Functional connectivity analysis conducted on the BOLD signal at rest and during a spatial working memory task enabled statistical assessment of functional brain networks in alcoholics and healthy subjects. We thresholded the correlational matrices to construct a set of unweighted binary graphs for each subject and weighted connectivity of each node of the network for which connectivity differed between the 2 groups.
At rest, the spontaneous slow fluctuations of the signal in the posterior cingulate and cerebellar regions in the alcoholics were less synchronized than in controls, indicative of compromised functional connectivity. To our knowledge, this is the first report of abnormalities in functional connectivity associated with the DMN in alcoholics. The DMN is defined as a set of functionally connected regions in the resting brain and is part of the default settings of the brain. The posterior cingulate cortex plays a key role in this network (Raichle et al. 2001
; Fransson and Marrelec 2008
), and the cerebellum has more recently been identified as a node in the default network (Habas et al. 2009
). In the healthy brain, modulation of the DMN plays a crucial role in normal functioning. Several reports have shown, for example, that weaker suppression of the DMN while engaged in a task is associated with inferior cognitive performance, for example, inferior memory formation (Daselaar et al. 2004
) and poor learning of cognitive skill (Mason et al. 2007
). Also, with increasing working memory demands, greater suppression of the DMN is required (McKiernan et al. 2003
). Thus, a failure in this suppression would compound a cognitive challenge (Whitfield-Gabrieli et al. 2009
). In the case of alcoholics, a deficit in DMN modulation between the posterior cingulate and cerebellar regions between rest and task could reflect processing inefficiency. Compromised connectivity detected in alcoholics represents a decoupling of synchronization between regions that are functionally synchronized in controls. Corrupted synchronization could result in disturbed connectivity, precluding direct transmission between nodes and requiring longer paths to reach nodes of the graph.
Task engagement elicited another pattern of group differences in cerebellar involvement. Alcoholics showed more robust connectivity than controls between the posterior cingulate cortex and the left anterior/medial lobules IV–V of the cerebellum. The group differences in connectivity cannot be a direct result of activation differences because the groups did not show such differences.
Brain networks other than the ones invoked by controls could thus be involved when alcoholics engage in a task. The addition of new direct functional connections may have led to greater connectivity and better than expected results in this sample of alcoholics who were able to reach the same level of performance as controls (cf. Chanraud, Pitel et al. 2010
). Such a mechanism, however, would involve additional connections within the network that would have been detected as an increase of the local efficiency in alcoholics. Yet, we detected an extra connection between the posterior cingulate cortex and a specific region within the cerebellum but outside of the DMN. With further probing, these results might reveal the presence of anatomical connections or functional synchronization in alcoholics not present in controls.
The presence of extra connections outside the DMN together with the observations of comparable performance of the controls and alcoholics led us to speculate that a compensatory system could operate through an alternate network, as also suggested by other studies (Pfefferbaum et al. 2001
; Desmond et al. 2003
; Chanraud-Guillermo et al. 2009
), and counteract compromise in another network (Caplan et al. 2007
). The anterior/medial cerebellar lobule IV/V complex, a region that was more correlated with the posterior cingulate region during the task in alcoholics, and thus a candidate for compensation, is usually involved in motor functions (Habas et al. 2009
; Marvel and Desmond 2010
). The alcoholics may have recruited the motor loop, potentially spared in alcoholism (Harper et al. 2003
), to compensate for the executive one, which is likely compromised in alcoholism (Harper et al. 2003
) and necessary for performing the spatial working memory task (Pfefferbaum et al. 2001
). Involvement of the motor network to compensate for inefficient functioning within a closed cerebrocerebellar loop, such as the executive network, comports with previous findings (Chanraud, Pitel et al. 2010
). Further to this point, the activity in the posterior cingulate cortex has been shown to be negatively correlated with the activity of nodes of the prefrontal-based motor control circuits, and its influence has been speculated to modulate the activity within this circuit involved in execution and observation of goal-directed actions (Uddin et al. 2009
Other cerebellar regions are also relevant for performance of the working memory task used. Local efficiency of the anterior cerebellar lobule III during the task was correlated with performance in both the controls and the alcoholics. While at rest, local efficiency in the inferior cerebellar lobule VIII, which is also a node of the DMN, was related to performance only of the alcoholics. Higher efficiency value in these regions also correlated with longer sobriety in alcoholics, suggesting that whatever the compensatory processes were used by alcoholics involving these regions, sustained abstinence was associated with better performance. A further implication is that longer sobriety provides extended time for recovery and repair, and this variability contributed to performance differences within the group of alcoholics. Similarly, variability of cognitive performance has been shown to correlate with path length across a combined group of control and patients with Alzheimer's disease (Stam and Reijneveld 2007
) and with network efficiency in groups of healthy subjects and schizophrenic patients (Rubinov et al. 2009
). Also, intellectual performance measured with IQ correlated with path length in healthy subjects (van den Heuvel et al. 2009). In heroin users, path length in the bilateral cerebellum is linked to duration of use, suggesting that length of consumption is influential in functional connectivity alteration (Yuan et al. 2010). In alcoholics, however, working memory performance was also positively related to local efficiency in a cerebellar node measured during the rest period (the inferior cerebellar lobule VIII). One interpretation is that this region is part of a network of compensation that facilitates cognitive performance. An alternative interpretation derives from the between-groups difference at rest. Specifically, alcoholics had less connectivity than controls between the posterior cingulate cortex and the inferior cerebellar lobule VIII, the efficiency of which correlated with working memory performance and duration of abstinence in alcoholics. One interpretation of these results suggests that this region needs to be available, even at rest, for alcoholics to meet high task demands. This hypothesis is further corroborated by the significant relationship between the efficiency in the cerebellar region and the strength of functional connectivity between the posterior cingulate cortex and the inferior cerebellar lobule VIII during the task.
Anatomically, cerebrocerebellar circuits are based only on indirect projections through the thalamus and pontine nucleus (Kelly and Strick 2003
). Functional connectivity reveals contralateral cerebellar correlations with cortex, consistent with polysynaptic connectivity of the cerebellum (Allen et al. 2005
). Thus, unlike imaging analyses that use anatomy directly (Sporns et al. 2007
; Hagmann et al. 2008
), connectivity patterns defined herein could reflect both direct and indirect anatomical projections. Functionally, regions of the cerebellum are active during rest and task engagement, and patterns of BOLD activity can be modified by the order of experimental blocks, that is, whether rest follows task or task follows rest. Evidence for such modulation derives from application of a probabilistic independent components analysis (Albert et al. 2009
), which identified 2 distinct resting-state networks, one of which was a cerebellar network that was modulated by learning as the network strength increased following a novel motor learning task but not following a simple motor performance task. Because the rest block followed the task block in this experiment, the modulatory process of the cerebellum may be disrupted in alcoholics and potentially underlie their spatial working memory impairment.
A principal assumption of the statistical approach taken by graph theory to model processing is that short path lengths between neural nodes can help preserve transmitted signals from decay with time and degradation over multilocal transmission. Events and conditions that disrupt short paths can impair connectivity. With respect to the untoward effects of alcoholism on the DMN, even if the principal network nodes are themselves not pathological, subtle disruption of their functional connections could result in inefficiency of information transmission and manifest as impairment (cf., Sullivan and Pfefferbaum 2005
). We conclude that the results obtained in this study comport with pathophysiological models of alcoholism as a functional disconnection syndrome (Geschwind 1965
; Chanraud, Zahr et al. 2010
) marked by processing inefficiency arising from compromised connectivity of required functional networks.