Down but not out in posterior cingulate cortex: Deactivation yet functional coupling with prefrontal cortex during demanding semantic cognition

Abstract

The posterior cingulate cortex (pCC) often deactivates during complex tasks, and at rest is often only weakly correlated with regions that play a general role in the control of cognition. These observations led to the hypothesis that pCC contributes to automatic aspects of memory retrieval and cognition. Recent work, however, has suggested that the pCC may support both automatic and controlled forms of memory processing and may do so by changing its communication with regions that are important in the control of cognition across multiple domains. The current study examined these alternative views by characterising the functional coupling of the pCC in easy semantic decisions (based on strong global associations) and in harder semantic tasks (matching words on the basis of specific non-dominant features). Increasingly difficult semantic decisions led to the expected pattern of deactivation in the pCC; however, psychophysiological interaction analysis revealed that, under these conditions, the pCC exhibited greater connectivity with dorsolateral prefrontal cortex (PFC), relative to both easier semantic decisions and to a period of rest. In a second experiment using different participants, we found that functional coupling at rest between the pCC and the same region of dorsolateral PFC was stronger for participants who were more efficient at semantic tasks when assessed in a subsequent laboratory session. Thus, although overall levels of activity in the pCC are reduced during external tasks, this region may show greater coupling with executive control regions when information is retrieved from memory in a goal-directed manner.

Keywords: Connectivity; Dorsolateral prefrontal cortex; Executive; Posterior cingulate cortex; Rest; Semantic control.

Fig. 1

Left-hand column shows overlap (yellow) of the pCC subdivisions with a semantic meta-analytic map (green) derived from Neurosynth (using ‘semantic’ as a search term). (A): Positive (blue) and negative (red) functional connectivity of pCC at rest (cluster correction, Z > 2.3, p < 0.05), and the overlap of these positive and negative networks with (B) semantic control (Noonan et al., 2013) and (C) the multiple-demand network (MDN; Duncan, 2010) shown in yellow (overlap of low connectivity at rest and semantic control/MDN) and cyan (overlap of high connectivity at rest and semantic control/MDN). Maps in Panels B & C are displayed with a fully saturated colour map to maximise the visibility of the regions of overlap.

Fig. 2

(A): task conditions for experiments 1 and 2, the target item is highlighted (green box). (B): efficiency scores (RT/ACC) for experiments 1 and 2, in milliseconds. (C): ROI analysis of percent signal change in pCC (mask shown in red). Error bars indicate standard error of the mean for both behavioural and ROI data (**p < 0.01, ***p < 0.001).

Fig. 3

Whole brain contrasts of feature selection > strong association for functional activation (A) and PPI of pCC (B; cluster correction, Z > 3.1, p < 0.005). Overlap with Duncan's (2010) multiple-demand network is shown in cyan for the functional activation and in yellow for the PPI. These maps are displayed with a fully saturated colour map to maximise the visibility of the areas of overlap.

Fig. 4

Task-based pCC functional connectivity masked by the multiple-demand network (Duncan, 2010). Overlap (white) of (i) pCC task-based functional connectivity for contrasts of feature selection > strong association (green) and feature > baseline (red) and (ii) lower resting state connectivity of pCC (blue). The grey panel displays the overlap of the pCC PPI contrasts and lower resting state connectivity revealed two clusters: one in dorsolateral prefrontal cortex (dlPFC) and one in pre-supplementary motor area (preSMA). These maps are displayed with a fully saturated colour map to maximise the visibility of the overlaps.

Fig. 5

Multiple regression result of resting state functional connectivity and response efficiency on semantic tasks (cluster correction, Z > 2.3, p < 0.05; search space restricted by a mask created using the final result of experiment 1: PPI masked by MDN contrast of feature > strong; blue). The correlation of performance and resting state connectivity is shown in the right hand scatterplots. A high score on the x-axis indicates better task performance.

Fig. 6

Higher (red) and lower (blue) resting state connectivity of dorsolateral prefrontal cortex (seed region: overlap of Experiments 1 and 2), and the corresponding terms derived from Neurosynth for these maps. The grey panel displays the positive connectivity map for this region overlaps with parts of the multiple-demand network (Duncan, 2010; yellow). In this panel the maps are displayed with a fully saturated colour map to maximise the visibility of the overlaps.

Fig. 7

Comparison of the patterns of functional connectivity in different cohorts used in this study. The upper panel illustrates the positive and negative resting state connectivity for the pCC region studied in three cohorts of participants used in this experiment. The lower panel presents the connectivity of the dorso-lateral prefrontal seed from Cohort 4. All maps in this figure are displayed with a fully saturated colour map.