Research theme #1

Imaging hippocampal substructures

While the hippocampus has been the focus of intense research over the last decades, a lot remains to be discovered. Notably, it is now fully acknowledged that it is not an homogeneous structure but rather a composite entity that can be divided in two orthogonal, complementary ways: its histologically-defined subfields (CA1-4, dentate gyrus, subiculum) and its longitudinal axis (classically: head – body – tail).

With the recent improvement of image resolution and the development of sophisticated neuroimaging analysis methods, these substructures are now accessible to in vivo neuroimaging research.

IN VIVO IMAGING OF HIPPOCAMPAL SUBREGIONS. The hippocampus can be divided along two orthogonal approaches: the anterior-posterior axis or histologically-defined subfields. Hippocampal subpart structure, function and connectivity patterns can be assessed separately using multiple neuroimaging techniques (structural MRI, activation functional MRI and resting-state functional MRI). Inversely, imaging metrics can sometimes be used to identify functionally-relevant subregions in a data-driven manner.

---

Hippocampal subfields

In the Inserm u1077 research unit in Caen, France, we developed a proton density weighted MR sequence on a 3T scanner together with a segmentation protocol in order to assess the volume of hippocampal subfields (CA1, Subiculum, and “others” including CA2-3-4 and dentate gyrus). This method was first published in 2010 and used on a group of 50 healthy individuals between 19 and 68 yo (La Joie et al., 2010). Further analyses included patients with amnestic Mild Cognitive Impairment (aMCI), Alzheimer’s disease (AD) and semantic dementia (SD), and a larger group of 98 healthy subjects between 19 and 84 yo (La Joie et al., 2013; de Flores et al., 2015).

HIPPOCAMPAL SUBFIELD VOLUMETRY IN AGING AND DEMENTIA. Images are derived from La Joie et al., Neuroimage Clinical 2013; de Flores et al., Human Brain Mapping 2015 (and Harding et al., Cerebral Cortex 1998 for the upper right corner)

Globally, our results showed that:

CA1 ATROPHY IN MCI. CA1 volumetry is better than whole hippocampus volumetry to distinguish aMCI from controls, and is found in patients whether they have amyloid-beta in their brain or not (images are derived from La Joie et al., Neuroimage Clinical 2013).

1) healthy aging, assessed cross-sectionally, is accompanied by a linear atrophy of the subiculum volume, and non linear changes in the volume of the CA1 subfield, with a decrease after 50 yo;

2) over the course of AD, atrophy of CA1 and the subiculum predominates from the early stage of MCI while the “other” subfields are relatively spared;

3) subfield volumetry has potential clinical value: CA1 volumetry outperforms global hippocampal volumetry in distinguishing MRI patients from controls (AUC= 0.88 versus 0.76, p=0.05);

4) CA1 atrophy is also found in patients with SD, and is detectable in MCI patients, whether they have Aβ deposition in their brain or not;

CA1 ATROPHY AND ENCODING EPISODIC MEMORY DEFICITS IN aMCI (Images are derived from Fouquet et al., Neuroimage 2012).

We also used a complementary method for assessing subfield atrophy, based on a 3D render of Voxel Based Morphometry (VBM) results onto a hippocampal surface (Chételat et al., 2008; La Joie et al., 2010). Doing so, we showed that episodic memory encoding deficits in MCI patients were specifically related to atrophy of the CA1 subfields.

Finally, we wondered whether discrepancies in the existing litterature on hippocampal subfields and AD could be due to differences in the variations between segmentation protocols used by different groups. We specifically tested the differences between our technique (manual delineation on high resolution images) and the automated segmentation included in the freely available Freesurfer package, which offers to segment subfields from classical T1-weighted MR images.We therefore used the two methods on the same sample of healthy individuals, MCI patients, and AD patients. The results, presented in de Flores et al (2015), showed that methodological differences, and especially variations in subfield definition have a heavy impact on the results regarding AD-related changes in subfield volume.

In line with this concern, I got involved with the hippocampal subfield group, gathering several experts of hippocampal subfield imaging who seek to develop a unified subfield segmentation protocol. The first step of this work was to quantitatively characterize the differences that exist between the main segmentation protocols, and has recently led to a publication (Yuchkevich et al., 2015).

---

Anterior-posterior gradient

ANTERIOR VS POSTERIOR HIPPOCAMPUS INTRINSIC CONNECTIVITY. Resting State fMRI analyses showed different large-scale cortical connections along the long axis of the hippocampus (image derived from La Joie et al., Neuron 2014).

---

Papers I (co)authored: La Joie et al., Neuroimage 2010; Fouquet et al., Neuroimage 2012; La Joie et al., Neuroimage clinical 2013; La Joie Neuron 2014; La Joie The Journal of Neuroscience 2014; de Flores et al., Human Brain Mapping 2015; Yushkevich et al., Neuroimage 2015.

Main collaborators: Gaël Chételat, Robin de Flores, Marine Fouquet, the hippocampal subfield group

Further reading: Small et al., Nature Review Neuroscience 2011; Ranganath & Ritchey, Nature Review Neuroscience 2012; Poppenk et al., Trends in Cognitive Neuroscience 2013; Maruszak & Thuret, Frontiers in Cellular Neuroscience 2014.