Functional connectivity in the resting brain: A network analysis of the default mode hypothesis

Results

Resting-State Connectivity Maps.

PCC.

The PCC showed significant resting-state connectivity with the following regions: a large cluster of MPFC, including vACC and orbitofrontal cortex (OFC); left DLPFC; IPC bilaterally; left inferolateral temporal cortex (ITC); and left parahippocampal gyrus (PHG) (Fig. 1, Table 1).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Map of the resting-state neural connectivity for the PCC. The blue arrow indicates the approximate location of the PCC peak [−2 −51 27]. The approximate locations of the eight significant clusters are labeled A–H in descending order of the cluster's t score (A corresponds to the cluster with the highest t score). A [−51 −65 27], left IPC. B has a maximum at [−2 55 −18] in the OFC, but extends superiorly into the MPFC and the vACC (seen at z = +2). C [53 −61 27], right IPC. D [−16 49 38], MPFC just left of midline. E [−44 20 41], left DLPFC. F [−12 −35 0], posterior left PHG. G [18 54 32], MPFC just right of midline. H [−58 −18 −14], left ITC. Map is superimposed on transverse sections of the group-average structural scans. The numbers below each image refer to the z plane coordinates of Talairach and Tournoux. The left hemisphere of the brain corresponds to the left side of the image. Height and extent thresholds were set at P < 0.001. t score scale is shown on the right.

View this table:

• View inline
• View popup

Table 1

Functional connectivity of the PCC during the resting state

vACC.

The vACC showed significant resting-state connectivity with the following regions: PCC, MPFC/OFC, nucleus accumbens, and hypothalamus/midbrain (Fig. 2, Table 2).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Map of the resting-state neural connectivity for the vACC. The blue arrow indicates the approximate location of the vACC maximum [2 38 −2]. A [2 −51 27], PCC extending into the precuneus. B [4 −14 34], rostral PCC. C [4 9 −6], nucleus accumbens. D [4 −16 −3] is in the hypothalamus with some extension into the rostral midbrain. For other details, see Fig. 1.

View this table:

• View inline
• View popup

Table 2

Functional connectivity of the vACC during the resting state

Visual Processing Task Connectivity Maps.

The connectivity patterns for the PCC and vACC during the visual processing task were virtually identical to those obtained from the resting-state data. Fig. 3 shows the connectivity patterns acquired for the PCC during the visual processing task compared with those acquired during the resting state. A similar figure (Fig. 5) comparing the vACC connectivity maps during the resting state and the visual processing task is published as supporting information on the PNAS web site. The precise locations for the cluster maxima for the PCC and vACC connectivity maps during the visual processing task are also published as Tables 6 and 7 in the supporting information on the PNAS web site.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Comparison of the PCC connectivity patterns during the visual processing task (Upper) and the resting-state (Lower). The blue arrows indicate the approximate location of the PCC peak [−2 −51 27]. For other details, see Fig. 1.

Prefrontal Regions and Inverse Correlations.

Fig. 4 shows the statistical maps for regions inversely correlated with the left VLPFC, right VLPFC, and right DLPFC during the resting state. In each case, only one cluster had resting-state activity that was significantly inversely correlated with the prefrontal ROI activity, and in each case the cluster was centered in the PCC.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

Connectivity maps showing regions inversely correlated with the left VLPFC, the right VLPFC, and the right DLPFC. In each case, the only significant cluster was in the PCC. Maxima were at [−4 −55 25] for the left VLPFC, [0 −49 30] for the right VLPFC, and [−4 −49 26] for the right DLPFC. Height threshold was P < 0.01; extent threshold was P < 0.05. For other details, see Fig. 1.

Discussion

This is the first study, to our knowledge, to demonstrate resting-state functional connectivity between brain regions implicated in the default mode network (1, 3). The vACC showed significant functional connectivity with the PCC and the MPFC (Fig. 2), two regions implicated in the default mode network. The PCC connectivity map (Fig. 1), however, provided a far more complete match with the hypothesized default mode network. Of the eight regions cited by Shulman et al. (1) in their metaanalysis of positron emission tomography “deactivations,” our PCC connectivity map includes all but the right amygdala and the left lateral inferior frontal cortex. Furthermore, only one of the regions (the left PHG) in the PCC connectivity map was not identified in the metaanalysis. This extensive overlap between our resting-state network analysis and the comparative approaches (baseline vs. experimental) combined in the metaanalysis suggests that the resting-state network identified here accounts, in large part, for the phenomenon of task-related decreases in brain activity. Taken together, our results demonstrate that a distinct set of brain regions, whose activity decreases during cognitive tasks compared with baseline states, shows significant functional connectivity during the resting state, thus providing the most compelling evidence to date for the existence of a cohesive, tonically active, default mode network.

Although the default mode network is thought to be most active during the resting state, it may also persist during passive sensory processing states (1, 2). To explore this possibility, we generated connectivity maps for the PCC and the vACC during our visual processing task and compared them to the resting-state maps. The PCC and vACC maps were virtually identical in the resting state and the visual processing task (Fig. 3 and supporting information on the PNAS web site), suggesting that the default mode neural network is minimally disrupted by sensory processing tasks with limited cognitive demand. It is important to note, in this regard, that the PCC and vACC were not “deactivated” in the visual processing task, further suggesting similar levels of ongoing activity in the default mode network during the flashing and still checkerboard epochs. This finding has important implications for the design and interpretation of functional imaging studies (4, 18). In particular, investigators wishing to study the functions subserved by regions such as the PCC or vACC should consider contrasting their experimental task with an equally demanding baseline task. Furthermore, because the default mode network was not significantly different between an eyes-closed and an eyes-open state, we may be able to deduce what cognitive processes the network does or does not support. For example, the default mode network appears to be distinct from the neural network that generates the alpha rhythm in resting electroencephalograms, given that the latter reproducibly attenuates when a subject opens their eyes (19).

Although the default mode network appears to persist during sensory tasks with low cognitive demand, it has been suggested that externally cued tasks with high cognitive demand do modulate the network, contributing to the phenomenon of task-related decreases in brain activity (4). If this network is suspended during performance of such tasks, we reasoned that during the resting state, the default mode network and brain regions that show task-related increases in brain activity may be inversely correlated. This hypothesis was strongly supported by the finding of significant inverse correlations between all three “activated” lateral prefrontal ROIs and the PCC during rest (Fig. 4). In this regard, it should be noted that Hampson et al. (10) also detected a negative correlation between activity in Broca's area and activity in the PCC during a passive listening task. These findings suggest the presence, during rest, of an inhibitory interaction between activated regions and regions in the default mode network, particularly the PCC. This raises the possibility that activated prefrontal regions, when called to duty, directly suppress the default mode network (20). Alternatively, it is possible that these regions are maintained in antagonistic states by a third region. The thalamus, for example, is one such subcortical region that has been shown to play an intermediary role in cortico–cortical interactions (21).

We will now consider the question of what mental processes the default mode network might subserve. Several findings from our study suggest that the PCC plays a central role in the default mode network. First, of the regions that showed task-related decreases during working memory and that also overlapped with regions in the default mode network (outlined by Raichle et al. in ref. 3), the PCC had the second highest z score. Second, compared with the vACC connectivity map, the PCC connectivity map provided a better match with this default mode network (accounting for six of eight regions cited in the initial metaanalysis) (1). Third, in the resting state, the PCC was the only region in the brain to show inverse correlations with any of the “activated” prefrontal ROIs. Finally, the PCC appeared not only in one of these inverse correlation maps, but in each of the three. The attribution of a central role to the PCC is consistent with data showing that the PCC is one of the most metabolically active regions during rest (3), that its activity decreases as a function of waning consciousness during sedation (22), and that its connections to prefrontal regions are impaired in the persistent vegetative state (23).

In view of these observations, understanding the cognitive functions of the PCC may yield insight into the functions supported by the default mode network. Unfortunately, imaging studies have revealed relatively little about the functions of the PCC, presumably because most paradigms have used a baseline state (such as fixation or sensory processing) in which the PCC is already active. Recently, however, a handful of imaging studies using more complex baseline states have implicated the PCC in retrieval of episodic memories (24–27). In this regard, the study by Cabeza et al. (27) is particularly instructive, in that the PCC did not show increased activity when an episodic memory retrieval task was compared with a baseline of visual fixation; however, the PCC did show increased activity when the retrieval task was compared with a more complex baseline. That the PCC may be critical to the retrieval of episodic memories is further supported by the fact that it is consistently among the earliest brain regions to show decreased metabolism in Alzheimer's disease (28, 29). Finally, axonal tracing studies (30, 31) have demonstrated neural connectivity between the PCC and medial temporal lobe regions, such as the entorhinal cortex and PHG, known to be key memory centers across species.

It would seem, therefore, that if the PCC is a critical node in the default mode network, this network should be integrally involved with episodic retrieval. Indeed, one prevailing view is that the default mode or “conscious resting state” involves retrieval and manipulation of past events, both personal and general, in an effort to solve problems and develop future plans (2, 5). In this framework, the left DLPFC is commonly implicated in working memory (32) as well as retrieval of episodic memories (33). Similarly, the bilateral IPC are frequently activated during working memory paradigms in both humans and animals (34, 35). The left ITC is known to be a major repository for semantic knowledge (36). The left PHG is associated with memory in humans (37), and its connections with the PCC may contribute to memory processes in animals (38). Perhaps the most substantial corroborative evidence to support our claim that the default mode network participates in retrieval of information, both personal and general, stems from the study by Maguire and Mummery (24). In that study, the authors compared a combination of four types of retrieval (personal memories, personal facts; nonpersonal memories, nonpersonal facts) with a relatively complex baseline (attending to syllable count while listening to words) and showed increased activity in a network that overlaps considerably with the default mode network outlined here. In particular, increased activity was detected in the PCC, the bilateral IPC, the MPFC, the left PHG, and the left ITC (ibid, Fig. 1).

Thus, the regions linked to the PCC in the default mode network are well-suited to support the mental processes presumed to be ongoing during the resting state. In contrast to the PCC and its connections to higher cortical regions, the vACC is linked primarily to paralimbic and subcortical regions associated with affective and autonomic processes including OFC, nucleus accumbens, and hypothalamus/midbrain (39, 40). The distinct roles demonstrated for PCC and vACC in our study converge on a model, derived from functional imaging studies of depression, positing that the vACC links a ventral “vegetative–somatic” compartment with a dorsal “attention–cognition” compartment that includes the PCC (41). Further support for this model is found in the overlap of the PCC and vACC patterns. The OFC and the MPFC, which have been implicated in the integration of cognitive and emotional stimuli (39, 42, 43) and also project to subcortical regions including nucleus accumbens and hypothalamus (44, 45), were the only regions common to both the PCC and vACC connectivity patterns. Thus, robust functional connections between PCC and vACC appear to form the critical link in the network between higher cortical (presumably conscious) processing and more basic (possibly subconscious) processing necessary for calibrating affective and autonomic states.

In summary, we have provided compelling evidence for the existence of a default mode network in the resting state. We have also shown that this network is minimally disrupted during a visual processing task and suggested that it is maintained in a dynamic equilibrium with lateral prefrontal regions that commonly show task-related increases in activity. Although the precise mental processes supported by the default mode network remain to be elucidated, we suggest that the retrieval and manipulation of episodic memories and semantic knowledge are likely candidates. A clearer understanding of the default mode network will be critical in the proper interpretation of task-activation studies that involve comparing an experimental task with a resting or simple-processing baseline. Understanding this network should also have broader implications. As interest builds in defining different components of consciousness and their neural correlates (46), several features of the default mode network, that it is widely distributed, internally cued, appears to integrate cognitive and emotional processing, and can be experimentally modulated, should make it an important and rewarding area of study in this nascent field.