findings in terms of the underlying neuronal activity is not straightforward and would require the integration of electrophysiological information (Maggioni et al., 2016). Despite these limits, rs-fMRI studies allow the whole-brain mapping of the functional connectomes of specific pathologies such as the BD and provide new data for the development of neurobiological models and theories.
In view of the recent conceptualisation of BD in terms of neural circuitry, future studies should consider a multimodal perspective in order to (1) investigate the complex link between structure and function in the DMN and other networks involved in BD pathophysiology, (2) identify neurodevelopmental and maturation trajectories of FC patterns in relation to the natural course of the disease and in response to different therapeutic strategies, (3) disentangle the contributions of genetics, environment and their interaction to the large-scale functional brain network abnormalities underlying BD symptomatology.

Data
The data that support the findings of this study (search query) are available on request from the corresponding author.