Transcriptomic gradient of the human hippocampus: A vertex-wise atlas of post-mortem gene expression

885 

Abstract Submission 

Alexander Ngo¹, Jordan DeKraker¹, Sara Larivière², Lang Liu³, Jessica Royer¹, Raúl Rodriguez-Cruces¹, Jacob Vogel⁴, Ziv Gan-Or³, Alan Evans¹, Boris Bernhardt¹

¹Montreal Neurological Institute and Hospital, Montreal, QC, ²Brigham and Women’s Hospital, Boston, MA, ³McGill University, Montreal, QC, ⁴Lund University, Lund, n/a

Alexander Ngo  
Montreal Neurological Institute and Hospital
Montreal, QC

Jordan DeKraker  
Montreal Neurological Institute and Hospital
Montreal, QC

Sara Larivière  
Brigham and Women’s Hospital
Boston, MA

Lang Liu  
McGill University
Montreal, QC

Jessica Royer  
Montreal Neurological Institute and Hospital
Montreal, QC

Raúl Rodriguez-Cruces  
Montreal Neurological Institute and Hospital
Montreal, QC

Jacob Vogel, PhD  
Lund University
Lund, n/a

Ziv Gan-Or  
McGill University
Montreal, QC

Alan Evans  
Montreal Neurological Institute and Hospital
Montreal, QC

Boris Bernhardt  
Montreal Neurological Institute and Hospital
Montreal, QC

The hippocampus is involved in multiple aspects of brain function and dysfunction. Unravelling its complex organization requires the integration of multiscale data, linking molecular features to macroscale hierarchies. Gene expression is a fundamental molecular phenotype, and its profiling can provide a reference description of how microstructural features are distributed across the brain. Post-mortem gene expression samples, however, are often spatially discontinuous and biased towards coarse brain parcellations, thus potentially overlooking fine-grained information. Here, we charted gene expression patterns within the hippocampus with unprecedented resolution, providing a unified atlas of the hippocampal transcriptome and relating our findings to its functional and structural hierarchies.

Allen Human Brain Atlas. We used the structural T1w magnetic resonance imaging (MRI) and microarray expression data of six deceased human donors (five males, mean ± SD age = 42.5 ± 13.4 years) from the Allen Human Brain Atlas-a brain-wide atlas comprised of bulk transcriptomic measures from over 20,000 genes sampled across 3,702 spatially distinct tissue samples [1].

Vertex-wise mapping of hippocampal gene expression. Donor-specific hippocampal surfaces were generated from individual structural scans using HippUnfold-an automated pipeline for hippocampal unfolding, subfield segmentation, and novel surface-based hippocampal registration [2, 3]. In parallel, we preprocessed the microarray expression data through intensity-based filtering of microarray probes, selection of a representative probe for each gene across both hemispheres, normalization, and aggregation across donors [4]. Tissue sampled within the hippocampus (n=125) were mapped to subject-specific hippocampal surfaces. We interpolated expression values across the hippocampus, weighted by the geodesic distance of a given vertex to its nearest sampled neighbour. Continuous donor-specific transcriptomic maps were averaged to generate a single expression map for each gene (Fig 1A).

Code and data availability. All code to preprocess imaging, microarray data, and hippocampal transcriptomic maps will be made available as part of the HippoMaps-a toolbox for multiscale and multimodal contextualization of the hippocampus (https://github.com/MICA-MNI/hippomaps).

We generated an atlas of vertex-wise maps of hippocampal expression for 13,561 genes. Dimensionality reduction using principal component analysis on the concatenated expression maps of all genes identified a main transcriptomic axis that explained 17.5% of the variance and differentiated anterior from posterior regions (Fig 1B), as previously observed in volumetric analyses [5]. Gene ontology enrichment analysis of the most influential genes (top 5% of PC1 loadings) revealed a consistent set of biological processes related to synapse organization, axonal growth, development, and behaviour (all pFDR < 0.05; Fig 2A). We further generated histology- and MRI-derived profiles of the hippocampus, revealing its structural and functional organization, and assessed their spatial correlation to the genetic gradient identified herein using autocorrelation preserving null models [6]. These multimodal comparisons revealed significant and specific spatial association between the predominant gene expression and the resting-state functional gradients (r=0.88, pspin<0.001; Fig 2B).

Capitalizing on recent imaging-transcriptomic initiatives, we generated vertex-wise maps of hippocampal gene expression from six post-mortem human brains, that mainly followed an anterior-posterior gradient. The presence of fundamental transcriptomic distinctions within the hippocampus may be associated with varying cognitive and functional roles along its longitudinal axis. Taken together, this continuous atlas may advance our understanding of human brain organization and offers a bridge to link multiple neural scales across the hippocampus, both in health and disease.

Transcriptomics ¹

Methods Development

Brain Atlases ²

Databasing and Data Sharing

Multi-Modal Imaging

Computational Neuroscience

Other - Imaging-transcriptomics; Hippocampus; Gene expression; Multimodal analysis