Dissociation of cortical regions modulated by both working memory load and sleep deprivation and by sleep deprivation alone

Abstract

Working memory is an important mental capacity that is compromised following sleep deprivation (SD). To understand how working memory load interacts with state to influence brain activation in load-sensitive regions, and the extent to which SD-related changes are common across different loads, we used fMRI to study twelve healthy subjects following 24 h of SD using a verbal n-back task with three load levels. Performance decline was observed by way of reduced accuracy and slower response times following SD. The left prefrontal region and thalamus showed load dependent activity modulation that interacted with state. The right parietal and anterior medial frontal regions showed load dependent changes in activity as well as an effect of state. The anterior cingulate and occipital regions showed activation that displayed state effects that were independent of working memory load. These findings represent a step toward identifying how different brain regions exhibit varying vulnerability to the deleterious effects of SD on working memory.

Introduction

Sleep deprivation (SD) is an important facet of modern life that is inescapable for many. A recent US National Sleep Foundation study found that 40% or fewer adults get at least 7 h of sleep on weekday nights. SD impairs performance in a number of cognitive tasks, but the extent to which it does so varies according to the cognitive domain tested. Tasks that engage the frontal lobes are well studied and among those that are vulnerable to SD (Jones and Harrison, 2001). The vulnerability of the frontal lobes to SD is further supported by the finding that predominantly frontal EEG changes occur after SD (Cajochen et al., 1999). Given that the frontal lobes are crucial for working memory (WM), it is not surprising that SD has an adverse effect on working memory.

Working memory is important to study because it underpins many higher cognitive processes. However, working memory itself is not a unitary entity and may be studied in terms of stimulus modality (verbal, non-verbal), temporally-defined components (encoding, delay period and probe-cued retrieval), process-defined components (maintenance vs. manipulation), and item load. Understanding how SD affects these elements is pivotal to helping us identify those that are particularly vulnerable to SD and should be avoided when accurate performance is critical. Furthermore, understanding the neural substrate underlying these vulnerable elements might provide us with ways to temporarily ameliorate the effects of SD using behavioral or pharmacologic approaches. In order to fulfill this goal, it is necessary to systematically evaluate the effect of SD on each element. Using this approach is supported by the divergent results of neuroimaging studies of task performance following SD. These studies illustrate that activation patterns following SD are exquisitely sensitive to task (Bell-McGinty et al., 2004, Drummond and Brown, 2001, Drummond et al., 2004).

Existing studies of WM during SD show that in some instances, decrements in performance may be correlated with reduced frontal activation (Drummond et al., 1999, Thomas et al., 2000) whereas in other instances performance is somewhat indifferent to changes in frontal activation (Habeck et al., 2004). We recently demonstrated that in accordance to prior behavioral studies, following SD, greater task complexity may paradoxically result in better preserved performance (Chee and Choo, 2004). This may in turn be associated with greater prefrontal and thalamic activation. In these experiments, the more complex task required additional manipulation of items held in WM while the less complex task required maintenance of items in WM.

In studies that did not involve SD, increasing task complexity by increasing task load generally results in increasing prefrontal activation (Braver et al., 1997, Callicott et al., 1999). At higher task loads, however, prefrontal activation has been found to increase monotonically (Braver et al., 1997) or peak and then decline in an inverted-U shape manner (Callicott et al., 1999). A decline in PFC activation at high WM loads has been interpreted as supporting evidence for a capacity constrained WM system (Callicott et al., 1999).

At present, it remains unclear how increasing difficulty in the context of a WM task interacts with SD to influence brain activation. In a recent experiment utilizing one-, three-, or six-letter arrays in a delayed-match-to-sample task (Habeck et al., 2004), the same level of reduced PFC activation was observed across all three load levels. In contrast, another study that involved WM manipulation and engagement of semantic processing, increasing task difficulty resulted in increased activation in several brain regions following SD including the left frontal and inferior parietal regions (Drummond et al., 2004). Given these conflicting results, we sought to clarify the manner in which task load interacts with state to influence brain activation as working memory is engaged following SD.

While the functional anatomy of the frontal lobes is the focus of many studies, the extent to which SD modulates cortical activation in other regions is also of interest. Some of these changes appear to generalize across different tasks and may be conceived as task independent but state dependent effects. For example, reduction in parieto-occipital activation with SD has been observed across several different tasks (Bell-McGinty et al., 2004, Chee and Choo, 2004, Drummond et al., 1999, Drummond et al., 2001, Habeck et al., 2004). Activity in the anterior cingulate has been found to increase following SD in a variety of tasks (Drummond and Brown, 2001, Drummond et al., 2004, Habeck et al., 2004). Reduced deactivation in the anterior medial frontal regions has been reported in two working memory tasks following SD (Chee and Choo, 2004). Although hinted at by Habeck et al. (2004), the generalizability of these findings merits further evaluation.

In light of past findings, we set ourselves two goals in the present study: to find a set of load-sensitive regions where load and state interact to modulate brain activation, and to find regions that show state-dependent effects across all loads. To accomplish these objectives, we scanned subjects using fMRI during rested wakefulness (RW) and following SD in a counterbalanced block design experiment. WM load was varied using a verbal n-back paradigm. We expected to find (WM) load-sensitive regions in the frontal and parietal lobes during RW (Braver et al., 1997, Callicott et al., 1999). Further, we hypothesized that in the presence of a capacity constrained WM system the detrimental effect of SD would be more evident at higher loads and that this would manifest as a reduction in activation during SD relative to RW at the same WM load. Lastly, we predicted that regions whose activation is influenced by SD but not by task load would lie in the occipital, anterior cingulate, and anterior medial frontal regions and that our results might extend those obtained recently (Habeck et al., 2004).