Brain WM forms the backbone of a large network connecting multiple segregated cortical regions, which occupies nearly as much of brain volume as the gray matter [39
]. There is a growing body of the literature suggesting active WM degeneration as part of the repertoire of pathophysiological processes underlying AD. Although somehow neglected, the impact of WM abnormalities in AD has been reported to be large [4
Here, we report increased MD (6-7%) occurring in approximately 28% of estimated WM and no FA declines in aMCI. In contrast, the WM of AD individuals had slightly higher MD (8.5%) and lower FA (11.3%) affecting a larger proportion (66%) of estimated WM. The magnitudes of MD or FA group differences reported here are above expected variability as revealed by previous DTI test-retest studies [40
]. In addition, we report, for the first time, brain areas disconnected from the hippocampus due to WM abnormalities in AD patients using FSL’s probabilistic tractography method [12
]. This technique provides the likelihood of connectivity between the hippocampus and any given brain region. Our results support the hypothesis that AD brain undergoes a progressive WM degeneration characterized firstly by increased MD, followed by declines in FA and reduction of hippocampal connectivity.
The modest but significant increase of MD found in aMCI (both MCSA and ADNI cohorts) supports the construct that increased WM water diffusivity constitutes an early neurodegenerative event associated to AD pathophysiological processes. In fact, high MD has been frequently reported in aMCI populations [41
]. In contrast using TBSS, Agosta and colleagues only found widespread changes in axial diffusivity but no significant MD differences in 15 aMCI individuals, however, these disparities could be explained by methodological issues such as MRI magnetic strength as well as low in-plane resolution [47
Absence of FA reduction in aMCI, as reported here, is consistent with previous studies using similar methodology (i.e. same MCI inclusion criteria, 3T acquisitions and strict TBSS analysis) [41
]. However, FA declines in MCI have been previously reported using VOI and voxel-based techniques [43
]. Such discrepancy might be associated with low sample size, diagnostic criteria for MCI, analytical methods (VOI vs. voxel-based) or correction for multiple comparisons (no correction, FDR, random field).
At least two previous studies focusing on FA and MD abnormalities in naMCI provide conflicting results [42
]. In fact, older age and the presence of multiple pathologies could account for these conflicting results [60
The two-fold increase of estimated WM showing increased MD in demented patients (as compared to aMCI) supports the hypothesis that WM pathology also progresses in AD. Higher estimated WM showing lower FA in demented individuals in comparison with aMCI or naMCI further corroborates WM progressive degeneration in AD. Reduced FA in AD has been previously reported using VOI and voxel-based techniques [52
] In fact, progression of WM abnormalities has been previously suggested by cross sectional and longitudinal DTI studies [41
]. As a whole, these findings suggest increased MD as an early WM degenerative event in AD.
In post-mortem tissue, increase of tissue water diffusivity, as measured as m2
, might occur due the reduction of tissue barriers imposed by various causes (i.e. tissue atrophy, neuroinflammation), however empirical evidence supporting this claim is undermined by the effects of tissue fixation typically utilized in post-mortem / DTI correlation studies [11
]. Possibly increased MD in aMCI might represent an early event of WM degeneration present in AD pathophysiology secondary to the reduction of brain extracellular matrix and shrinkage of WM tissue in preclinical AD [70
In addition to the axonal degeneration originating due to death of cortical cell lesions, recent advances in pathophysiological mechanisms underlying FA and MD abnormalities described in AD can be partially attributable to independent WM neurodegeneration. A growing body of literature suggests that WM DTI abnormalities might be secondary to neuroinflammatory factors [72
]. Corroborating this hypothesis, imaging studies with PET and molecular imaging agents show an increase of microglia activation and astrocytosis in the WM of aMCI [73
]. In addition, it is noteworthy that ventricular enlargement (a surrogate of WM atrophy) is able to accurately distinguish between MCI and AD and has been proposed as tool for measuring disease progression in the short term [76
We also demonstrated hippocampal WM connectivity abnormalities in AD dementia. Hippocampal connectivity abnormalities affected hippocampal connections predominantly to the temporal, parietal, occipital and frontal polymodal associative areas (). The structural connectivity outcome measure described here (HSC) represents the probability of a given brain area to be connected with the hippocampal seed region. The hippocampal seed typically encompasses all cornus amonius (CA) sectors, the dentate gyrus and subiculum, since the limits between these areas remains below the resolution of the present DTI acquisition. For example, connectivity declines observed in the inferior occipitofrontal and superior longitudinal fascicles possibly reflect hippocampal connectivity declines mediated via subiculum, which constitutes a critical hub between the hippocampal system and the cortex [77
]. WM areas with significant decline of probability to be connected to the hippocampus were interpreted as depleted from normal hippocampal connectivity.
The results from our HSC CN maps are in excellent agreement with the post-mortem data describing hippocampal connectivity [83
]. Individual connectivity maps generated by this study in CN captured the classical reciprocal connections between the hippocampal system and the temporal, cingulate, inferior parietal or frontal cortices. Since the nature of DTI does not permit the inference of directionality, one cannot discriminate between hippocampal-petal and hippocampal-fugal fibers. Areas with reduced structural connectivity revealed by the HSC technique in AD was consistent with functional disconnections frequently reported by the literature. For example, reduced hippocampal connectivity in the cingulate bundle might explain functional disconnection reported between the hippocampus and the posterior cingulate / precuneus frequently described by numerous AD rsfMRI studies [85
]. Moreover, the WM connectivity reduction of the arcuate fascicles, inferior longitudinal, uncinate fascicles and superior longitudinal fascicles is a possible mechanism underling the [18
F]FDG signature of AD (hypometabolism in the posterior cingulate, inferior parietal, temporal and prefrontal cortices) [87
]. Hippocampal WM disconnections as described here corroborate the theoretical framework that emphasizes cortical disconnections as a key feature of AD [90
]. In fact, neuropathological, functional neuroimaging and neuropsychological evidence indicate WM disconnections as a pathophysiological mechanism involved in AD.
The relative integrity of hippocampal connections in aMCI or naMCI, as predicted in our hypothesis, is supported by histopathology evidence showing tangle pathology and WM disconnection affecting predominantly transentorhinal projections to the dentate gyrus (perforant path) in aMCI, while projections from the hippocampus and subiculum and the rest of entorhinal cortex are affected in more advanced stages of the disease [2
]. Possibly, a seed point in the transentorhinal cortex could better capture perforant path depletion in MCI stage, however transentorhinal connectivity is beyond the scope of this specific study.
Some methodological issues limit the interpretations of the present study. Since this is a cross sectional study, inferences regarding progression of WM pathology in the spectrum of AD clinical manifestations are speculative. The hypothesis posing MD increases as an early WM AD pathophysiology change followed by FA declines should be assessed by appropriate longitudinal studies.
Vascular pathology is certainly a potential confounder in all studies of this nature, since vascular insults and small vessel disease constitute a frequent finding in the MRI of elderly populations. For example, white matter intensities detected in T2 or FLAIR images may have an impact on various DTI outcome measures. However, since all patients recruited in this study, had Hachinski scores lower than 4, the impact of these lesions are unlikely to be clinically significant. In contrast the impact of vascular pathology on DTI outcomes is always a limitation for dementia studies since Hachinski score < 4 does not preclude vascular lesions. Particularly on the MCSA cohort, the presence of white matter intensities was minimal and monitored with 3D FLAIR MRI. Since the results obtained from the MCSA and ADNI cohorts were identical, it seems that vascular pathology affects these two populations in a similar fashion.
Although ADNI provides a powerful database for AD research, the drawback of utilizing DTI acquisitions acquired in 14 different scanners might represent a limitation for this study. While there were large efforts taken to cross-validate MRI scanners, multiple scanners acquisition is an undeniable confounding factor. Although, using the same analytical pipeline, the MCSA cohort yielded higher FA and MD control values in comparison with the same outcome measures from ADNI cohort. High FA and MD values on the MCSA cohort were obtained due to higher signal to noise ratio from the utilization a 32-channel head coil as well as 99 diffusion directions.
Regarding statistical analysis, TBSS is a conservative but extensively utilized method to compare WM change in numerous experimental populations [42
]. Particularly in the case of this study, the results using voxel-based non-parametric statistics provide similar results to TBSS (data not shown). Analytical protocols can potentially constitute a bias particularly for those studies utilizing voxel-based parametric statistical analysis without correcting for multiple comparisons.
In conclusion, we found in aMCI WM abnormalities are characterized by high MD, which are possibly secondary to brain inflammatory changes or WM axons or myelin content depletion. Furthermore, the concomitance of MD and FA abnormalities observed in AD suggests higher degree of WM microstructural lesion, which impacts in large-scale brain structural connectivity. Further longitudinal studies are necessary to corroborate whether a progression of WM disease occurs in the spectrum of clinical manifestation of dementia.