Abstract
Keywords
INTRODUCTION
Prion diseases are a phenotypically heterogeneous group of neurodegenerative disorders caused by the deposition of misfolded prion proteins (PrPSc) in the brain. Creutzfeldt-Jakob disease (CJD), its most common form, is characterized by progressive dementia, ataxia, and myoclonus, with spongiform change, neuronal loss, gliosis, and PrPSc deposits widely distributed throughout the brain. This phenotype differs from fatal insomnia (FI), another prion disease that can be either sporadic (MM2T or thalamic form of CJD) or genetic (caused by D178N substitution in the prion protein gene), and is distinguished by disrupted circadian rhythms, cognitive deterioration, dysautonomic and motor abnormalities, with severe thalamic, inferior olivary, and cerebellar degeneration [1].
While visual assessment of computed tomography scans or T1-weighted magnetic resonance imaging (MRI) is usually normal or shows nonspecific atrophy at late stages of prion diseases [2, 3] diffusion-weighted imaging (DWI) has demonstrated high sensitivity for the identification of hyperintense signal in cortical and subcortical grey matter (GM) of patients with CJD [4, 5]. This contrasts with the lack of specific structural neuroimaging findings in FI [1, 6]. Diffusion tensor imaging (DTI) and quantitative volumetric MRI have been used to characterize and quantify brain changes in different neurological disorders [7–9]. These tools provide relevant information concerning both macro and microstructural changes, might reveal specific alterations that cannot be captured by naked eye visual assessment and can potentially be used as markers of disease progression in clinical trials [10]. Scarce studies have applied these methodologies in prion diseases and several have used a region-of-interest (ROI)-based methodology, which is suitable for hypothesis-driven studies, but entails the potential risk of overlooking regional changes that occur beyond ROIs[7, 11–15].
In this study, we use a whole-brain, voxel-based approach, in combination with a ROI-based analysis, to investigate volumetric and DTI changes in patients with prion disease with either CJD or FI phenotype, and correlate them with neuropathological features. Our aim is to understand the in vivo histopathological changes that occur in the brain of these patients. This information may help in the interpretation of voxel-based morphometry (VBM) and DTI data in case these tools are used as biomarkers for priondiseases.
MATERIALS AND METHODS
Participants
Twenty subjects with definite or probable prion diseases according to the World Health Organization and/or the European Consortium diagnostic criteria [5, 16] were prospectively recruited at the Alzheimer’s and other Cognitive Disorders Unit of Hospital Clínic de Barcelona between February 2011 and April 2015. Participants underwent an extensive work-up for rapidly progressive dementias. Clinical evaluation included Mini-Mental State Examination (MMSE) and MRC Prion Disease Rating Scale [17, 18]. PRNP genotype was determined by direct sequencing. According to their clinical phenotype (irrespective of their genetic status), patients were classified as CJD or FI, and analyzed separately. The rationale for pooling genetic and sporadic cases together in a same group is based on the aim to evaluate clinico-pathologically defined phenotypes, independently of their cause, and on that patients with genetic and sporadic CJD present a similar clinico-radiological phenotype, as patients with sporadic and familial FI do [1, 4]. Forty age- and gender-adjusted healthy volunteers were enrolled as controls. All participants signed an informed consent before their enrolment in the study, which was approved by the Hospital Clínic EthicsCommittee.
Image acquisition and analysis
MRI scans were performed on a 3T MRI scanner (Magnetom Trio Tim, Siemens Medical Systems, Germany). A high resolution 3D structural T1-weighted dataset (MP-RAGE, TR = 2300 ms, TE = 2.98 ms, 240 slices, FOV = 256 mm, matrix size = 256×256, slice thickness = 1 mm) was acquired for all subjects. The DTI protocol consisted on an echo-planar imaging (EPI) sequence with 30 directions +b0 image, with 2 repeated acquisitions, TR = 7600 ms, TE = 89 ms, 60 slices (thickness = 2 mm), distance factor = 0%, FOV = 250 mm, matrix size = 122×122, voxel size = 2×2×2 mm. All images were visually inspected to ensure their quality. Due to motion artifacts, four subjects (three CJD, one FI) were excluded from the volumetric analysis and two subjects with CJD from the DTI analysis. Images were reoriented along the anterior–posterior commissure line. VBM analysis was performed with SPM12 (Statistical Parametric Mapping, Welcome Trust Centre for Neuroimaging, UK; http://www.fil.ion.ucl.ac.uk/spm). Total GM, white matter (WM), cerebrospinal fluid (CSF), and total intracranial volume (TIV) were calculated based on the tissue probability maps obtained with the new segment tool. Images were normalized to the Montreal Neurological Institute space using diffeomorphic anatomical registration through exponentiated lie algebra (DARTEL) and smoothed convolving by an Isotropic Gaussian kernel with 8 mm of FWHM [19]. The two runs of DTI images acquired for each subject were concatenated in a single file. Next, images were corrected for eddy current distortion and head motion, and registered to 3D T1 using FLIRT and FNIRT (as part of the package FMRIB Software Library (FSL); http://www.fmrib.ox.ac.uk/fsl). Fractional anisotropy (FA), mean diffusivity (MD), axial diffusivity (AD), and radial diffusivity (RD) maps were obtained from whole-brain images using FSL. Skeletonized maps of DTI indices were obtained using the Tract-Based Spatial Statistics (TBSS) tool implemented in FSL [20]. Significant clusters were identified using the Johns Hopkins University WM tractography atlas and the International Consortium of Brain Mapping ICBM–DTI WM labels atlas [21]. Averaged values of FA, MD, AD, and RD were extracted from clusters with significant differences. Mean MD and volume values were extracted from ROI masks of thalamic nuclei (anterior, lateral dorsal, lateral posterior, medial dorsal, midline, ventral anterior, ventral lateral, ventral posterior lateral and ventral posterior medial nuclei, lateral and medial geniculate bodies, and pulvinar). These masks were generated using WFU_pickatlas software implemented in SPM12, based on the Talairach Daemondatabase [22, 23].
Neuropathological assessment
Postmortem neuropathological brain examination was performed in 13 subjects at the Neurological Tissue Bank of the Biobanc-Hospital Clinic-IDIBAPS. Neuropathological diagnoses were established according to current standard diagnostic criteria. Prion diseases were classified based on their PrPres western blot pattern and/or PRNP genotype and histotype [24]. In 11 subjects (seven CJD and four FI), who also had available DTI data, a semiquantification of spongiform change, neuronal loss, gliosis and presence of PrPSc deposits (0 = absent; 1 = mild; 2 = moderate; 3 = extensive) was performed in four thalamic nuclei: anterior, medial dorsal and ventral lateral nuclei, and pulvinar.
Statistical analyses
Two-sample t-test was used to compare whole-brain VBM and DTI maps between CJD and FI groups and controls, with age and gender entered as covariates. Voxels with significant differences at an uncorrected threshold (punc <0.001) were identified and subsequently corrected for multiple comparisons using false discovery rate (FDR) at cluster level, corrected p-value (pcorr) <0.05, in a Gaussian Random field framework using SPM12. For TBSS analysis, the number of permutations was set at 5000, and age and gender were entered as covariates. Statistical significance was set at pcorr <0.05, after family-wise error (FWE) correction, using the threshold-free cluster enhancement option (TFCE), implemented in FSL [25]. Mann-Whitney U test was used to assess differences in volume and DTI metrics in thalamic ROIs between patients and controls, as well as differences in neuropathological features between CJD and FI groups. Linear regression was used to explore the influence of spongiform change, neuronal loss, gliosis, and PrPSc deposition on MD values. Time between MRI acquisition and neuropathological study was included in the analysis as a confounding factor. For these analyses, STATA 12 software (Stata Statistical Software: Release 12. StataCorp LP, Tex., USA) was used and significance threshold was set at an FDR adjusted p-value of 0.05, by using the Simes method for multiple hypothesis testing [26].
RESULTS
Patients
Prion disease was neuropathologically confirmed in 13 patients. Among the remaining seven, three carried mutations in PRNP (two E200K and one D178N) and the other four fulfilled the European CJD Consortium diagnostic criteria for probable sporadic CJD. Fifteen patients (13 sporadic CJD and two E200K mutation carriers) displayed typical CJD clinical features, while the other five (four D178N mutation carriers and one sporadic case with MM2T+C molecular subtype) presented a clinic-pathological FI phenotype. Clinical, demographic, and genetic data are presented in Table 1.
Whole-brain VBM analysis
In CJD, VBM revealed significant GM loss in bilateral thalamus (anterior, lateral dorsal, lateral posterior, medial dorsal, midline, ventral anterior, ventral lateral, ventral posterior lateral and ventral posterior medial nuclei, and pulvinar), putamen, fusiform gyrus, cerebellum, and the left perirolandic cortex (Fig. 1A and Table 2). WM loss was observed in corpus callosum in CJD patients at the uncorrected analysis (punc <0.001), but these differences disappeared with FDR correction. In FI, we observed significant volume loss in anterior-medial thalamus, cerebellum and medulla oblongata (Fig. 1B and Table 2). No areas of volume increase were identified in patients with FI or CJD compared with controls.
Whole-brain DTI analysis
MD, AD, and RD were significantly decreased in thalamus in CJD (MD was decreased in ventral lateral, ventral posterior lateral, ventral posterior medial and medial dorsal nuclei, and pulvinar), putamen, and right medial globus pallidus (Figs. 1C, 2A and C, Table 3). MD was increased (mainly at the expense of RD) in frontal, temporal, parietal and occipital lobes (predominantly involving WM), anterior corpus callosum, and cerebellum (Fig. 1D). FA was reduced in frontal, temporal, parietal and occipital lobes (mainly WM), corpus callosum, cingulate gyrus, cerebellum, thalamus (mainly posterior and medial thalamus), basal ganglia, hippocampus, amygdala, and parahippocampal gyrus (Fig. 2E). No areas of FA increase were detected.
In FI, MD was increased in bilateral thalamus (bilateral anterior, ventral anterior, ventral lateral and medial dorsal nuclei, and right pulvinar), right parahippocampal area, left cerebellar hemisphere, and midbrain (Fig. 1E and Table 3). FA reduction, as well as AD and RD increase, was observed, with similar distribution than MD. No areas with significant decrease in MD, AD, or RD, or significant increase in FA were identified.
VBM and whole-brain DTI analysis excluding genetic CJD and sporadic FI cases
In order to exclude a bias due to the inclusion of two cases with the E200K mutation in the CJD group and a sporadic case in the FI group, we repeated the VBM and whole-brain DTI analyses excluding these patients from their respective groups. The exclusion of genetic cases from the CJD group yielded slightly larger clusters in the same cerebral regions in the VBM and DTI analysis. In the FI group, significant differences in the VBM analysis remain but at a less exigent threshold (punc <0.005), which suggest that differences would be attributable to real group differences rather than to an individual effect. In the DTI analysis, the exclusion of the sporadic case was associated to a mild decrease in the clusters’size.
Tract-based spatial statistics analysis
In CJD, FA was significantly decreased in WM tracts from brain hemispheres, brainstem, and cerebellum (Fig. 3A). Within these clusters, averaged FA was decreased by 10.7%, MD and RD values were increased by 5.9% and 11.4% respectively, while AD was only increased by 0.7%, compared with controls. MD and RD were diffusely increased in hemispheric WM, and RD was also increased in brainstem and cerebellar WM tracts, while AD was increased in areas of the anterior corona radiata (Fig. 3B, C, E). In contrast, AD was significantly decreased in intrathalamic WM tracts, anterior and posterior limbs and retrolenticular part of internal capsule, cerebral peduncles, right superior cerebellar peduncle, and right external capsule (Fig. 3D).
In FI, RD was increased in thalamus, cerebellum, superior, middle and inferior cerebellar peduncles, fornix, and several right hemisphere subcortical regions (corona radiata, superior longitudinal fasciculus, internal and external capsule, and cerebral peduncle) (Fig. 3H). MD was increased in the body of corpus callosum and thalamus and in several right hemisphere locations (corona radiata, superior longitudinal fasciculus, internal and external capsule, and cerebral peduncle) (Fig. 3G). FA was decreased in right corona radiata, internal capsule, and cerebral peduncle, as well as in the splenium of corpus callosum (Fig. 3F). We did not observe areas with significant changes in AD values.
Thalamic nuclei ROI analysis
ROI analysis revealed volume loss in thalamic nuclei in CJD and FI compared with controls (Fig. 4). Differences were statistically significant in all thalamic nuclei in CJD. In FI, differences were statistically significant in the anterior, medial dorsal, and midline nuclei at the uncorrected analysis (pcorr <0.03), but not after FDR correction. Regarding DTI metrics, both patterns of increased and decreased MD were observed in CJD (Fig. 4A). In these patients, MD was significantly increased in anterior, lateral dorsal, midline and ventral anterior nuclei, while significantly reduced in ventral posterior medial and ventral posterior lateral nuclei (Fig. 4A). In FI, a significant increase in MD was observed in all thalamic nuclei except for midline nucleus(Fig. 4B).
Neuropathological evaluation and correlation with DTI metrics
Nine out of 13 patients with neuropathologically confirmed prion disease had sporadic CJD (two MM1, two VV2, two MV2K+C, one MV2+1, one MM2C+1, and one VV2+1). The remaining four had a FI phenotype (three with D178N mutation and one sporadic case with MM2T+C subtype). Table 4 shows the severity scores of neuropathological alterations in thalamic nuclei. CJD patients exhibited higher scores of spongiform change and PrPSc deposition than FI patients. These differences were statistically significant for PrPSc in all nuclei (pcorr <0.03) and for spongiform change in all but the anterior nucleus (pcorr <0.03). Additionally, a trend was observed toward higher neuronal loss (pcorr = 0.07) and gliosis (pcorr = 0.06) in FI in the anterior nucleus. In other nuclei, neuronal loss and gliosis were also higher in FI but differences were not statistically significant. Globally, spongiform change was more prominent in VV2 than in MM1 CJD subtype, and gliosis was frequently gemistocytic in VV2 compared to the chronic and fibrillary glial reaction seen in FI cases. Linear regression did not show any significant correlation between MD and individual neuropathological alterations in patients with prion diseases, despite reduced MD values were only observed in patients with CJD, who showed higher spongiform change and PrPSc deposition scores in thalamic nuclei than FI patients. However, this correlation was not evaluated in ventral posterior medial and ventral posterior lateral nuclei, as the protocol used for preparing brain slices did not allow a reliable identification of these thalamic subnuclei.
DISCUSSION
We quantified volumetric and DTI changes in patients with prion diseases with either CJD or FI phenotypes and found prominent thalamic alterations in both groups. In CJD, both decreased and increased MD was observed across different thalamic subnuclei. MD was also decreased in basal ganglia, and increased in subcortical WM and cerebellum. We report for the first time quantitative MRI findings in FI, which were characterized by increased MD in thalamus and cerebellum, overlapping with areas of GM loss.
VBM revealed significant atrophy in thalamus and cerebellum in CJD and FI. Volume loss was also present in putamen, fusiform gyrus, and left perirolandic cortex in CJD, and in medulla oblongata in FI. Structural imaging in prion disease is usually considered normal at visual inspection except for advanced cases. There are scarce volumetric studies in prion diseases and, to the best of our knowledge, no published studies of quantitative MRI data in FI [1]. Volume loss has been reported in cerebral cortex in the slowly evolving prion disease linked to 6 octapeptide repeat insertion in PRNP and in cerebellum in genetic CJD [11, 14]. In contrast, one study failed to show significant GM loss in subjects with sporadic CJD [7]. The use of a stronger magnet, which allows a higher signal-to-noise ratio and therefore a higher spatial resolution, and a voxel-based approach, which overcomes the possibility of diluting regional changes and missing regions that have not been considered in a priori hypotheses, may account for some differences between ours and previously reported findings, concerning volumetric changes.
Our findings of decreased MD in basal ganglia and thalamus in CJD are in line with previous studies [7, 13]. Both thalamic ROI-based and whole-brain analyses evidenced significantly decreased MD in ventral posterior medial and ventral posterior lateral nuclei. In contrast, the MD increase observed in anterior, lateral dorsal, midline and ventral anterior nuclei in the ROI-based analysis did not arise in the whole-brain analysis, although areas of increased MD were observed in the thalamus at the uncorrected maps (data not shown). Discordant MD changes were also found in medial dorsal nucleus when comparing whole-brain and ROI-based analyses. These discrepancies may be explained by methodological differences, as areas with significant MD changes in whole-brain analysis may be masked by surrounding areas. The coexistence of areas with decreased and increased MD in CJD is not surprising, as it has been reported that MD may increase during the disease course in brain areas that initially had restricted diffusivity, presumably as a consequence of parenchymal loss [7, 27]. Unexpectedly, we did not find significantly decreased MD neither in caudate nor in cerebral cortex, despite DWI hyperintensities are frequently observed in these locations in CJD. However, cortical alterations that are easily identified at an individual level may not be detected by quantitative group analysis, because they do not always occur in the same regions. Moreover, spatial normalization can result in a biased estimation of the morphology of cerebral structures, especially in patients with enlarged ventricles, [28] which might explain the absence of significant changes in caudate nucleus in our study, as its small size and periventricular location make it particularly susceptible to this undesired effect [29]. Finally, certain differences may lack statistical significance due to insufficient statistical power.
In CJD, whole-brain and TBSS analyses revealed increased MD, RD, and AD in subcortical WM, overlapping with areas of decreased FA. Decreased FA is thought to reflect pathological processes such as axonal loss or myelin degeneration. Increased RD has been associated with a disruption of the myelin sheath, while both decreased and increased AD values have been associated with WM tract alterations [9, 30]. The pattern of reduced FA with increased MD, RD, and AD observed in our study has also been reported in Alzheimer’s disease, and may reflect secondary WM degeneration due to primary neuronal loss [8]. Increased diffusivity has also been described in WM from patients with inherited prion diseases, [12, 14] but not in sporadic CJD. On the contrary, Caverzasi and collaborators found significant reductions of MD in WM from patients with sporadic CJD [15]. This finding was surprising, as reduced diffusivity in CJD has been associated with spongiform change and PrPSc deposition, [31] which are typically absent in WM from these patients. The authors also highlighted the presence of astrocytosis and microglial activation in WM, although these were not statistically correlated with reduced MD. Microglial activation was also found to be increased in sporadic CJD but not in FFI in a different study, which allows us to speculate that different patterns of microglial activation might also play a role in different patterns of MD changes in prion diseases [32]. In our study, we only found significantly decreased AD in intrathalamic WM tracts, internal capsule, superior cerebellar peduncle, and right external capsule, without significant changes in MD in those areas. Even if we do not have a solid explanation for the discrepancies between our results and those reported by Caverzasi and collaborators, differences in the mean time between symptoms onset and MRI (12 months in their study and 5.2 months in ours) and total disease duration (19 months in their study and 11.5 months in ours) could account for some of these differences, as more prolonged disease course might produce a higher degree of gliosis in WM. Therefore, assuming a hypothetical restriction in diffusivity caused by gliosis, at the time MRI scans were acquired in our study, the increase of diffusivity caused by axonal loss might have outweighed the diffusivity restriction caused by gliosis.
In FI, whole-brain analysis revealed increased MD, AD, and RD in thalamus, parahippocampal region, cerebellum, and midbrain, probably reflecting neuronal loss, as these alterations overlapped clusters with volume loss. Both VBM and DTI data evidenced diffuse thalamic damage, with higher involvement of anterior and medial dorsal nuclei, in line with previous studies [1]. TBSS analysis showed much more limited WM abnormalities compared to CJD.
The fact that MD was reduced in subcortical GM in CJD but not in FI, and spongiform change scores in thalamus were higher in CJD than in FI suggests that MD reduction in CJD may be predominantly caused by spongiform change, which is in line with previous studies [31, 34]. However, we did not find any significant correlation between neuropathological findings and MD values in the thalamus, probably because MD changes are influenced by competing processes (e.g., spongiform changes and neuronal loss) and that neuropathological changes likely evolved since MRI scanning until postmortem examination. In addition, we could not analyze those thalamic nuclei with significant MD reduction in CJD due to technical limitations.
The broad and remarkably variable anatomical distribution of the neuropathological alterations observed in CJD is a striking feature of this disease. In this regard, whether PrPSc initially spreads from a single location or starts simultaneously in multiple regions remains uncertain. However, clinical and experimental evidence point to thalamic involvement as an early event in prion diseases [35, 36]. Frequent clinical and cognitive features observed in CJD, such as myoclonus, episodic unresponsiveness, interference effects, verbal and motor perseveration, and the presence of periodic sharp wave complexes in electroencephalogram also suggest a critical role of the thalamus [37, 38] The thalamus conveys input and output information from different sensory modalities and regions involved in motor regulation, such as the cerebellum, basal ganglia, and motor cortex. It also participates in the maintenance of attention, and the anterior nucleus is a key component of Papez circuit [39]. Considering its multiple inbound and outbound connections, and in the light ours and previously reported findings, we believe the thalamus might act as the main common hub in CJD (at least in the subtypes represented in our study) and FI from which PrPSc would spread trans-synaptically to other regions, preferentially to structures that belong to neural networks where the thalamus has an important role, such as striatum, cerebellum, sensory-motor cortex, and medial temporal structures, which were atrophied or showed DTI alterations in patients from our study. Whether these diseases start at the thalamus, in either genetic or sporadic cases, or this is just a common pathway for PrPSc spreading cannot be elucidated from our data.
The main limitation of our study is the sample size, which prevented us to adjust the analyses by disease stage or CJD molecular subtype. Due to the limited sample size, we decided to analyze genetic and sporadic cases in the same group, as our main focus was to perform a neuroradiological characterization of patients based on their common phenotypic features, although we acknowledge that in larger sample sizes, it would be desirable to analyze these groups separately. We also acknowledge a possible pre-inclusion bias that may have led to a misrepresentation of the MM1 CJD subtype in our sample. Patients with MM1 CJD subtype present a faster disease progression, that could limited their chances of being referred to our study from other centers. However, our results encourage us to try to validate them in a larger cohort, ideally, with a multicentric study approach.
In summary, volumetric and DTI data point to the thalamus as a common central structure among regions vulnerable to CJD and FI, and therefore it may be considered as a candidate region for monitoring disease progression. DTI metrics show changes in opposite directions in CJD, with decreased MD in the striatum, increased MD in most WM tracts, and both increased and decreased MD values across different thalamic nuclei, which reflect the regional heterogeneity and dynamic nature of CJD neuropathological changes. In addition, the present study reports for the first time quantitative MRI changes in FI, characterized by atrophy and MD increase in thalamus and cerebellum. We believe that our findings, together with prior quantitative MRI studies in prion diseases reported in the literature, contribute to set the basis for using these tools as disease biomarkers.
Footnotes
ACKNOWLEDGMENTS
This work has been partially supported by a research grant to RSV (FIS1200013, Plan Nacional de I+D+I, co-funded by Instituto de Salud Carlos III and the European Region Development fund, FEDER). We thank all the volunteers for their participation in this study, and Dr. Juan Fortea and Dr. Berta Pascual-Sedano from Hospital de la Santa Creu i Sant Pau, Barcelona, Spain, for their collaboration in patient recruitment. We are indebted to the Neurological Tissue Bank of the Biobanc-Hospital Clinic-IDIBAPS, Barcelona, Spain, for sample and data procurement.
