In one of the largest TBM studies to date, and one of the largest MRI studies of AD and MCI, we found that baseline temporal lobe atrophy (1) correlates with cognitive impairment (measured using CDR-SB, MMSE, and logical memory test scores), (2) predicts future cognitive decline (in terms of the CDR-SB), in all of the AD, MCI, and normal groups, and (3) predicts conversion from MCI to AD, over a subsequent one-year period. We found a dose-dependent association of the ApoE4 gene with greater structural atrophy, and for the first time, we established neuroimaging evidence for a protective effect of the ApoE2 gene in healthy controls. ApoE2 carriers – who made up around 1/6 of the normal group - showed more limited ventricular expansion, which likely reflects preservation of brain parenchyma.
In an initial pilot study (N = 120
) (Hua et al., 2008
), our main aim was to determine which types of parameter selections in TBM (e.g., 9- or 12-parameter global alignment, cross-subject versus group template registration) would be optimal to detect group differences between AD and normal subjects. In such a small sample (only 40 per group), we were barely able to differentiate MCI and normal groups, and with 120 subjects in total, correlations with clinical scores were significant only when all three groups were pooled together. It is common practice to combine various diagnostic groups (e.g., AD, MCI and normal) to achieve a greater range of disease severity for correlations with MRI volumetric measurements. In the current study, we used a sample size that was almost five times greater, which allowed us to detect correlations even within
the diagnostic categories. Not only did we detect significant differences between MCI and normal, but we were also able to correlate temporal lobe atrophy at baseline with a variety of clinical measures and cognitive test scores, independently within each diagnostic group. TBM-based computations of temporal lobe atrophy predicted future cognitive decline (annual change in CDR-SB) even in the healthy subjects. The risk of deteriorating from MCI to AD in the one year period after the baseline scan was considerably greater among those with greater temporal lobe atrophy. Our results agree with prior studies suggesting that the level of atrophy in the hippocampus, entorhinal cortex, whole brain, and the degree of lateral ventricular expansion, are typically greater in MCI converters versus stable MCI subjects (Bozzali et al., 2006
; Jack et al., 2004
). Given the effect sizes, our samples are close to the minimal sizes necessary to detect these effects with TBM; this information may be useful in planning future studies, including therapeutic trials.
Our finding that MCI converters have greater baseline atrophy than those who remain stable agrees with several prior reports. It has been shown that MRI can predict likelihood of (or time to) progression from MCI to AD (Jack et al., 1999
). In Apostolova et al. (Apostolova et al., 2006
), we found distinct patterns of hippocampal atrophy in MCI subjects who remained stable or recovered, versus those who declined, over a 3-year follow-up period. In a recent VBM study, MCI converters showed more widespread areas of reduced gray matter density than MCI non-converters, but in a similar pattern to that seen in AD (Bozzali et al., 2006
AD pathology follows a characteristic anatomical trajectory, with the earliest observable changes in the entorhinal cortex and hippocampus, then moving to temporal and parietal lobes, and finally affecting the frontal lobes in the late stages of AD (Braak and Braak, 1991
; Thompson et al., 2003
; Thompson and Apostolova, 2007
). With that in mind, we were interested in assessing genetic influences on structural differences in the temporal lobes and ventricular regions, as they are among the earliest to show disease-related atrophy.
The apolipoprotein E (ApoE) gene is located on chromosome 19 with three alleles (ApoE2, ApoE3, ApoE4) (Zannis and Breslow, 1982
; Zannis et al., 1982
). ApoE3 is the commonest allelic variant, while ApoE4 increases and ApoE2 decreases susceptibility to AD (Corder et al., 1993
; Saunders et al., 1993
; Corder et al., 1994
; Blacker et al., 2007
;). ApoE4 is over-represented in the AD population () and confers a dose-dependent risk, in which two E4 alleles confer greater risk than one (Corder et al., 1993
; Geroldi et al., 1999
). ApoE4 homozygotes (ε4/ε4) typically show greater deficits than ApoE4 heterozygotes (ε3/ε4) with similar demographic risk factors (Geroldi et al., 2000
; Lehtovirta et al., 2000
; Martins et al., 2005
We were able to demonstrate genetic effects on brain structure, even among healthy elderly subjects. 17% - or around one-sixth - of the normal controls carried an ApoE2 allele, and had smaller ventricular volume than homozygous ApoE3 carriers, which is the commonest genotype. As progressive ventricular enlargement is a typical process during normal aging and brain degeneration, this neuroimaging evidence may reflect a protective effect of ApoE2 on the surrounding brain parenchyma, detected as an effect on ventricular size. We also found a dose-dependent effect of ApoE4. In MCI, one copy of ApoE4 was not detectably associated with temporal lobe differences, but it was associated with elevated CSF volume in the Sylvian fissures, a region where changes are commonly seen in the initial phase of temporal lobe degeneration. An additional copy of ApoE4 was associated with further hippocampal atrophy in MCI, and was correlated with still further deterioration of temporal lobe structures in AD. The dose-dependent effect of ApoE4 seems to mirror an accelerated AD progression in the same topography as typical AD development, starting in the entorhinal cortex and hippocampus at the MCI stage, then moving on to temporal and parietal lobes in AD.
Our ApoE4 genetic findings are consistent with prior MRI studies. Several MRI studies have shown reduced hippocampal cross-sectional area (Tohgi et al., 1997
; Moffat et al., 2000
) and greater volumetric atrophy in the entorhinal cortex (Juottonen et al., 1998
; Geroldi et al., 1999
; Jack et al., 2007
) in ApoE4 carriers versus non-carriers. Using the brain boundary shift integral method, and iterative principal component analysis, accelerated whole brain atrophy was associated with ApoE4 gene in a dose-dependent way (Chen et al., 2007
). Recently, voxel-based morphometry (VBM) has made it easier to map the profile of genetic influences throughout the entire brain, without restriction to pre-defined regions-of-interest. In a large VBM study of 750 healthy elderly subjects, ApoE4 homozygotes displayed significant medial temporal lobe deficits as compared to ApoE4 heterozygotes and non-carrier subjects (Lemaitre et al., 2005
). In another VBM study of cognitively normal subjects, ApoE3/ApoE4 carriers, relative to non-carriers, showed regionally reduced gray matter density in right medial temporal and bilateral frontotemporal regions (Wishart et al., 2006
). Although we did not find an ApoE4 effect on brain morphology in normal subjects here, we recently detected abnormal ventricular expansion in a different sample of healthy elderly ApoE4 carriers versus non-carriers, by developing a automated method to extract 3D surface-based models of the lateral ventricles (Chou et al., 2008b
) As that method combined multiple fluid registrations to create increasingly accurate models of specific anatomical structures, it may be that an ApoE4 effect in normals could be detectable using a multi-atlas version of TBM (cf. Chou et al., 2008
) or by using other methods that directly model the affected anatomical structures. For example, we recently applied a cortical flattening and thickness mapping approach to the entorhinal cortex, and found thinner cortex in specific hippocampal subregions in cognitively normal ApoE4 carriers versus non-carriers (Burggren et al., 2008
). Lastly, another reason we may have found an ApoE4 effect in MCI but not in normals is that our MCI group was almost twice the size (N=323) of our normal group (N=179), and around half of the MCI subjects but only a quarter of the controls carried ApoE4. The power to detect an effect was therefore greater in the MCI group. In other words, our detection of ApoE4 effects in MCI but not controls does not imply the effect size is any greater in MCI. Conversely, the protective ApoE2 effect we found in controls may also be equally present in MCI or AD subjects, but the genotype is so rare in AD and MCI that sufficient numbers of ApoE2 subjects could not be found to provide adequate power. Using placebo subjects in a recent study of vitamin E and donepezil in MCI, Jack et al. (2007)
found that rates of hippocampal atrophy were greater in E4 carriers than non carriers (but not significantly different between E4 homozygotes than heterozygotes – although this latter finding could just be due to small numbers).
VBM is an unbiased technique that has been applied to various MRI studies of AD, MCI and normal subjects to provide a comprehensive view of normal aging as well as brain degeneration. Good et al. (2001)
used VBM to examine the effects of age on brain tissue volumes in 465 normal adults (Good et al., 2001
). Global gray matter volume was shown to decrease linearly with age, with accelerated loss in the insula, superior parietal gyri, central sulci, and cingulate sulci bilaterally. In contrast, regions such as the amygdalae, hippocampi, and entorhinal cortices were relatively preserved during normal aging. Similar time-lapse trajectories were computed in our recent cortical thickness analysis of 176 subjects aged 7 to 87, with temporal lobe atrophy accelerating in normal old age (Sowell et al., 2007
; Sowell et al., 2003
). Smith et al. (Smith et al., 2007
) prospectively followed 136 cognitively normal elderly and 5 MCI subjects with MRI for an average of 5.4 years. 23 of them converted to MCI and 9 of the 23 further deteriorated to meet criteria for AD. At baseline, the group consisting of the 23 pre-MCI/pre-AD and the 5 MCI subjects demonstrated significant gray matter atrophy in the anteromedial temporal and left angular gyri, and the left lateral temporal lobe. VBM revealed clusters of gray matter loss in bilateral medial temporal, posterior cingulate, and temporoparietal structures in AD (Baron et al., 2001
). Optimized VBM was used to demonstrate gradually reduced global gray matter volume from MCI to AD compared to normal, characterized by medial temporal lobe damage in MCI and additional parietal and cingulate cortices loss in AD (Karas et al., 2004
), in a pattern that we recently confirmed using cortical mapping methods (Apostolova et al., 2007
; Frisoni et al., 2008
). Chetelat et al. examined 18 MCI subjects of whom 7 converted to AD during an 18-month follow-up interval (Chetelat et al., 2005
). Fastest atrophy rates were identified in the temporal pole, entorhinal and lateral temporal cortices (2.5–4.5%). The prefrontal cortex was more affected in non-converters. At baseline, greater cortical involvement was seen in the parahippocampal, fusiform, lingual and posterior cingulate cortices in MCI subjects who later converted to AD versus those who did not.
TBM-based analysis is a powerful tool with potential to monitor structural atrophy in AD at the incipient stage, before severe cognitive impairment takes place. Recent developments in TBM have included a method called DARTEL, an algorithm for diffeomorphic image registration (Ashburner, 2007
). It was applied to a large MRI dataset (N = 471) to automatically extract information regarding gender and age effects on brain morphometry. These studies and ours show that TBM is a highly automated technique, capable of revealing voxel-level associations between neuroimaging markers and clinical or genetic variation. TBM should facilitate the discovery of factors that influence disease progression, as well as protective factors, in large-scale MRI studies.