This work presents the only study of brain activation in asymptomatic offspring from confirmed cases of late-onset familial Alzheimer’s disease and clearly identifies functional differences in this high-risk sample. In response to the challenge of an episodic memory task, where individuals need to both encode and retrieve new information (in this case unrelated word-pairs), at-risk individuals prove capable of completing this task with the same degree of accuracy as age- and education-matched controls. However, the patterns of activation that accompany each of these phases of the task in those at risk differ markedly from that seen in the control sample. Familial risk is associated primarily with increases in activation during memory encoding, irrespective of APOE allele status, and decreases in activation during cued recall.
Explicit memory impairment is generally considered the earliest clinical symptom of Alzheimer’s disease, with studies identifying distinct brain regions involved in the encoding of new information into memory and the retrieval or recall of this stored information. MTL structures, particularly the hippocampus, entorhinal cortex and parahippocampal gyrus, as well as regions in the frontal cortex, have all been shown to be intimately involved in the encoding of new material as demonstrated in both animal and human studies (
Henke et al., 1997;
Desgranges et al., 1998;
Buckner et al., 1999;
Rombouts et al., 1999;
Schacter and Wagner, 1999;
Cabeza and Nyberg, 2000b;
Gron et al., 2003;
Pihlajamaki et al., 2003). These medial temporal structures, particularly the entorhinal cortex and hippocampus, are the site of initial Alzheimer’s disease pathology, which begins while individuals are clinically asymptomatic (
Hyman et al., 1984). As shown by the activation maps for the healthy controls, the paired-associates paradigm used in this study has proven robust for examining haemodynamic changes in these important brain regions during encoding, and thus provides a window on the neural processing associated with intentional memory processing. Individuals at familial risk for Alzheimer’s disease included in this study evidence more extensive and intense activation in these same brain regions in order to successfully complete the task. This provides support for the compensatory hypothesis that postulates that additional neural resources are recruited to compensate for neuronal loss (
Grady et al., 2003;
Grossman et al., 2003;
Dickerson et al., 2004).
The retrieval of information placed in long-term storage appears to rely on a somewhat different processing network, which varies with the type of stimuli and the manner of retrieval. In this paradigm, recall is initiated with an auditory cue (i.e. the first word of the pair), with individuals asked to think of the second word of the pair. The few studies that have used a similar method of assessing the recall of learned pairs [reviewed by
Cabeza and Nyberg (2000a)] find increased activation in the frontal cortex, as well as in the thalamus. Other areas of increased activation are less consistent but can include temporal cortex, cingulate and lateral parietal areas. MTL structures, such as the hippocampus and entorhinal cortex, important during encoding, are less involved in the retrieval of material, with their involvement decreasing over repeated trials (
Petersson et al., 1997). While both groups in this study show activation in the frontal and temporal cortices, in contrast to encoding where the at-risk group produced more extensive and intense activation, here, the at-risk show significantly less activation. These less-activated structures include the thalamus and specifically the anterior nucleus, a region important for memory function with major connections to the hippocampal complex and both the anterior and posterior cingulate (
Van der Werf et al., 2003;
Taber et al., 2004). In fact, the thalamus is consistently less active in the at-risk group across both conditions of this study. Of note, bilateral thalamic hypometabolism was reported as the distinguishing feature of asymptomatic carriers of the APP717 mutation for early-onset Alzheimer’s disease (
Perani et al., 1997).
The activation patterns seen in these asymptomatic individuals at elevated risk for Alzheimer’s disease differ from those reported for individuals already diagnosed with AD, where the most consistent finding across fMRI studies employing episodic memory tasks has been significant reductions in MTL activation including the hippocampus and parahippocampal gyrus (
Small et al., 1999;
Rombouts et al., 2000;
Sperling et al., 2003;
Gron and Riepe, 2004). They also differ from those patterns reported for individuals with mild cognitive impairment, where studies report both decreased and increased MTL activation. These studies have not investigated other structures that are probably important, such as the cingulate and thalamus (
Small et al., 1999;
Machulda et al., 2003;
Dickerson et al., 2004). Further work is required to understand whether there is some continuum of dysfunction in specific brain structures and regions or whether there will emerge a picture of temporal alterations that occur in response to changes in the efficiency of brain circuitry, in this case for memory function.
The absence of volumetric differences between the at-risk sample and the controls, in the face of functional differences, is in keeping with previous reports distinguishing functional and structural differences. For example, Reiman, using both resting PET and MRI, contrasted 11 individuals with APOE 4/4 with 22 individuals with APOE 3/3 and found that while the ε4 allele was associated with decreased metabolism, particularly in the cingulum, there were no volumetric differences (
Reiman et al., 1996,
1998), demonstrating, as does this study, that functional differences occur in the absence of volumetric loss. It appears that volumetric loss is discernable only in the presence of cognitive decline, which is absent in the at-risk sample included here (
Wolf et al., 2001). Recent imaging work suggests that surface deformity of the hippocampus, rather than volume of the hippocampus, may be predictive of subsequent cognitive decline and appearance of Alzheimer’s disease (
Csernansky et al., 2005).
This study has several methodological shortcomings that must be noted. The scan FOV provided only partial coverage of the brain. This was done to enhance the scan resolution in the MTL; however, it prevented us from investigating other potential differences between groups in regions outside of the FOV. Declarative memory involves structures in the frontal lobe, and thus complete frontal lobe coverage would have been helpful. With the onset of technological advances in fMRI, and the ability to scan at high tesla (3+) with more optimized parameters, it will be possible to use scan parameters that provide adequate coverage of the whole brain, as well as optimal resolution in the MTL in future studies. Secondly, it is possible that there were differences between groups in recall speed, which could have caused the activation differences observed between groups during recall. One could hypothesize that the at-risk sample would be slower, and therefore may show increased activation, reflecting more effort. However, the study finds the opposite. At-risk participants demonstrate decreased activation in several structures during recall, which makes the possibility of a difference in recall speed less likely. It is also unlikely that recall speed was the driving force behind the differences in medial temporal activation, since most of these differences were only found during the encoding phase and not during retrieval. Finally, we did not use a control for auditory stimuli, in order not to interfere with the auditory encoding of test stimuli. Auditory activation, as expected, was present and was reported in the superior temporal gyrus. Since this is anatomically distinct from medial temporal activation, it was not considered a confounding factor in our analysis.
The current study highlights a number of areas for further examination, including the relationship of APOE ε4 to changes in haemodynamic response. This is particularly important since the APOE ε4 allele is associated with a number of chronic medical conditions that impair cognition. For example, while this study demonstrates the lack of association between increases in activation and the presence of the APOE ε4 allele, PET studies have found the ε4/ε4 genotype associated with significantly reduced rates of glucose metabolism in the posterior cingulate, parietal, temporal and prefrontal regions, in both middle-aged and young adults. These are the same areas that have shown metabolic reductions in individuals experiencing memory complaints and those diagnosed with Alzheimer’s disease (
Reiman et al., 1996,
2004). Thus, an area of further inquiry is the relationship of APOE allele status to decreases in haemodynamic response in this memory task. In addition, it may be important to identify unique activation patterns within this high-risk sample, as these may reflect genetic heterogeneity in late-onset Alzheimer’s disease and be linked to specific genes or genetic haplotypes (
Kennedy et al., 1995;
Rossor et al., 1996;
Bassett et al., 2005). Such patterns may also provide information on clinical prognosis as suggested in a recent PET study (
Ishii et al., 2003). Both of these may, in turn, ultimately be useful for individualizing treatments. Finally, the consistent reduction in activation in the cingulum and thalamus, which was apparent both in the encoding and cued recall comparisons, will require further investigation. The thalamus and cingulum have been shown to be activated during memory retrieval in a number of studies
see [
Desgranges et al. (1998) for a review], and PET studies have reported hypoperfusion in both of these regions in the resting state.
In conclusion, this study finds that individuals at risk for familial late-onset Alzheimer’s disease, although clinically asymptomatic, demonstrating normal cognitive performance, present a different pattern of functional brain activity when completing tasks requiring memorization and recall. These functional differences are apparent even though these individuals are more than a decade from the age at which their parent became clinically symptomatic. In addition, these functional differences occur in the absence of volumetric loss, suggesting that compensatory mechanisms may be present before any significant neuronal loss. Indeed, neuropathology studies of individuals showing significant amyloid deposits in the context of normal cognition do not find neuronal loss in temporal lobe structures (
Price et al., 2001;
West et al., 2004). Finally, the increased brain activation seen in the at-risk subjects during memory encoding is unrelated to the APOEε4 allele. In summary, this work provides evidence that familial risk for late-onset Alzheimer’s disease, related to a yet-unidentified gene or genes, is associated with altered brain function. This work sets the stage for the examination of temporal and spatial patterns of functional alterations in those at genetic risk for this disease. Longitudinal study of these high-risk individuals will enable us to determine whether these patterns, or a subset of them, are ultimately predictive of disease onset.
The publisher's final edited version of this article is available free at 
