Brain Struct Funct DOI 10.1007/s00429-015-1084-x ORIGINAL ARTICLE Distinct hippocampal functional networks revealed by tractography-based parcellation 1 2,5 3 2,4 Areeba Adnan • Alexander Barnett • Massieh Moayedi • Cornelia McCormick • 2,5 2,5 Melanie Cohn • Mary Pat McAndrews Received: 27 February 2015 / Accepted: 7 July 2015 Ó Springer-Verlag Berlin Heidelberg 2015 Abstract Recent research suggests the anterior and pos- parahippocampal, inferior temporal and fusiform gyri and terior hippocampus form part of two distinct functional the precuneus, predicted interindividual relational memory neural networks. Here we investigate the structural under- performance. These findings provide important support for pinnings of this functional connectivity difference using the integration of structural and functional connectivity in diffusion-weighted imaging-based parcellation. Using this understanding the brain networks underlying episodic technique, we substantiated that the hippocampus can be memory. parcellated into distinct anterior and posterior segments. These structurally defined segments did indeed show dif- Keywords DTI Á fMRI Á Hippocampus Á Recognition ferent patterns of resting state functional connectivity, in that memory Á Networks the anterior segment showed greater connectivity with temporal and orbitofrontal cortex, whereas the posterior segment was more highly connected to medial and lateral Introduction parietal cortex. Furthermore, we showed that the posterior hippocampal connectivity to memory processing regions, There is a growing consensus for the specialization of hip- including the dorsolateral prefrontal cortex, pocampal function along the anterior–posterior axis, arising out of differential involvement in large-scale brain networks (Poppenk et al. 2013). Functional neuroimaging has shown A. Adnan, A. Barnett and M. Moayedi contributed equally to this work and share first authorship. that the anterior and posterior hippocampi (antHC, postHC) are nodes within two distinct intrinsic networks: the antHC is Electronic supplementary material The online version of this functionally connected to perirhinal, lateral temporal, and article (doi:10.1007/s00429-015-1084-x) contains supplementary material, which is available to authorized users. ventromedial prefrontal cortices, and the postHC is con- nected to regions of the default mode network (DMN) & Areeba Adnan including medial and lateral parietal areas, medial prefrontal [email protected] cortex (PFC), and posterior cingulate cortices (PCC) (Kahn 1 et al. 2008; Poppenk and Moscovitch 2011; Libby et al. Department of Psychology, York University, Toronto, ON M3J 1P3, Canada 2012). This distinction has also found support in studies examining co-activation of regions during memory retrieval. 2 Krembil Neuroscience Center and Toronto Western Research Institute, University Health Network, Toronto, ON M5T 2S8, The postHC has been shown to be a key node of the recol- Canada lection memory network, a group of regions including the 3 Department of Neuroscience, Physiology and Pharmacology, parahippocampal gyrus (PHG), retrosplenial posterior cin- University College London, London WC1E 6BT, UK gulate cortex (rsPCC), and lateral parietal cortices that are 4 Centre for Developmental Cognitive Neuroscience, preferentially engaged when individuals are able to recall University College London, London WC1E 3BG, UK rich contextual details about events (e.g., Henson et al. 1999; 5 Department of Psychology, University of Toronto, Toronto, Eldridge et al. 2000; Yonelinas et al. 2005; Rugg and Vil- ON M5S 3G3, Canada berg, 2013). In contrast, the antHC is considered to be a node 123 Brain Struct Funct in the lateral temporal network, which includes the perirhi- investigate the cingulate, parietal, and temporoparietal nal/entorhinal and anterior lateral temporal cortices (Kahn junction, frontal polar cortices (Beckmann et al. 2009; et al. 2008; Libby et al. 2012; Poppenk and Moscovitch Johansen-Berg et al. 2004; Mars et al. 2011, 2012; 2011), subserving memory processes including familiarity Tomassini et al. 2007; Moayedi et al. 2014). Our analysis or the sense of prior occurrence in the absence of recall of involved four steps: (1) probabilistic tractography seeded other qualitative details (e.g., Yonelinas et al. 2005; Ran- from the hippocampus to the rest of the brain, (2) deter- ganath 2010). mination of the number of spatially consistent subregions Distinct structural connectivity profiles of the antHC and using a K-means clustering algorithm of the tractographic postHC could play an important role in supporting the data in each subject, (3) characterization of the functional distinct functional engagement of these regions that connectivity of these subregions, and (4) examination of underlie different memory processes (Honey et al. 2009). the relationship between functional connectivity of these Indeed, the antHC has direct connections with the amyg- subregions and memory performance on a verbal recogni- dala (Duvernoy 2005), and to the temporal pole, insula and tion memory task. ventromedial PFC via the uncinate fasciculus (Kier et al. 2004; Catenoix et al. 2011), while the postHC is connected Data acquisition to frontal and parietal neo-cortical regions via a polysy- naptic pathway involving the fornix projections to mam- All subjects provided informed consent to procedures millary bodies, anterior nucleus of the thalamus, and approved by the UHN Research Ethics Board. Diffusion- anterior cingulum (Duvernoy 2005; for review see Poppenk weighted images (DWI) were acquired for 15 healthy et al. 2013). While behavioral and functional neuroimaging subjects (5 women and 10 men; mean ± SD age: findings suggest the presence of functional subregions 34.33 ± 11.9 years) on a 3T Sigma MR System (GE along the longitudinal axis of the hippocampus, the struc- Medical Systems, Milwaukee). DWI were acquired along tural basis of such divisions has not been previously 25 non-colinear diffusion-encoding directions investigated in humans. Thus, we used diffusion-weighted (b = 1000 s/mm2). For each acquisition, one B0 imaging (DWI) to elucidate the structural connectivity of (b = 0 mm/s2) volume was acquired at the beginning of the hippocampus in vivo to assess whether there are dis- the run, and the parameters were as follows: cernible subregions within the hippocampus along the TR = 12,000 ms, FOV = 24 9 24 cm2, 128 9 128 longitudinal axis. We hypothesized that the functional matrix, 1.875 9 1.875 mm2 in-plane resolution, 3-mm heterogeneity attributed to the antHC and postHC arises thick, 46 axial slices. from differences in their structural connectivity. Also, a whole brain high-resolution anatomical scan was These structural connections can be estimated using also acquired for each subject using a 3D fast spoiled tractography (Johansen-Berg et al. 2004), which we applied gradient echo (FSPGR) sequence. The parameters were as to identify two hippocampal regions with unique white follows: flip angle: 12°,TR= 7.88 ms, TE = 3.08 ms, matter connectivity to the rest of the brain. We then 120 sagittal slices, FOV: 22 9 22 cm2, 256 9 256 matrix, investigated whether these structurally defined subregions 0.9 9 0.9 9 1 mm voxels, 1-mm thick. were differentially functionally connected to the rest of the Task-free rs-fMRI T2*-weighted functional MRI scans brain based on resting state fMRI (rs-fMRI). Finally, we with an echo-planar pulse imaging (EPI) sequence were examined the relationship between the intrinsic functional also acquired for every subject. The parameters were as connectivity of hippocampal segments with performance follows: TR = 2000 ms, TE = 25 ms, FOV: on a memory task. Specifically, we tested the hypothesis 24 9 24 cm2,649 64 matrix, 3.75 9 3.75 9 5 voxels, that connectivity of the postHC would predict relational 5-mm thick, 28–32 slices depending on head size, for 180 memory (Poppenk and Moscovitch 2011), which we volumes. Subjects were instructed to lie still, clear their operationalized as the performance advantage obtained on thoughts and ‘‘not to think about anything in particular,’’ a verbal recognition task when the study context was with their eyes open (the entire field of view was black). reinstated compared to when decisions were based on single items (Cohn et al. 2009a, b). Anatomical parcellation using DWI Seed region definition Materials and methods A hippocampal mask for each hemisphere was selected We investigated the structural connectivity of the hip- based on the Harvard-Oxford subcortical atlas (Desikan pocampi to determine whether there are subregions within et al. 2006), available in FSL v4.0 (Jenkinson et al. 2012). this structure with an approach that has been used to The probabilistic mask was then thresholded to ensure full 123 Brain Struct Funct Fig. 1 Hippocampal masks used for tractography-based parcellation. The left hippocampal mask is shown in the top panel (green) and right hippocampal mask is shown in the bottom panel (red). All masks are displayed on the MNI152 brain coverage of the hippocampus while eliminating voxels Cluster-based parcellation outside the structure such as in the amygdala or the PHG. Based on this requirement, we thresholded at 62 % for each DWI data were preprocessed using tools from FDT, part of hemisphere (Fig. 1). The hippocampus mask was trans- FSL. Motion and eddy current correction were performed formed to individual space using the linear registration tool using affine registration of all volumes to a target b0 vol- (FLIRT) implemented