Longitudinal hippocampal axis in large-scale cortical systems underlying development and episodic memory

Abstract

The hippocampus is functionally specialized along its longitudinal axis with intricate interactions with cortical systems, which is crucial for understanding development and cognition. Using a well-established connectopic mapping technique on two large resting-state functional MRI datasets, we systematically quantified topographic organization of the hippocampal functional connectivity (hippocampal gradient) and its cortical interaction in developing brains. We revealed hippocampal functional hierarchy within the large-scale cortical brain systems, with the anterior hippocampus preferentially connected to an anterior temporal (AT) pathway and the posterior hippocampus embedded in a posterior medial (PM) pathway. We examined the developmental effects of the primary gradient and its whole-brain functional interaction. We observed increased functional specialization along the hippocampal long axis and found a general whole-brain connectivity shift from the posterior to the anterior hippocampus during development. Using phenotypic predictive modeling, we further delineated how the hippocampus is differentially integrated into the whole-brain cortical hierarchy underlying episodic memory and identified several key nodes within PM/AT systems. Our results highlight the importance of hippocampal gradient and its cortical interaction in development and for supporting episodic memory.

Keywords: development; episodic memory; functional connectivity; gradient; hippocampus.

Fig. 1.

(A) Primary hippocampal gradient for the left and right hippocampus in HCP-D and NKI datasets. (B) Comparing hippocampal gradient k-means clusters with subdivisions of Melbourne subcortical atlas, i.e., hippocampus head lateral and medial division, body, and tail. (C) Violin plots of the primary gradient distribution across four hippocampal subdivisions based on the Melbourne subcortical atlas, color-coded according to Melbourne subcortical atlas in panel B.

Fig. 2.

Projection maps associated with the primary hippocampal gradient for the left and right hippocampus in HCP-D (A) and NKI datasets (B).

Fig. 3.

(A) Seed-based functional connectivity along the hippocampal longitudinal gradient (seed regions are color-coded based on k-means clusters). For visualization purposes, only regions with the top 20% connectivity strength were shown. (B) Seed-based functional connectivity map as a function of seed location. Red reflects regions with increasing connectivity and blue for decreasing connectivity as the seed moves from the posterior to the anterior hippocampus.

Fig. 4.

Developmental analysis of the primary hippocampal longitudinal gradients with the HCP-D dataset. (A) Hippocampal age-related correlation maps for the left and right hippocampal gradient. (B) Age-related correlation maps of corresponding projection maps.

Fig. 5.

A graphical overview of the predictive framework. (A) Projection values from the primary projection maps were extracted using the CAB-NP whole-brain atlas. (B) The extracted mean projection matrix was entered into a LASSO model, significantly predicting the memory scores in the left-out validation set (r = 0.20, p_perm < 0.001). (C) The memory-predictive pattern for left hippocampal projection maps and the violin plot of model weights for each network across 10 × 10 CVs. Positive weights mean increased connectivity to the anterior hippocampus, while negative weights mean increased connectivity to the posterior hippocampus.