The hippocampus is a core region in the study and understanding of brain diseases, in particular, in Alzheimer’s related pathology. Previous studies have evidenced that the accumulation of amyloid-beta and tau pathologies in the hippocampus is associated with metabolic and structural changes of the hippocampus itself, but also of other brain regions in healthy aging and dementia[1, 2], hence suggesting a role of hippocampal networks in Alzheimer’s disease pathology. Understanding the neurobiological properties of the hippocampus is therefore a crucial step to understand pathophysiological processes in Alzheimer’s disease. In that perspective, it has been pointed out that the disruption of its’ white matter fiber bundles, the spatiotemporal patterns of alteration of metabolism in spatially remote brain regions and the pattern of macrostructural atrophy in the hippocampus in the early stages of Alzheimer’s disease were related events occurring as the neuropathology of Alzheimer’s disease spreads [3–7].
In view of the fact that the hippocampus shows heterogeneous patterns of microstructural features, as well as heterogeneous anatomical connectivity, the pathophysiological processes could be better understood by considering different subregions rather than looking at the hippocampus as a whole. However, this perspective is currently limited by the fact that cytoarchitecture studies and structural and functional neuroimaging studies of the hippocampus resulted in divergent parcellation maps. Based on cyto- and receptor-architecture features, a pattern of differentiation within the hippocampus (Fig. 1A) can mainly be identified along the medio-lateral and ventro–dorsal axes [8, 9]. While structural covariance patterns to a large extent follow this differentiation, structural covariance also mainly changes along the longitudinal axis, that is, between anterior (head) and posterior (body-tail) regions (Fig. 1B) [10]. Furthermore, differentiation based on functional connectivity (functional coupling during a task or at rest) appears primarily along this longitudinal axis, a differentiation pattern that converges with the different behavioral involvement of the anterior and posterior hippocampal regions [11–14]. Thus, heterogeneity within the hippocampus has been evidenced both based on local microstructure, connectivity profiles, and behavioral engagement, but how this heterogeneity influences metabolic patterns remains an open question.
Our recent examination of hippocampal structural covariance networks in healthy populations, including older populations, revealed three hippocampal subregions with specific structural covariance networks (Fig. 1B) [15] the hippocampal head as part of an anterior network, the lateral posterior hippocampus (corresponding to the body and tail of Cornu Ammonis (CA) fields) as part of a subcortical network and the medial posterior hippocampus (corresponding to the body and tail of the subiculum) showing a very vast cortical structural covariance pattern. As illustrated in Fig. 1 (C and D), these structural covariance networks appear, to a great extent, to follow white matter tracts which connect the medial temporal lobe to subcortical and cortical regions suggesting that morphological covariance across the brain in a healthy population is, to a great extent, influenced by anatomical connection [16, 17]. Nevertheless and importantly, our previous study also suggested that in patients with dementia, the posterior regions differentiate together from the more anterior regions in their co-atrophy patterns suggesting that the complex pathophysiological process in Alzheimer’s disease would not result in atrophy patterns following specifically structural covariance networks seen in the healthy population (such as a specific alteration within the cortical network), but rather, would affect several subregions [15]. Because atrophy patterns that can be observed at the macroscale in dementia are the ultimate consequences of a long and complex pathological process, a first step towards understanding how this process finally leads to complex macrostructural atrophy patterns is to identify the metabolic changes that affect the hippocampal metabolic networks. However, a better understanding of pathological processes in Alzheimer’s disease first requires a robust identification and understanding of the hippocampal subregion and their metabolic networks in healthy older populations.
To address these questions in the current study, we first performed a hippocampus parcellation based on a molecular profile computed from 18FDG-PET in a large cohort of healthy elderly participants pooled from two independent databases. We hence computed a data-driven parcellation of the hippocampus based on co-metabolism profiles. Then, capitalizing on the robustly defined subregions, we characterized their metabolic networks. Furthermore, to help therapeutic actions and advance our understanding of behavioral phenotype in hippocampal disorders, we characterized the disentangled hippocampal metabolic networks with regards to behavioral and neurotransmitter systems using quantitative decoding. We finally examined how the local metabolism in the hippocampal subregions is influenced by Alzheimer’s disease pathology in a cohort of ADNI participants (n = 580).