Parallel hippocampal-parietal circuits for self- and goal-oriented processing

Abstract

The hippocampus is critically important for a diverse range of cognitive processes, such as episodic memory, prospective memory, affective processing, and spatial navigation. Using individual-specific precision functional mapping of resting-state functional MRI data, we found the anterior hippocampus (head and body) to be preferentially functionally connected to the default mode network (DMN), as expected. The hippocampal tail, however, was strongly preferentially functionally connected to the parietal memory network (PMN), which supports goal-oriented cognition and stimulus recognition. This anterior-posterior dichotomy of resting-state functional connectivity was well-matched by differences in task deactivations and anatomical segmentations of the hippocampus. Task deactivations were localized to the hippocampal head and body (DMN), relatively sparing the tail (PMN). The functional dichotomization of the hippocampus into anterior DMN-connected and posterior PMN-connected parcels suggests parallel but distinct circuits between the hippocampus and medial parietal cortex for self- versus goal-oriented processing.

Keywords: brain networks; functional connectivity; hippocampus; individual variability; resting state.

Fig. 1.

Hippocampal parcellation using a winner-take-all approach. (A) Resting-state networks in cortex, shown for exemplar subject (MSC01). Refer to SI Appendix, Fig. S1 for individual-specific resting-state networks of all subjects. (B) Winner-take-all parcellation of the hippocampus based on its functional connectivity to all cortical networks (MSC01), demonstrating a DMN-anterior and PMN-posterior organization. Parasagittal slices are shown. Refer to SI Appendix, Fig. S2 for individual-specific hippocampal parcellations of all subjects. (C) Quantification of the relative resting-state network representation in the left and right hippocampus for each MSC subject (1–10) and the MSC Average (Top), using a winner-take-all approach including all networks. (D) A two-network (DMN, PMN) winner-take-all approach. Refer to SI Appendix, Fig. S5 for the two-network winner-take-all parcellations for all subjects.

Fig. 2.

DMN and PMN parcels’ functional connectivity to cortical networks. Displayed is the mean RSFC, Fisher z-transformed correlations z(r), to the cortical DMN (red) and PMN (blue) for individual-specific winner-take-all–derived hippocampal DMN and PMN parcels. Black circles indicate the null distribution, generated from hippocampal winner-take-all parcellations between all possible network pairs; plotted are each generated parcels’ mean functional connectivity to its winner network. Although the null distribution for all participants is shown here, significance testing to demonstrate hippocampal parcels did not occur by chance and only occurred within subjects against the participant-specific null distribution. ***P < 0.001 for all subjects, n.s. P > 0.05.

Fig. 3.

Functional connectivity of individual-specific hippocampal DMN and PMN parcels. (A) Functional connectivity of right hippocampal DMN parcel to cortex for exemplar subject (MSC01, Top) and MSC average (Bottom). (B) Functional connectivity of right hippocampal PMN parcel to cortex for exemplar subject (MSC01, Top) and MSC average (Bottom). (C) Difference between cortical connectivity of DMN and PMN parcels (DMN–PMN), showing that hippocampal functional connectivity respects individual-specific network borders. Warm colors represent greater DMN connectivity, and cool colors represent greater PMN connectivity. Functional connectivity values are Fisher z-transformed z(r). Refer to SI Appendix, Fig. S10 for all subjects.

Fig. 4.

Hippocampal functional connectivity to DMN and PMN along the anterior–posterior axis. (A) Schematic of the hippocampus (MSC01) with the longitudinal (AP) axis drawn and the parcels outlined. For each subject, we determined each voxel’s position along the longitudinal axis in the hippocampus as well as its parcel identity. (B) Scatterplots depicting the pairwise differences in functional connectivity to DMN and PMN as a function of coordinate position along the longitudinal axis. The amount of variance in functional connectivity differences explained by the gradient (grad) and parcel (par) models are noted, suggesting equal variance explained and thereby both a gradient and parcel organization. Functional connectivity values z(r) are Fisher z-transformed.

Fig. 5.

Functional connectivity of anatomically defined hippocampal segments (head/body versus tail). (A) The hippocampus was split into two segments, head/body and tail, based on either anatomical landmarks or a percentage-based approach (exemplar subject MSC01 shown). Refer to SI Appendix, Fig. S12 for landmark and percentage-based anatomical hippocampal parcellations for all subjects, demonstrating a similar anatomical segmentation. (B) Functional connectivity seed map for anatomical head/body (Left), tail (Middle) parcels, and the difference between the two (Right), for MSC01 as an exemplar, recapitulating functional parcels. Functional connectivity values z(r) are Fisher z-transformed. Refer to SI Appendix, Fig. S13 for all subjects.

Fig. 6.

Task deactivations in DMN and PMN. (A) Whole-brain task-general fMRI signal decreases with RSFC defined network borders (DMN, red; PMN, blue) overlaid. Task deactivations (negative fMRI signal relative to baseline), are shown for all MSC subjects (right hemisphere, medial). Mean task activations (z-scores) in DMN and PMN were calculated for (B) all of cerebral cortex, (C) the medial parietal cortex, and (D) the hippocampus. On average, there were greater task-general activity decreases in the DMN compared to the PMN across the whole brain during the task state. ***P < 0.001; *P < 0.05

Fig. 7.

Schematic of parallel self- and goal-oriented circuits between the hippocampus and medial parietal cortex. Medial parietal cortex is the primary target of hippocampal functional connectivity, but connectivity is segregated by functional network. The bulk of the hippocampus (anterior) is functionally connected to the DMN (red) and CAN (pearl white) networks. The tail of the hippocampus is preferentially connected to the PMN (blue) and fFPN (yellow) networks. This functional connectivity dichotomy maps onto those parts of hippocampus and medial parietal cortex that deactivate during goal-oriented tasks (DMN and CAN) and those that do not (PMN and FPN). This functional organization suggests that human cognition can draw on two variants of hippocampal and medial parietal circuitry. The anterior circuit might support sequencing and navigating spacetime (68, 69) in the service of the self, while the posterior circuit might carry out very similar operations in the service-specific, goal-directed, attention-demanding tasks.