Abstract
Background:
Different clinical syndromes can arise from Alzheimer’s disease (AD) neuropathology, including dementia of the Alzheimer’s type (DAT), logopenic primary progressive aphasia (lvPPA), and posterior cortical atrophy (PCA).
Objective:
To assess similarities and differences in patterns of white matter tract degeneration across these syndromic variants of AD.
Methods:
Sixty-four subjects (22 DAT, 24 lvPPA, and 18 PCA) that had diffusion tensor imaging and showed amyloid-β deposition on PET were assessed in this case-control study. A whole-brain voxel-based analysis was performed to assess differences in fractional anisotropy, mean diffusivity, axial diffusivity, and radial diffusivity across groups.
Results:
All three groups showed overlapping diffusion abnormalities in a network of tracts, including fornix, corpus callosum, posterior thalamic radiations, superior longitudinal fasciculus, inferior longitudinal fasciculus, inferior fronto-occipital fasciculus, and uncinate fasciculus. Subtle regional differences were also observed across groups, with DAT particularly associated with degeneration of fornix and cingulum, lvPPA with left inferior fronto-occipital fasciculus and uncinate fasciculus, and PCA with posterior thalamic radiations, superior longitudinal fasciculus, posterior cingulate, and splenium of the corpus callosum.
Conclusion:
These findings show that while each AD phenotype is associated with degeneration of a specific structural network of white matter tracts, striking spatial overlap exists among the three network patterns that may be related to AD pathology.
INTRODUCTION
Alzheimer’s disease (AD) is a progressive neurodegenerative disorder characterized pathologically by the deposition of amyloid-β (Aβ) senile plaques and tau neurofibrillary tangles [1, 2]. Dementia of the Alzheimer’s type (DAT) is the most common clinical manifestation of AD, characterized by early and prominent episodic memory impairment [3]. However, subjects with AD can also present with other more atypical clinical syndromes, including the logopenic variant of primary progressive aphasia (lvPPA) [4] and posterior cortical atrophy (PCA) [5]. Clinically, the syndromes of DAT, lvPPA, and PCA present quite differently. DAT is characterized by a predominance of episodic memory impairment; lvPPA is characterized by deficits in language, with anomia, difficulty repeating sentences, and phonological errors [4]; and PCA is characterized by early onset visuospatial and perceptual deficits and is often associated with features of the Balint’s syndrome and the Gerstmann syndrome [6]. The vast majority of subjects diagnosed with either lvPPA or PCA have AD pathology on autopsy or on amyloid-PET imaging [5, 7–9], although it appears as though these atypical clinical presentations of AD are associated with different distributions of neurofibrillary tangles compared to subjects with DAT [10].
Gray matter atrophic patterns in these diseases have been previously studied, demonstrating volume loss in the left temporoparietal region for lvPPA versus atrophy predominantly in the primary visual cortex and visual association cortex for PCA [11–13]. These patterns differed from those typically observed in subjects with DAT, with DAT showing relatively greater involvement of the medial temporal lobe [12, 13]. While patterns of grey matter atrophy have been well studied across these clinical variants of AD, little is known about whether they show different underlying patterns of damage to the white matter. In this study we aimed to use diffusion tensor imaging (DTI) to investigate regions of white matter tract degeneration associated with these three AD syndromes, including those regions common to all three diseases and those specific to each. DTI measures the restricted diffusion of water along white matter tracts and allows the assessment of a number of different white matter metrics that have been hypothesized to measure different aspects of white matter pathology, including mean diffusivity (MD), fractional anisotropy (FA), axial diffusivity (AxD), and radial diffusivity (RD). One previous study analyzed white matter atrophy in these diseases using voxel-based morphometry [14], but DTI may provide more accurate imaging of white matter tracts. Better characterization of white matter degeneration in DAT, lvPPA, and PCA could lead to an improved understanding and a more precise radiologic diagnosis of these clinically challenging syndromes.
METHODS
Subject recruitment
We identified 24 subjects fulfilling clinical criteria for lvPPA [4] and 18 subjects fulfilling clinical criteria for PCA [5] who had Aβ deposition on Pittsburgh Compound B (PiB) PET imaging. All 42 subjects had been prospectively recruited from the Department of Neurology, Mayo Clinic, between 1 November 2010 and 1 March 2014, and had undergone a detailed neurological examination, a speech-language assessment, and a neuropsychological test battery, as previously described [15]. Subjects were excluded if they had a chief complaint of memory impairment or if memory impairment was identified on cognitive testing as a prominent deficit at disease onset or presentation.
We also identified two comparison cohorts: 1) a cohort of 22 subjects that presented with memory loss, met established criteria for DAT [3] and demonstrated Aβ deposition on PiB PET scanning, and 2) a cohort of 22 cognitively normal individuals who did not show evidence of Aβ deposition on PiB PET scanning. The DAT subjects were recruited by the Mayo Clinic Alzheimer’s Disease Research Center (ADRC) and were selected from the ADRC database to match the lvPPA and PCA cohorts on age and gender. None of the DAT subjects met clinical criteria for lvPPA [4] or PCA [5]. The cognitively normal subjects were recruited by the Mayo Clinic Study of Aging (MCSA), a population-based study, and were selected from the MCSA database to match the lvPPA and PCA cohorts on age and gender. The DAT and cognitively normal subjects had all undergone the identical MRI and PiB PET imaging protocol to the lvPPA and PCA subjects, and had undergone neuropsychological testing.
The study was approved by the Mayo Clinic IRB. All patients consented for enrolment into the study.
PiB-PET acquisition and analysis
All subjects underwent PiB PET imaging using a PET/CT scanner (GE Healthcare, Milwaukee, Wisconsin) operating in 3D mode. Subjects were injected with PiB (average, 614 MBq; range, 414–695 MBq) and after a 40–60 min uptake period, a 20-min PiB scan was obtained consisting of four 5-min dynamic frames following a low dose CT transmission scan. Standard corrections were applied. Individual frames of the PiB dynamic series were removed if motion was detected. Each PiB scan was co-registered to the MPRAGE for each subject using 6 degrees-of-freedom affine registration. An in-house modified version of the Automated Anatomical labeling atlas was then transformed into the native space of each subject using the inverse spatial normalization parameters from performing SPM5 unified segmentation on each subject’s magnetization-prepared rapid gradient-echo (MPRAGE) scan. A global PiB retention summary was generated by calculating median uptake for 6 regions of interest (ROIs): temporal, parietal, posterior cingulate/precuneus, anterior cingulate, prefrontal, and orbitofrontal cortex (left and right were combined for all ROIs). These regions were originally chosen to reflect the distribution of Aβ in the brain in DAT [16]. Median PiB uptake in each region was divided by median uptake in cerebellar grey matter to create uptake ratios, and a global PiB standardized uptake value ratio (SUVR) was formed by calculating median uptake ratio values across all 6 regions. The global PiB SUVR was used to classify each subject as PiB-positive using a global cortical-to-cerebellar ratio cut-point of 1.5 [16].
MRI acquisition
All subjects in the study underwent an identical 3T MRI acquisition protocol on a GE scanner that included a MPRAGE and a DTI sequence. The DTI acquisition consisted of a single-shot echo-planar (EPI) pulse sequence in the axial plane, with TR = 10,200 ms; in-plane matrix 128/128; FOV 35 cm; phase FOV 0.66; 41 diffusion encoding directions and five non-diffusion weighted (b0) T2w images; slice thickness 2.7 mm (2.7 m isotropic resolution). Parallel imaging with a SENSE factor of two was used.
DTI analysis
Each volume of the DTI image was registered to the first volume (a b0) using affine transformations to correct for head motion and minimize distortions due to eddy currents. Images were brain-extracted and FA, MD, AxD, and RD maps were generated using linear least-squares optimization with the FMRIB Software Library (FSL) version 4 software [17]. A whole-brain voxel-based analysis (VBA) was performed on the FA, MD, AxD, and RD images to determine a set of voxels with statistically significant intensity differences across groups by using a specific in-house DTI-VBA implementation previously described and demonstrated to provide superior sensitivity and specificity versus tract-skeleton-based approaches [18]. In brief, FA images of all subjects were nonlinearly coregistered via an iterative, groupwise registration algorithm (advanced normalization tools [19]) and normalized to a 1 mm isotropic Montreal Neurological Institute (MNI) 152 standard space via the FMRIB58_FA template [17]. The other DTI metrics (MD, AxD, and RD) were also analyzed by projecting these inputs through the previously calculated FA-based transformations prior to performing voxel-based comparisons. To increase the validity of the Gaussian field assumptions used in voxel-wise parametric tests, the images were smoothed using a Gaussian kernel with 8 mm FWMH. Finally, regions of cerebrospinal fluid and gray matter were removed from consideration by masking out regions where the mean FA across coregistered subject images was below 0.2.
SPM5 was used to analyze the smoothed DTI images at the voxel-level. A full-factorial model was used to compare lvPPA, PCA, and DAT to controls, assessed after correction for multiple comparisons using the false discovery rate (FDR) correction at p < 0.05. A masking analysis was performed in order to identify regions of degeneration that were common to lvPPA, PCA, and DAT when compared to controls. In addition, statistical analyses directly comparing the lvPPA, PCA, and DAT groups were performed and assessed at p < 0.05 uncorrected for multiple comparisons using an extent threshold of 200 voxels to exclude small clusters, as well as at p < 0.05 using FDR correction. All analyses were performed separately using each of the different DTI metrics (FA, MD, AxD, RD). Age and gender were included in all analyses as covariates. The Johns Hopkins University (JHU) single-subject white matter atlas was registered to MNI space and overlaid on the resultant statistical maps in order to aid in the identification of specific white matter tracts and in order to output mean regional values.
Statistical analyses
Subject demographics and clinical data were compared across groups using Wilcoxon Rank Sum tests for continuous data and Chi-squared tests for categorical data. Statistical analyses were performed utilizing the JMP computer software (JMP Software, version 8.0; SAS Institute Inc., Cary, NC) with α set at 0.05.
RESULTS
Study demographics
No differences were observed across groups in age at MRI, age at onset, gender, or education (Table 1) (by design for AD and control subjects). By design, the PiB-PET SUVRs were lower in controls but consistent across disease groups. Disease duration was also relatively consistent across groups, although was somewhat lower in lvPPA compared to DAT. Mini-Mental State Examination scores [20] were higher in controls but did not differ significantly between disease groups. On neuropsychological testing, DAT performed the worst on the Auditory Verbal Learning Test delayed recall [21], lvPPA performed the worst on the Boston Naming Test [22], and PCA performed the worst on the Rey-Osterrieth complex figure test [23].
White matter tracts showing degeneration across all three syndromes
All three disease groups showed bilaterally reduced FA in the fornices, and body of the corpus callosum, as well as the left posterior thalamic radiation, when compared to controls (Fig. 1). There was increased MD in these same regions, as well as significant bilateral temporal lobe, moderate parietal lobe, and minimal frontal lobe white matter involvement. Specific tracts with increased MD bilaterally included the fornix, internal capsule, entire corpus callosum, cingulum, inferior fronto-occipital fasciculus (IFOF), inferior longitudinal fasciculus (ILF), superior longitudinal fasciculus (SLF), and uncinate fasciculus (UF). AxD and RD showed overlap across groups in the same regions as MD (Supplementary Figure 1).
White matter tract degeneration in DAT
Overall, DAT subjects showed reduced FA in several tracts compared to controls (Fig. 1), including bilateral hippocampal cingulum, fornix, superior temporal white matter, posterior thalamic radiation, and body of the corpus callosum, as well as left superior parietal white matter. MD was increased in the same tracts with additional involvement of bilateral parietal white matter and minimally in the bilateral frontal white matter. Also, the entire corpus callosum, corona radiata, internal capsule, UF, ILF, IFOF, and SLF had bilaterally increased MD. AxD and RD showed increased signal in the same regions as MD (Supplementary Figure 1).
Only a few regions, predominantly in the right SLF, showed reduced FA in DAT compared to lvPPA (Fig. 2). Similarly, tracts where DAT had increased MD compared to lvPPA were mainly in the right hemisphere, including ILF, IFOF, cingulum, SLF, corona radiata, splenium of the corpus callosum, and posterior thalamic radiation. The fornix was bilaterally involved. Very few regions showed greater involvement in DAT compared to PCA, with only some mild regions of increased MD observed in the frontal white matter and right anterior corona radiata, anterior limb of the internal capsule, and genu of the corpus callosum (Fig. 2). The AxD and RD results were similar to the MD findings (Supplementary Figure 2). None of these disease group differences survived after correction for multiple comparisons using FDR.
White matter tract degeneration in lvPPA
The lvPPA subjects showed reduced FA predominantly in the temporal white matter, UF, IFOF, and posterior thalamic radiation, particularly in the left hemisphere, compared to controls (Fig. 1). Asymmetry scores for the UF, IFOF, and posterior thalamic radiation are shown in Table 2. Asymmetry was evident in the FA of the UF and MD of the posterior thalamic radiation, with greater involvement of the left hemisphere. The body of the corpus callosum, cingulum, and fornix had bilateral FA reduction. MD was increased in the same regions, with additional involvement of the ILF, left parietal white matter, SLF, internal capsule, and corona radiata. AxD and RD showed increased signal in the same regions as MD (Supplementary Figure 1).
The lvPPA subjects showed decreased FA in temporal and frontal white matter tracts, predominantly in the left hemisphere, including the UF, IFOF, anterior cingulum, internal capsule, and genu of the corpus callosum, compared to DAT (Fig. 3). Regions of increased MD in lvPPA compared to DAT were only observed in the left temporal white matter. The lvPPA subjects showed decreased FA and increased MD in the left UF and IFOF compared to PCA (Fig. 3), with additional regions of increased MD observed in genu of the corpus callosum and anterior cingulum. AxD and RD showed no additional differences between these groups (Supplementary Figure 2). None of these disease group differences survived after correction for multiple comparisons using FDR.
White matter tract degeneration in PCA
The PCA subjects demonstrated widespread regions with bilateral FA reduction and MD increase with a notable posterior predilection compared to controls (Fig. 1). The posterior thalamic radiation was the most significantly affected region with more signal than seen in other disease groups. Other involved regions included the superior parietal and occipital white matter, SLF, IFOF, ILF, UF, fornix, cingulum, and splenium of the corpus callosum. Increased AxD and RD were seen in similar regions (Supplementary Figure 1).
The PCA subjects showed widespread abnormalities in both FA and MD compared to the other disease groups at both uncorrected and FDR corrected statistical thresholds (Fig. 4), predominantly involving white matter tracts in the parietal and occipital lobes, including the SLF, posterior thalamic radiations, splenium of the corpus callosum, and posterior cingulate, as well as posterior regions of the ILF and IFOF. Regions that showed greater involvement in PCA compared to lvPPA showed more severe involvement of the right hemisphere. Similar differences were observed with AxD and RD (Supplementary Figure 2).
DISCUSSION
We have described DTI patterns of white matter degeneration in three syndromic variants of AD using relatively large well-matched cohorts of patients with DAT, lvPPA, and PCA. DTI metrics including FA, MD, AxD, and RD were evaluated in our analysis. We identified specific white matter tracts that were involved across all three AD groups, regardless of presenting syndrome, as well as some subtle regional differences across the variants that are likely associated with the differing clinical presentations.
A network of white matter tracts was involved across DAT, lvPPA, and PCA. Abnormalities in all DTI metrics were observed in the fornices, corpus callosum, and posterior thalamic radiations, suggesting that these regions are particularly associated with AD pathology. Although previous studies have associated degeneration of the fornix specifically with DAT and episodic memory loss [24, 25], our findings suggest that degeneration of this tract is not specific to DAT. It is possible that fornix involvement may reflect the fact that episodic memory loss, while severe in DAT, was also observed to a lesser degree in lvPPA and PCA, as others have reported [26, 27]. Degeneration of the corpus callosum is commonly observed on autopsy in AD and likely reflects Wallerian degeneration of interhemispheric fibers [28]. The finding of common involvement of the posterior thalamic radiations is more surprising, given that the regions of reduced FA were identified in occipital regions, which are typically involved primarily in PCA. Involvement of these tracts may reflect the fact that the disease spreads posteriorly in all three of these AD variants, although further studies will be needed to confirm the finding. More widespread patterns of overlap were observed with MD, AxD, and RD, with additional involvement of the temporal and parietal white matter, including IFOF, ILF, UF, and SLF. These findings are largely concordant with previous grey matter studies showing that atrophy of the temporal and parietal cortices provide excellent biomarkers of the presence of AD pathology, regardless of syndrome [29]. Our findings suggest that temporal and parietal white matter tract measures may also provide useful biomarkers for AD. One previous study that assessed atrophy of the white matter also observed regions of common involvement in DAT, lvPPA, and PCA predominantly in the temporal and parietal lobes [14].
With regard to DAT, the most notable finding was the predominance of right-sided disease compared to lvPPA. This illustrates that lvPPA tends to spare right hemisphere tracts, while DAT demonstrates more bilateral disease. Additionally, tracts involved in memory such as the fornix and cingulum were slightly more severely affected in DAT than they were in lvPPA, likely contributing to the former’s predominance of memory loss. The parahippocampal components of the cingulum were also affected in DAT, consistent with previous DTI studies [24, 30]. Very few regions showed greater involvement in DAT compared to PCA, reflecting the general severity of white matter degeneration in PCA. There was, however, some very subtle evidence that DAT showed relatively greater involvement of frontal white matter tracts compared to PCA.
Several interesting patterns of disease were also seen in the lvPPA group. Degeneration was bilateral but asymmetric, with greater involvement of left-sided white matter tracts, relative to controls, concordant with the fact that these subjects present with language impairment. This finding persisted when lvPPA was compared to the other syndromes. Degeneration was primarily observed in temporal, parietal, and frontal tracts, including the UF, ILF, IFOF, and SLF, consistent with previous DTI studies in lvPPA [31, 32]. The lvPPA subjects showed more degeneration than both the DAT and PCA groups in the left UF and IFOF in the temporal lobe and the anterior cingulate and genu of the corpus callosum in the frontal lobe. These tracts therefore appear to be predominantly associated with lvPPA and may serve as important biomarkers of lvPPA. The IFOF has indeed been previously shown to play an important role in language comprehension and degeneration of this tract may reflect disruption of the language functional network [33].
The PCA group, as compared to controls, was mostly characterized by degeneration of posterior parietal and occipital white matter tracts. Although many association fibers were involved, there was a tendency for the posterior portion of any given tract to be most severely affected. Our findings are consistent with VBM studies of gray and white matter atrophy in PCA, which show volume loss with a posterior predominance [11, 14] and a couple of DTI studies that have similarly shown involvement of parietal and occipital white matter tracts [34, 35]. Among the tracts identified, the posterior thalamic radiations that project into the occipital lobe showed the most significant degeneration bilaterally. We also observed striking involvement of the SLF, posterior cingulate, splenium of the corpus callosum, and posterior ILF and IFOF. These tracts were, in fact, involved to a greater degree in PCA compared to both DAT and lvPPA, suggesting they may have diagnostic utility and likely contribute to the visuospatial and visuoperceptual deficits observed in PCA patients. It is important to stress that the PCA, lvPPA and DAT groups were relatively well matched for general cognitive severity, age, and disease duration, and hence differences observed across groups were unlikely driven by imbalances in thesevariables.
In general, we observed that MD changes across the AD syndromes were more extensive and widespread than the changes in FA, although there was overlap across the two DTI metrics with MD tending to identify the same tracts as FA but also multiple additional tracts. It therefore appears that MD is a more sensitive measure of white matter tract degeneration in these disorders, possibly because FA measures lack sensitivity in regions of crossing fibers or because MD measures may be more affected by partial volume averaging due to atrophy [36]. Also, our DTI acquisition had only 41 directions and relatively large voxel sizes. A longer, more intensive DTI acquisition with smaller voxel size and greater angular resolution might have led to different conclusions about the relative value of FA versus MD. The findings with AxD and RD were very similar to the MD findings and hence did not appear to add any additional information. These findings concur with previous studies involving DTI in neurodegenerative diseases [37], and with studies that have similarly found more widespread MD than FA findings in lvPPA [31] and DAT [38]. A major strength of our study was the large sample of well-matched subjects with DAT, lvPPA, and PCA that all had Aβ deposition on PiB-PET supporting an underlying neuropathological diagnosis of AD. Our findings should generalize to other PiB-PET studies of these syndromes. The use of the voxel-based analysis approach to analyze FA, MD, AxD, and RD was also a major strength. This approach is more accurate and sensitive than traditional skeleton-based techniques, such as tract-based spatial statistics [39] with greater specificity in resisting false positives from misregistration [18]. While the comparisons between each disease group and controls survived a correction for multiple comparisons, the differences across disease groups were only observed at a relatively lenient uncorrected statistical threshold. The FDR correction controls the expected proportion of rejected null hypotheses that were false positives. Therefore, we cannot rule out the possibility that some of the uncorrected findings may reflect false positives or type I errors. It is typically difficult to identify findings that survive a correction for multiple comparisons when comparing two inherently variable disease groups, with findings being more powerful when disease groups are compared to a less variable control cohort. The fact that the results did not survive correction for multiple comparisons may argue that there is a large degree of anatomical overlap across these AD spectrum disorders.
Our study provides neuroanatomical data detailing structural connectivity in three syndromic variants of AD with the precision of DTI. The findings show that while each AD phenotype is associated with degeneration of a specific structural network of white matter tracts, significant spatial overlap exists among the three network patterns, which may be related to the underlying AD pathology. Not only do these findings contribute to our understanding of underlying disease mechanisms, but they may also have value as imaging biomarkers that supplement clinical diagnoses. White matter tract measures should, therefore, be considered for future clinical trials targeting these AD variants. It will be important, however, for future studies to investigate the progression of both gray matter atrophy and white matter tract degeneration early in these syndromes in order to elucidate whether gray and white matter changes occur concomitantly or in succession, which may further our understanding of the underlying disease mechanisms in AD.
Footnotes
ACKNOWLEDGMENTS
The study was funded by National Institute of Health grants R01 DC010367 to KAJ, R01-AG11378 and R01-AG041851 to CRJ, R01-AG040042 to KK, and P50-AG016574 to RCP, Alzheimer’s Association grant NIRG-12-242215 to JLW and the Elsie and Marvin Dekelboum Family Foundation. We would like to acknowledge Drs. Bradley Boeve, David Knopman, Robert Ivnik, and Glenn Smith, Mayo Clinic Rochester, MN, who performed clinical or neuropsychological evaluations on some of the DAT and healthy control subjects included in the manuscript.
