The neural basis of semantic memory: Evidence from semantic dementia

Abstract

Semantic dementia (SD) is a syndrome of progressive impairment in semantic memory. Fifty-eight brain regions were measured in seven post mortem SD cases, ten normal controls and two disease controls (diagnosis frontotemporal dementia and motor neuron disease, FTD–MND). Manual segmentation of the whole brain has not previously been undertaken in a series of SD cases, either post mortem or during life. Widespread volume loss relative to controls was found in SD, with anterior temporal lobe regions bearing the brunt (>60% atrophy of temporopolar and perirhinal cortices bilaterally). Comparison of regional volumes in SD and FTD–MND found greater atrophy in SD only in temporopolar and perirhinal volumes. The sole region showing atrophy relative to controls in FTD–MND but not SD was motor cortex. Posterior temporal and frontal regions were not consistently affected and no significant asymmetry of atrophy was found. In summary, whole-brain regional evaluation in SD, in comparison with normal controls and FTD–MND, found anterior temporal atrophy encompassing the perirhinal cortex with relative sparing of adjacent posterior temporal regions.

Introduction

Semantic dementia (SD) is a syndrome of progressive impairment in conceptual knowledge associated with anomia, impaired comprehension and speech that is fluent but lacking in content (Bozeat et al., 2000, Hodges et al., 1992, Snowden et al., 1989). Other cognitive domains, including day-to-day memory, are relatively preserved (Hodges and Patterson, 1996, Hodges et al., 1992, Hodges et al., 1999).

In terms of neuropathology, most SD cases (72%) conform to a single subtype of FTD, with intra-neuronal deposits containing the protein ubiquitin (Davies et al., 2005). Such deposits are also described in motor neuron disease (MND) and are sometimes termed MND inclusions (Jackson et al., 1996). Patient with clinical features of FTD (typically behavioural disturbance and reduced fluency of speech) with signs of MND are increasingly recognized, and may be labelled ‘FTD–MND’ (Lomen-Hoerth et al., 2002, Lomen-Hoerth et al., 2003, Neary et al., 1990, Rakowicz and Hodges, 1998).

Manual tracing methods and automated voxel-based morphometry (VBM) in SD show asymmetric atrophy of the temporal lobe, worse on the left than the right (Chan et al., 2001, Galton et al., 2001b, Mummery et al., 2000). Comparisons of atrophy in SD and AD, moreover, find relatively greater atrophy in anterior (rostral) than posterior (caudal) temporal regions in SD (Chan et al., 2001, Davies et al., 2004). Regional measurements with reference to cytoarchitectonic boundaries found the most severe atrophy in perirhinal cortex (PRC), incorporating Brodmann areas (BA) 35 and 36 in the collateral sulcus and extending over the temporal pole (Davies et al., 2004, Saleem and Tanaka, 1996, Suzuki and Amaral, 2003). Volume loss was also noted in the anterior entorhinal cortex (ERC), to which PRC projects, but not in the posterior ERC (Davies et al., 2004, Suzuki and Amaral, 1994). Semi-quantitative assessment of neuronal loss in twelve SD cases likewise found the most severe abnormality in the antero-medial temporal lobe, in the region of the PRC (Davies et al., 2005).

By contrast, a single case study quantified atrophy and neuronal loss in language-associated gyri, finding involvement of the angular and the posterior temporal gyri (BA 37 and 39) bilaterally and involvement of the lateral temporal gyri (BA 20, 21 and 22) on the right, the case having been one of the minority with SD in which right sided atrophy predominated (Harasty et al., 1996b). Functional abnormalities in the posterior temporal lobe (BA 37) were also found in a PET activation study in SD employing a semantic association task, although this area was not atrophic as assessed by VBM (Mummery et al., 1999).

A difficulty in assimilating existing studies that differ in the emphasis of atrophy reported is that the measurements have often been limited to selected regions (Chan et al., 2001, Davies et al., 2004, Galton et al., 2001b). Whole-brain assessments in SD, to date, have been confined to automated analyses in the form of VBM (Mummery et al., 2000, Williams et al., 2005). This method, however, may be insensitive to differences between small regions, especially those adjacent to CSF (Good et al., 2002, Mummery et al., 2000). Warping and smoothing required for image processing are particularly problematic when, as in SD, the disease causes marked anatomical distortion. The neural basis of semantic memory remains controversial in part because patients with isolated semantic impairment from focal brain lesions do not exist.

Our primary aim was to assess regional atrophy, in comparison with normal controls, across entire post mortem brain specimens in SD, to illuminate the neural basis of semantic memory. A second aim was to compare regional atrophy in SD with a disease control group, namely FTD–MND, contrasting in neuropsychological impairments but similar in microscopic neuropathology. The third aim was to examine data on regional volumes and disease duration by factor analysis to identify underlying patterns.