In the present study, parental family history was associated with altered brain microstructure manifested as lower FA measured by DTI. Furthermore, FA was lower with increased risk load. Specifically, possessing both positive family history and ApoE ε4 was associated with the lowest FA compared to possessing only one or no risk factors. Affected brain regions were those that are also affected in AD, including the corpus callosum (26
), the cingulum (27
), and medial temporal gray and white matter (29
). Our findings in the vicinity of the hippocampus proper and white matter adjacent to the hippocampus are similar to previous findings associated with ApoE ε4 (8
). To fully understand FA measurements in gray matter, multiexponential modeling of slow and fast diffusion components is needed (53
); however, FA measurements in the hippocampus are plausible. The hippocampus has an organized cytoarchitecture that restricts water diffusion in such a way as to cause anisotropic diffusion (54
), and hippocampal differences in diffusion have also been reported in MCI (55
) and AD (29
Participants with family history of AD in the present study also showed lower FA in left tapetum. The tapetum is a tract with fibers running between posterior corpus callosum and medial temporal lobe, and is thought to form a connection between the medial temporal lobe and the contralateral temporal cortex. Degeneration of the tapetum that is secondary to temporal lobe abnormality has been observed in temporal lobe epileptics (56
). It is possible that preclinical AD involves a similar course of degeneration, with early AD pathology affecting the medial temporal lobe, followed by degeneration of cross-connecting fibers. This is supported by several studies in AD indicating that posterior corpus callosum is affected to a greater degree than the anterior (genu) portion of the corpus callosum (26
). A difference was also found in the left uncinate fasciculus, a tract connecting anterior temporal lobe with frontal orbital cortex. Alzheimer’s patients show lower anisotropy in the uncinate compared to control (27
), and amnestic MCI is associated with lower FA on the left, with uncinate FA also being correlated with episodic memory in aMCI (58
In general, regions where we found an effect of family history were clustered on the left side of the brain. Our hippocampal results were bilateral, but we found a slightly larger cluster on the left. Several of the microstructural findings in people with increased risk for AD due to MCI are asymmetrical, with several results favoring greater left medial temporal lobe alterations (29
), which may reflect a unique pattern of AD development. It is also possible that the changes are bilateral but that limitations in imaging or statistical analysis fail to reveal bilateral alterations. This is somewhat supported by our observation that raising the statistical threshold in our analysis yielded greater unilateral effects.
Only a few studies have examined the effect of risk for AD on DTI measures in a cognitively normal population. Using transverse relaxation rate (R2), a measure that—based on its sensitivity to small changes in tissue water content—is sensitive to myelin breakdown, Bartzokis et al. have demonstrated an effect of ApoE ε4 on the trajectory of aging (61
). In cognitively normal adults aged 55 to 75 years, ApoE ε4 resulted in a steeper age-related decline in myelin compared to controls. In a later study, Bartzokis et al. also demonstrated that ApoE ε4 associated myelin alterations are related to cognitive processing speed (62
The effect of ApoE ε4 in cognitively normal populations has also been investigated using DTI. Honea et al. (63
) examined the effect of ApoE ε4 on FA in older adults (over the age of 60 years) and found that ε4 was associated with lower FA in the left parahippocampal gyrus white matter. Similarly, Nierenberg et al. (7
), compared ApoE ε4 carriers to non-carriers, and found that carriers exhibited lower FA in parahippocampal white matter. In the present study, there was no main effect of ApoE ε4, however, our participant sample was substantially younger than that of Nierenberg et al. (7
) and Honea et al. (63
). In another study, Persson et al. (8
) found that ApoE ε4 was associated with decreased FA in left hippocampus in a location similar to our finding. However, when ApoE status was taken into account in our sample, we found that decreased FA was more strongly related to family history rather than ApoE ε4, and that ApoE ε4 was additive.
The lack of main effect of ApoE ε4 was an interesting finding of the present study, and suggests that previous findings associated with ApoE ε4 may in fact have been due to family history of AD (except in cases where the contribution of family history was controlled). It is well known that the ε4 allele is overrepresented in family history positive cohorts (33
), which could make it difficult to interpret the effects of ε4, if there is something embedded in family history that exerts an effect beyond ε4 as suggested by the present results. The mechanism by which family history of AD bestows increased risk for the disease is ambiguous. Unlike ApoE ε4, there is no clear biomarker associated with family history, and presumably the effects of this risk factor are mediated through an as of yet unknown gene or genes. Although it is surprising that there was no main effect of ApoE ε4, family history by itself and in combination with ε4 is known to increase risk for the development of AD, and exert effects that suggest a unique role in AD pathogenesis. Asymptomatic middle age persons with a family history have elevated plasma A-beta (64
) and neurocognitive changes consistent with early AD (65
) that are independent of ApoE genotype. Once AD has developed, presence of both family history and ApoE ε4 together are associated with an identifiable cognitive phenotype of AD, namely an amnestic presentation (66
). Family history of AD is also associated with a deficit in odor identification that is independent of ApoE genotype. Anosmia is known to be a marker of incipient AD, and the single study of this phenomenon in siblings of AD patients revealed a similar pattern to ours, indicating that even though ApoE is not an independent predictor of deficit, it is additive: siblings of AD patients who also carried the ApoE ε4 genotype showed the greatest deficit. Finally, the potential existence of novel genetic markers embodied in family history is supported by neuroimaging studies that have stratified by family history and ApoE genotype. Family history affects the magnitude of MRI perfusion measurements and BOLD signals (14
) as well as PET measurements of CMRglc (19
Further complicating the family history effect are data indicating that the effect of family history is multi faceted. Evidence supports a higher rate of maternal inheritance of AD (67
), compared to paternal inheritance, and maternal family history is associated with a lower CMRglc in AD susceptible brain regions compared to paternal family history and absence of family history (19
). Furthermore, having both maternal and paternal family history confers a higher risk of developing AD compared to having only one parent with the disease, or no parental family history (12
). Consistent with these studies, our analysis indicated that maternal family history is associated with significantly lower hippocampal FA compared to absence of family history. Although it was not a significant effect, there was a trend toward lower hippocampal FA in the small group of participants who had two parents diagnosed with AD. Parental family history of AD is an understudied risk factor that clearly needs increased attention, given the results of this study and others (14
Although AD has historically been considered a disease of gray matter, white matter clearly undergoes an alteration in the disease course. Post mortem assessment of white matter in AD has revealed several pathological differences compared to control, such as denudation of the ventricular ependyma, gliosis, and loss of myelinated axons in the deep white matter (70
). Biochemical analysis of post mortem white matter in AD patients has shown increased quantities of Aβ40 and Aβ42, less total myelin protein, decreased white matter cholesterol, and increased total fatty acid content (71
). Interestingly, the pattern of Alzheimer-related neurofibrillary changes in the neocortex inversely recapitulates the development of myelin (72
), an observation that suggests that break down of white matter is a key event in the development of AD pathology (23
). Our findings of altered white matter integrity in a largely middle-aged cohort lend support to the hypothesis that white matter change is an early event in the development of AD pathology.
Several underlying pathologies may be responsible for the diffusion differences observed in the present study. Future studies, preferably those employing biomarkers or additional modalities of imaging, will be needed to evaluate the contributions of inflammation (73
), amyloid deposition (77
), or tau pathology (79
), to alterations in white matter. Vascular health is associated with white matter integrity, and future work will also be needed to ascertain the contribution of vascular risk factors to white matter alterations and AD risk. Arterial spin labeling perfusion methods in combination with DTI hold promise for this avenue of work.
Although we found alterations in radial diffusion suggesting differences in myelin, as opposed to axonal changes, further work will be needed to assess whether preclinical AD involves primary myelin degradation (80
), or white matter loss that is secondary to neuronal or axonal degradation. Finally, it should be noted that the interpretation of differences in FA and the eigenvalues is limited. Only a handful of studies have been carried out in humans that relate DTI measures to underlying anatomy and pathology in humans using ex vivo brains (81
). In particular, attempts to differentiate between axonal loss, Wallerian degeneration, and demyelization by analyzing the eigenvalues of the tensor are largely based animal studies not human brains.
A few additional limitations of the present work should be acknowledged. First, our study adopted a voxel-wise approach of analysis. This approach has been widely embraced by many investigators, benefits from automation, provides the capability of quickly analyzing large quantities of brain images, allows for the evaluation of brain regions that may not have been considered a priori, and appears to produce results that largely concur with region of interest (ROI) based methods (27
). However, it is limited by spatial normalization methods and has been challenged as a result (86
). In the present study, images were inspected visually for errors in registration, and a field mapping procedure was used to overcome susceptibility-related distortion, however, we cannot completely rule out normalization differences as the source of our results. Furthermore—due to large number of statistical comparisons—voxel-wise results are usually evaluated using stringent criteria for significance. This can result in artificially high correlations in highly localized brain regions (87
). In the present study, we employed a threshold that was less stringent. Our study population consisted of largely middle-aged adults who are increased risk for AD, rather than a patient population, hence we expected that structural effects would likely be subtle. Still, it should be recognized that employing a threshold that is not corrected for multiple comparison brings with it the risk of a type 1 error and that significant results must consequently be regarded with caution.
The results of our study may also be limited by the resolution of our DTI scan. The resolution of typical DTI (3 mm thick sections in the present study) means that the technique is prone to partial volume effects. Depending on the location of acquired voxels over brain regions where we found significant effects, it is possible that the FA measures could be influenced by the inclusion of cerebrospinal fluid (CSF). This is particularly true for our hippocampal cluster, which was located in the posterior part of this structure. Future studies with smaller voxel sizes are needed to contribute toward more accurate measurements in structures that are proximal to CSF.
T2-wieghted imaging (with or without suppression of CSF) can provide complementary information to DTI. Abnormal white matter appears hyperintense on T2-weighted scans and in healthy adults, these lesions are most likely attributable to ischemia. T2-weighted scans can be used for quantifying total lesion burden, or for comparing regional distribution of lesion burden between groups. We did not examine white matter lesions in this study; thus, we can not conclude whether our results obtained with DTI were due to differences in the distribution of ischemic white matter damage between groups. The mechanisms that underlie DTI measured white matter alterations in preclinical AD require further study.
Additional limitations of the present study include the fact the number of people in the group of participants who had both a mother and a father with AD was small, limiting our analysis of family history load. Although these results are suggestive, they require replication in additional study samples. The sample was on average highly educated, largely Caucasian, from higher socioeconomic strata, and self-selected based on a shared interest in memory research with a specific focus on family history of AD, limiting the study’s generalizability. Additional studies are needed to further evaluate the unique interplay between family history and ApoE ε4 risk factors in the pathogenesis of AD.