Differences in verbal and spatial working memory in patients with bipolar II and unipolar depression: an MSI study

Up to 60% of patients with bipolar disorder are misdiagnosed as unipolar disorder what may considerably increase the risks of switching to suicide and poorer treatment responses [1]. Bipolar and unipolar disorder are sometimes confused in clinical practice which both could be manifested as symptoms of depression. Additionally, depressive symptoms may be either a risk factor or prodrome for cognitive deficits [2]. Cognitive deficit is one of the common symptoms of both bipolar and unipolar depression, even seems to be present in individuals in the remitted state [3, 4]. Furthermore, working memory is the basis for individuals to successfully perform various cognitive tasks, also plays an important role in complex cognitive activities [5]. Our and other’s previous studies suggested that decreased working memory performance in bipolar and unipolar disorder reflected neurofunctional deficits [6, 7].

The brain oscillations integrate functional brain systems across multiple spatiotemporal scales. Gamma oscillation (30–120 Hz) is a significant indicator of the functional activity of neurons. Many researchers found gamma oscillation was instrumental to the process of working memory [8]. Roux et al. [9] used magnetoencephalography (MEG) and a delayed match-to-sample task and found that there was a linear relationship between working memory-load and gamma oscillation activity in the prefrontal cortex, as the increased difficulty of working memory task, the activity also increased. Further, they suggested that gamma oscillation as well is involved in the maintenance of working memory information [10]. The findings of Haegens et al. supported the views of Roux et al., who found that gamma oscillation significantly increased during working memory tasks especially information coding and information retention in healthy control, besides they observed a significant positive correlation between gamma activity and task performance [11].

The frontoparietal network (FPN) includes portions of lateral precuneus, dorsolateral prefrontal cortex (DLPFC), and ventrolateral prefrontal cortex (VLPFC), and is associated with externally-oriented and cognitively demanding mental activity [12]. On the one hand, patients with bipolar disorder presented abnormal activation in regions typically involved in the FPN, according to a meta-analysis [13]. A study investigating multiple networks in euthymic bipolar I patients also identified an aberrantly increased connectivity in the FPN [14]. On the other hand, a meta-analysis identified that reduced connectivity within FPN and imbalanced connectivity with other networks involved in internal or external attention may reflect depressive biases toward internal thoughts at the cost of engaging with the external world in unipolar disorder patients [15]. Activation within FPN could be obviously observed during cognitive tasks, researchers believed that FPN contributed to functions such as working memory, cognitive control, and decision-making processes, moreover might be associated with regulation of emotion [16, 17]. Some evidence indicated that verbal working memory (VWM) activation mainly implicates left FPN, while the right FPN is associated with spatial working memory (SWM) [18]. However, more studies concluded that the completion of SWM has resulted in both two sides of FPN [19].

However, the results on working memory in depression are very heterogeneous, some studies believed that working memory was not impaired in bipolar or unipolar depression patients [20, 21]. It is perhaps due to different stimulation of the experimental paradigm is inconsistent. To avoid it, a 2-back task in which stimulus presentation and required response matched was used to compare the impaired condition of verbal and spatial working memory. Simultaneously, we used MEG to explore the neuroelectrophysiological mechanism between working memory deficits and gamma oscillation within FPN.

Our study showed that the SWM of patients with whether UD or BP II was impaired, but VWM wasn’t impaired. Even in the depressive episode, the patterns of FCs within these two groups of patients were different when performing working memory tasks. The altered FCs of FPN suggest that the IPL may be the key component to distinguish between BP II and UD.

The performances of UD and BP II groups were similar in two working memory tasks, and the accuracy was independent of clinical characteristics. Some research suggested that depression might be associated with insufficient activity of γ-aminobutyric acidergic (GABAergic) neurons which was the main condition for gamma oscillations generation [24‐26]. Additionally, human and animal studies reveal that gamma oscillations show a significant reduction in Alzheimer’s disease which is characterized by memory decline [27, 28]. Furthermore, both high and low gamma oscillations were found to be the key to the successful execution of working memory [29, 30]. We hypothesized that depressive individuals may have impaired GABAergic network inhibition mechanisms resulting in gamma oscillations reduction and working memory deficits.

Previous studies demonstrate an increase in both high and low gamma activity in healthy controls during verbal and non-verbal memory tasks [31‐33]. However, we found that the deficits of working memory were inconsistent, and only SWM performed abnormally in both UD and BP II patients. Previous researches indicated that negative emotions, such as depression and anxiety disrupted the performance of SWM selectively [34, 35]. From the work of Baddeley and Hitch, they proposed a multi-component model of working memory, the central executive, visuospatial sketchpad, episodic buffer, and phonological loop, and each component had different processing methods to cope with corresponding stimulation [36]. Verbal and spatial location information is processed by different components of working memory. Our tasks of working memory used in this study mainly involve the updating function of the central executive, phonological loop, and visuospatial sketchpad. The phonological loop is equivalent to the buffer which is mainly responsible for the storage and processing of verbal information, furthermore maintains memory accomplished by repetition [37]. Visuospatial sketchpad processes location information by spatial selective attention, which will consume more resources of exogenous attention [38, 39]. Therefore, our study suggested that there might be different components of working memory that was impaired varying degrees in patients with depression, resulting in different performance in two working memory tasks. Besides, depressive patients always have a tendency of rumination which makes them involve negative thoughts deeply [40]. For this reason, when the patients were in depression, cognitive resources were occupied by negative moods and rumination, which made it difficult to put resources into external stimuli. Hence the decline in SWM that requires higher cognitive resources is more obvious. Conversely, based on attentional control theory, we considered that the reason why VWM was complete might be compensatory [41]. When interfered with by emotional factors, individuals will invest more cognitive processing resources to ensure the successful completion of the main task and the same level of homework. In other words, negative emotions are likely to specifically weaken an individual’s ability to work and remember, but this effect is masked by compensatory. When interfered with by emotional factors, individuals would invest more cognitive processing resources to ensure the successful completion of the task. In other words, negative moods might reduce individuals’ accuracy of VWM, but concealed by compensatory.

Next, we will further discuss the neurophysiological mechanisms of working memory deficits in patients. More recently studies provided evidence that working memory impairments appeared to emerge altered gamma oscillatory activity in the frontoparietal regions [42‐45]. Furthermore, there was a study that systematically examined markers of GABA signaling across cortical nodes responsible for SWM. SWM is carried out by communication across a distributed cortical network including the primary visual cortex, visual association cortex, posterior parietal cortex, and DLPFC [46]. Inhibition control was considered to be the basic component of the central execution system of working memory. The decline in Inhibition Theory of Hashcr and Zacks believed that the cause of the decreased working memory in the elderly was actually the declining inhibition, rather than the reduction of working memory capacity. Subsequent studies provided evidence to support this theory, moreover found that inhibition was highly correlated with the frontal lobe. Therefore frontal lobe lesions would lead to failed inhibition, then leading to cognitive deficits [47, 48]. However, Redick et al. [49] proposed the opposite theory that the size of working memory capacity could affect the efficiency of inhibition. Both MFG and IFG were considered important components for inhibition and preSMA was also closely related to inhibition. Some findings suggested that bilateral IFG, right MFG, and parietal lobe were co-activated regions when subjects performed inhibition and working memory tasks [50]. The degree of activation in IFG was related to the rate of completion of inhibitive tasks [51]. Whereas Smith and Jonide [52] speculated that the preSMA was responsible for the retelling of spatial information, and activation of preSMA was also be found during SWM performance in the fMRI study. For non-verbal working memory, preSMA exhibited high aggregation with the DLPFC [23]. In short, activation of the prefrontal lobe was abnormally reduced when completing the n-back task [53]. For IPL, as Alain noted, the IPL activation overlapped with that observed during the auditory SWM task [54]. Some studies indicated altered FC of IPL was associated with lower levels of happiness or depression and rumination [55, 56]. IPL was found to play an important role in selective attention, WM rehearsal, and capacity [23, 57, 58]. Apart from this, other results highlighted the role of the IPL in modulating frontal lobe attention network activity [59]. Previously, a resting-state fMRI research in patients with depression showed a lower FC between DLPFC and bilateral posterior parietal cortex, which might directly affect cognitive control function [15]. Hence, we speculated that impaired top-down regulation from the prefrontal lobe might cause damage to the parietal lobe and alteration of the parietal lobe results in SWM deficit.

Our results suggested the patterns of FCs within these two groups of patients were different even performing similarly during working memory tasks. This may be due to the GABA levels evolved differentially between patients with unipolar and bipolar depression. A meta-analysis provides evidence that plasma GABA changes in those with unipolar depression were associated with symptomatic states, whereas plasma GABA changes in those with bipolar disorder seemed to be more closely associated with medication use [60]. Therefore, we believe that the gamma oscillation activity of UD patients would be more affected when they are also in a depressive episode. Additionally, our results also indicated that the FCs of IPL may be a key component to distinguish between BP II and UD. Some brain structural and functional studies have reached similar conclusions. The average kurtosis of gray matter in the IPL could help distinguish bipolar disorder from UD with high accuracy [61, 62]. More recently fMRI study found that UD showed increased fALFF values in the IPL compared to bipolar disorder [63]. Our study provides new evidence to support this conclusion.

In summary, since the stimulus was presented in the same way, the MSI reflected the endogenous processing of the subjects on both tasks. We demonstrated that only SWM was impaired in both UD and BP II patients. Furthermore, we suggested that the patterns of FCs within these two groups of patients were different even performing similarly during working memory tasks, and IPL may be a key component to distinguish between BP II and UD. Although there were some limitations in our study. Bipolar disorder patients were characterized by increased FC within regions of the FPN [64], which was inconsistent with our results. There was a higher FC strength but not significant between BP II and HC. The reason might be that the size of the participants was too small caused by limited conditions and long duration MSI. In further study, we will include more participants.