Background The thalamus is a major neural hub, with selective connections to virtually all cortical regions of the brain. The multisystem neurodegenerative syndrome amyotrophic lateral sclerosis (ALS) has pathogenic overlap with frontotemporal dementia, and objective in vivo markers of extra-motor pathological spread are lacking. To better consider the role of the thalamus in neurodegeneration, the present study assessed the integrity of the thalamus and its connectivity to major cortical regions of the brain in a longitudinal manner.
Methods Diffusion-based MRI tractography was used to parcellate the thalamus into distinct regions based on structural thalamo-cortical connectivity in 20 patients with ALS, half of whom were scanned at two time points, and 31 matched controls scanned on a single occasion.
Results At baseline, widespread diffusivity alterations in motor- and extramotor-associated thalamic parcellations were detectable. Longitudinal decline selectively affected thalamic regions associated with frontal and temporal lobe connectivity. Diffusivity measures were significantly correlated with clinical measures of disease burden. Progression of functional disability, as indicated by change on the ALS functional rating scale, was associated with longitudinal change in mean diffusivity of the right frontal lobe thalamic parcellation (r=0.59, p=0.05).
Conclusions Regional thalamic connectivity changes mirror the progressive frontotemporal cortical involvement associated with the motor functional decline in ALS. Longitudinal MRI thalamic parcellation has potential as a non-invasive surrogate marker of cortical dysfunction in ALS.
- amyotrophic lateral sclerosis
- motor neuron disease
- diffusion tensor imaging
Statistics from Altmetric.com
The thalamus is a major neural hub that receives and projects information to all cortical regions of the mammalian brain, where it has a critical role in sensory, motor and cognitive processes.1 Although traditionally viewed as a relay structure, increasing evidence suggests that the thalamus plays an active role in modulating cortico-cortical information transfer with respect to behavioural context.2 The cytoarchitecture of the thalamus itself is complex and unique, comprising numerous subnuclei with distinct projections to a diverse array of cortical and subcortical neural structures that can result in selective deficits of varying severity across a wide domain of brain functions when disrupted. These thalamic subnuclei can be parcellated in vivo, into distinct nuclei with divergent patterns of structural and functional connectivity using diffusion and functional MRI, respectively.3 4
Across the neurodegenerative disease spectrum, amyotrophic lateral sclerosis (ALS) is characterised by the loss of upper motor neurons combined with loss of lower motor neurons arising from the brain stem and spinal cord anterior horns. ALS has a clinical, pathological and genetic overlap with frontotemporal dementia (FTD).5 Further clinical heterogeneity includes site of initial symptom onset as well as rate of progression, with survival duration ranging from a few months to decades. Up to half of patients with ALS may have demonstrable mild neuropsychological deficits.6 Despite this clinical heterogeneity, nearly all cases of ALS are associated with a common underlying histopathology, namely deposition of a 43 kDa phosphorylated transactive response DNA binding protein (pTDP-43).7 ALS is a syndrome that can be seen as a common end point of multiple upstream molecular pathways.8 Stereotyped patterns of postmortem cerebral histological change have been identified,9 lending support to wider evidence for a corticoefferent pathogenesis,10 in which there might be propagation of protein misfolding through functionally connected brain networks.11 While pTDP-43 aggregates appear to be concentrated in motor neurons, all cortical projection neurons that generate a long axon are posited to be vulnerable.12 Thalamic involvement in ALS has been reported across in vivo neuroimaging platforms,13–16 as well as postmortem histopathological studies,7 17 although its role in the neurodegenerative process has not been articulated. As such, the present study tested the hypothesis that regional alterations in thalamic connectivity may serve as a holistic marker of the widespread cortical pathology and disease involvement in ALS.
A group of patients with ALS (n=20) were selected from the longitudinal Oxford Study for Biomarkers in Motor Neurone Disease cohort, matched in age (p=0.26) and education (p=0.13) with a group of healthy controls (n=31). The patients were chosen from 65 potential MRI datasets according to a priori criteria: <70 years at time of scan (to improve age matching to controls and minimise independent age-related thalamic atrophy), apparently sporadic (no family history of ALS or FTD), classical ALS (not flail arm or other regionally limited phenotypes), no overt dementia and a disease progression rate (DPR) >0.4 ALSFRS-R points per month.
The diagnosis of ALS was established by two experienced neurologists (MRT, KT). Patients were all presumed to be sporadic, given no family history of ALS or FTD. Patient clinical examinations were carried out at baseline and, for a subset (n=11), again after an approximate 6month interval (mean=5.6 months, SD=0.8 months). Disease duration was calculated from date of symptom onset to date of scan. Severity of functional disability was assessed at each clinical examination using the revised ALS functional rating scale (ALSFRS-R), a four-domain disability questionnaire with scores ranging from 0 to 48, with a lower score reflecting greater disability.18 DPR, as indicated by increasing disability, was calculated by dividing the change in the ALSFRS-R sum score at baseline and follow-up by the time between assessments. General cognitive function was assessed using the Addenbrooke’s Cognitive Examination-Revised (ACE-R), covering attention/orientation, memory, verbal fluency, language and visuospatial ability. For ACE-R total score, as well as verbal fluency and memory recall subscores, a higher score indicates better cognitive function.
All participants were scanned at the Oxford Centre for Clinical Magnetic Resonance Research using a 3T Siemens Trio scanner with a 12-channel head coil on the day of the clinical examination. Whole-brain T1-weighted MRI scans were acquired using a magnetization-prepared rapid gradient echo sequence (TR/TE=2040/4.7 ms; flip angle=8°; 1 mm isotropic resolution). Whole-brain diffusion-weighted images were acquired using an echo-planar sequence (60 isotropic directions; b-value=1000 s/mm2; echo time/repetition time=94/10 000 ms; 2×2×2 mm3 voxel size; 65 slices). Four additional b0 images without diffusion weighting were acquired. A field map was acquired using a gradient echo imaging sequence (2×2×2 mm3 voxel size; 65 slices; echo time 1/echo time 2/repetition time=5.19/7.65/655 ms) to account for distortions present in diffusion-weighted data caused by field inhomogeneities.
Volumetric and vertex analysis
Participant thalami were segmented from their whole-brain T1-weighted MRI scan using FIRST19 in FSL. Resulting segmentations were visually inspected for accuracy. For each participant, grey matter volume of the thalamus was then calculated after normalising for interindividual differences in head size via the volumetric scaling factor derived from SIENAX.20 Briefly, brain and skull images were extracted and affine-registered to MNI152 space, using the skull image to determine the registration scaling, to obtain a volumetric scaling factor accounting for the difference between the subject’s image and standard space.
Focal morphological change along the surface of the thalami was then examined between patients and healthy controls using vertex analysis. Vertex analysis is a powerful statistical tool that provides complementary structural characteristics for specified structures of interest. Specifically, vertex locations from each participant were projected onto a deformable mesh surface of an average template as scalar values using FIRST in a mode of operation that assessed group differences on a per-vertex basis. Intergroup differences were assessed using permutation based non-parametric testing. Results with p<0.05 were considered significant after cluster-based multiple comparison correction.
Each individual’s structural MRI scan was also processed in FreeSurfer V.6.0 to generate cortical masks of interest to be used as target masks in subsequent tractography analyses. The standard cortical processing pipeline was carried out,21 whereby scans were preprocessed and linearly aligned to the MNI305 average brain and segmented into tissue types. Images were aligned to a common spherical template surface using cortical folding-based co-registration patterns. After alignment, the cortex was partitioned on the basis of gyral and sulcal structures using an automated segmentation procedure. For each individual, left and right hemispheric cortical masks comprising the frontal, somatosensory, parietal, temporal and occipital cortices, as defined by the Desikan-Killiany atlas,22 were obtained, as well as Brodmann defined areas of premotor (BA6) and primary motor (BA4) cortices. All cortical masks of interest were transformed from spherical-space back to native space using FSL’s affine registration tool FLIRT23 and dilated to ensure masks reached white matter boundaries. Grey matter volumes of all cortical masks were extracted from FreeSurfer for each participant’s scan.
Diffusion preprocessing and tractography
All diffusion-weighted scans were corrected for head motion and eddy currents, then brain-extracted to remove all non-brain voxels. Field map correction was performed with FUGUE to correct for field inhomogeneities and to unwarp scans. Following preprocessing, DTIFIT was used to apply a diffusion tensor model at each voxel, resulting in maps of three eigenvalues (L1, L2, L3), which allowed the calculation of fractional anisotropy (FA), mean diffusivity (MD), radial diffusivity (RD) and axial diffusivity (AD) maps for each participant’s scan.4
Using each individual’s preprocessed diffusion-weighted image, scans were processed in FSL’s BEDPOSTX to estimate the fibre orientation distribution function within each voxel for subsequent probabilistic tractography. Probabilistic tractography was performed independently within left and right hemispheres of the brain and in each subject’s native structural space using PROBTRACKX, using the thalamus as a seed mask and the cortical structures obtained with FreeSurfer, as described above, as classification targets. Following tractography, seed-to-target information defining the connectivity between seed and target masks was used to perform a winner-takes-all clustering to assign each voxel within the thalamus to the target with which it showed the highest probability of connection.4 Mean diffusion metrics (FA, MD, RD, AD) of each participant’s left and right thalamic parcellations were extracted and compared between patients and controls.
For longitudinal analysis of 11 patients with ALS, all 6-month follow-up structural and diffusion-weighted scans were processed using the same protocol, and independent of baseline scans. Tractography-based thalamic parcellation was performed in the same manner as at baseline, using cortical masks derived from follow-up structural patients scans processed in FreeSurfer. To ensure longitudinal diffusivity change was independent of volumetric differences of thalamic parcellations at baseline and follow-up, patient scans were transformed into a common ‘halfway’ space using SIENAX and diffusion metrics were extracted for each time point only where the parcellations intersect. Mean diffusion metrics of each patient’s left and right thalamic parcellations at 6-month follow-up were extracted and compared with baseline.
Statistical analyses comparing thalamic volume, cortical volume and diffusivity of thalamic parcellations between ALS and healthy controls were carried out using SPSS V.21.0. The Shapiro-Wilk test was used to test normality. Total thalamic volume was compared between participant groups, and longitudinally within patients with ALS using two-tailed independent samples t-test and paired samples t-test, respectively. Cortical mask volumes, cluster size and diffusion tensor imaging metrics averaged across thalamic parcellations relating to the frontal, premotor (BA6), primary motor (BA4), somatosensory, parietal, occipital and temporal cortices, were assessed using one-way multivariate analysis of variance (MANOVA) at baseline in patients and controls, followed by planned comparisons of each dependent variable between cohorts using Bonferroni correction. Longitudinal change within 11 patients with ALS was analysed using repeated measures MANOVA. Two-tailed Pearson’s correlation was used to examine correlations between imaging metrics and clinical measures. Results were adjusted for multiple comparisons using Bonferroni correction. In all analyses, p values<0.05 were considered to be significant.
Patient data and demographics
ALS and healthy control cohorts were well matched for age (p=0.26) and years of education (p=0.13) (table 1). Site of symptom onset in patients with ALS varied, but was equally represented in upper (n=8) and lower (n=9) limb cases, and three bulbar onset cases. Patients were cognitively intact as indicated by overall performance on the ACE-R (mean=93.6, SD=4.4), scoring within 1 SD of the normative population score for their age group (mean=95.82, SD=3.8), and above the high end cut-off limit of 88 for suspected dementia. The ALS cohort recruited for the current study is representative of classical sporadic ALS.
Volumetric and morphological thalamic change
Patients with ALS showed reduced overall grey matter volume for the left (p=0.02) and right (p=0.03) thalamus (figure 1A). This finding was further supported by complementary vertex analyses of the thalami, which indicated changes involving the medial surface of the left and right thalamus in patients with ALS (figure 1B). No significant volumetric or morphological thalamic changes were observed longitudinally.
Structural connectivity-based thalamic parcellation
Each participant’s left and right thalamus was reliably and consistently parcellated using tractography according to their ipsilateral structural connectivity to frontal, premotor (BA6), primary motor (BA4), somatosensory, parietal, temporal and occipital cortices (figure 2). Grey matter volume of the cortical masks used as classification targets in tractography was compared between patients with ALS and controls. A difference in volume of cortical masks was approaching significance based on cohort at baseline (p=0.06). Follow-up comparisons for each cortical mask indicated patients with ALS had significantly reduced bilateral grey matter volume of primary motor cortex (BA4), relative to healthy controls (all p<0.01; online supplementary table 1). Longitudinally, there was a significant change in volume within subjects across cortical masks in patients with ALS (p=0.03), with follow-up comparisons indicating significant bilateral grey matter volume reduction in the primary motor cortex (BA4) between baseline and 6-month follow-up (all p<0.02). Cluster size of thalamic parcellations was also assessed at baseline and follow-up. No significant difference in cluster size was observed across thalamic parcellations between patients with ALS and controls at baseline (p=0.61), or longitudinally within patients with ALS (p=0.29; Online supplementary table 2).
Diffusion metrics indicated increased diffusion in thalamic parcellations associated with the left frontal lobe, bilateral premotor cortex, bilateral motor cortex, right somatosensory cortex and bilateral parietal lobe, in patients with ALS compared with healthy controls (all p<0.05; table 2). This pattern of increased diffusivity in patients with ALS was consistent across diffusion metrics of MD, RD and AD, and reflects greater diffusion occurring in these subthalamic regions as a result of reduced cellular integrity. No differences in FA were observed between patients with ALS and healthy controls across thalamic parcellations.
Longitudinal diffusivity change across thalamic parcellations in patients with ALS was more selective (table 3). The right frontal lobe thalamic parcellation showed a significant increase in MD, RD and AD (all p<0.02). The right and left temporal lobe thalamic parcellations also showed a significant increase in MD, RD and AD (all p<0.05).
Clinical correlates of thalamic integrity
Volumetric measures of total grey matter volume of the thalami did not show any significant correlation with clinical measures of disease duration, absolute ALSFRS-R score or rate of progression. Diffusion metrics, however, showed a positive correlation with disease duration for the right premotor (MD r=0.48, p=0.03; RD r=0.48, p=0.03; AD r=0.51, p=0.02), right primary motor (MD r=0.6, p<0.01; RD r=0.56, p<0.01; AD r=0.45, p=0.05) and right somatosensory (MD r=0.61, p<0.01; RD r=0.59, p<0.01; AD r=0.48, p=0.03) cortices, respectively, associating increased diffusion with longer disease duration (figure 3A; online supplementary figure 1A). Conversely, diffusion metrics showed a significant negative correlation with ALSFRS-R score for the right premotor (MD r=−0.52, p=0.02; RD r=−0.53, p=0.02; AD r=−0.59, p<0.01), right primary motor (MD r=−0.57, p<0.01; RD r=−0.59, p<0.01; AD r=−0.53, p=0.02) and right somatosensory (r=−0.6, p<0.01; r=−0.62, p<0.01; r=−0.61, p<0.01) cortices, respectively, associating reduced diffusion with lower functional disability (figure 3B; online supplementary figure 1B). DPR, as indicated by change in ALSFRS-R at 6-month follow-up, showed a negative correlation with two diffusion metrics in the left temporal lobe (RD r=−0.43, p=0.05; AD r=−0.45, p=0.05) (figure 3C; online supplementary figure 1C). Relative to baseline, a positive correlation was detected between DPR and longitudinal change in MD of the right frontal lobe thalamic parcellation (r=0.59, p=0.05), but of subthreshold significance with RD (r=0.57, p=0.06) and AD (r=0.58, p=0.06). MD, RD and AD are all indicators of degree of diffusion, with increases indicative of reduced cellular integrity. The correlation between diffusion metrics and DPR suggests integrity of the left temporal and right frontal thalamic subregions are clinically tied to progressive functional disability in patients with ALS.
This study demonstrated widespread changes in thalamic cortical structural connectivity in ALS, involving extra-motor frontotemporal parcellations longitudinally, highlighting thalamic change as a holistic marker of disease burden in ALS. Some studies of thalamic grey matter change in ALS have reported reduction only in ALS-FTD24 25; in the 10% of patients with ALS associated with a hexanucleotide repeat expansion (HRE) in C9orf72 24 26; or no detectable reduction.27 Diffusion MRI studies in ALS consistently report markers of reduced thalamic cellular integrity.28 29 Thalamic abnormalities extend beyond conventional grey matter and diffusion change to altered functional connectivity,28 increased iron deposition,30 altered metabolism,31 decreased texture features,32 altered cerebral blood flow,33 reduced motor task-related activation34 and increased microglial activation.35 Thus, despite the variability of detectable grey matter atrophy, thalamic abnormalities are a very robust finding consistent across a wider range of neuroimaging modalities (table 4).
Thalamic grey matter changes
Our cohort of patients with ALS showed a significant bilateral reduction in overall volume of the thalamus at baseline, but no significant longitudinal change. Complementary vertex analysis supported this finding and also demonstrated significant bilateral boundary reduction along the midline of the thalamus, a region typically associated with connectivity to the frontal lobe. Factors that have been associated with more generalised cerebral atrophy in ALS are the presence of cognitive impairment,36 comorbid FTD37 and the C9orf72 HRE.26 It is possible that one or two of our presumed sporadic ALS patient cohort were carriers of the C9orf72 HRE, though highly unlikely to be a significant confound if so. Likewise, while the suboptimal cognitive assessment used in this study predated tools with more sensitivity to ALS-related cognitive impairment and behavioural change such as the ECAS,38 none of this study’s participants showed features indicative of dementia.
Diffusion changes in thalamic parcellations
At baseline, widespread increased apparent diffusion was detected in parcellated thalamic regions associated with structural connectivity to frontal, premotor, primary motor, somatosensory and parietal cortices. Mathematically, the lack of difference in FA across thalamic parcellations observed here is likely the result of MD, RD and AD all showing a similar pattern of increased diffusivity in patients maintaining a similar ratio. This might reflect a lack of preferential diffusivity change along specific radiating directions due to the lack of isolated bundles of parallel fibres within the grey matter of the thalamus.
Longitudinal diffusivity change was considerably more selective, affecting only the right frontal lobe and bilateral temporal lobe thalamic areas. The observed thalamic abnormalities were also found to relate to disease duration, functional disability rating and progressive functional decline. Disease duration and ALSFRS-R score significantly correlated with degree of diffusivity in the premotor, primary motor and somatosensory parcellated regions of the right thalamus at baseline. This is consistent with previous studies that have linked disease duration and ALSFRS-R score with motor-related integrity in ALS.29 Our finding of negative correlation between DPR and diffusivity of the baseline left temporal lobe thalamic parcellation, and positive correlation between DPR and longitudinal change in diffusivity of the right frontal lobe thalamic parcellations requires consideration. One might infer that cellular integrity of left temporo-thalamic connectivity and the increased longitudinal decline in cellular integrity of the right fronto-thalamic connectivity at baseline are associated with rate of progression in ALS. Rather than integrity of motor-related thalamic connectivity, rate of disease progression might then be linked to imaging markers of frontotemporal involvement, even in ALS without overt cognitive impairment.
The interpretation of observed diffusivity change in the thalamus remains conservative, with limited assumptions on the cellular impact of diffusion alterations in grey matter. This remains important when dealing with iron-rich deep grey matter structures in an elderly participant population. Not only does level of ferritin-bound iron impact MR signal, including diffusivity metrics, but levels naturally elevate as a consequence of ageing.39 Of relevance, this confound has been reported to have little effect on the diffusivity of the thalamus, compared with the basal ganglia.39 The importance of a demographically well-matched patient and control cohort for deep grey matter investigation remains critical, and undoubtedly there remains a major issue of clinical heterogeneity inherent to single-centre studies of limited size, which multicentre initiatives are looking to address. This study only considered the thalamus, focused as it was on a potential surrogate of widespread cortical pathology. The structural and functional relationship of the broader basal ganglia (eg, caudate) to thalamic changes in ALS is an important concept for future study. Similarly, this study was not powered to explore potential hemispheric lateralisation of thalamic pathology, which remains more broadly unanswered in the ALS neuroimaging literature. Pooling of larger numbers of clinically homogeneous cases through multicentre ALS cohorts arising from the Neuroimaging Society in ALS40 offers real potential for addressing these more intricate clinicopathological correlates.
In summary, thalamic involvement is a consistent feature of the ALS cerebral pathological signature. This study has demonstrated the utility of diffusion-based thalamic parcellation as a non-invasive marker of disease progression in ALS.
The authors thank all the study participants for their efforts and enthusiasm for clinical research.
Contributors ST contributed to design, analysis and manuscript preparation. RAM contributed to design, data acquisition, analysis and manuscript preparation. KT contributed to data acquisition and manuscript preparation. MCK contributed to manuscript preparation. MRT contributed to design, data acquisition and manuscript preparation.
Funding ST is funded by the Australian National Health and Medical Research Council CJ Martin Early Career Fellowship (APP1121859). MCK was supported by the Australian National Health and Medical Research Council Program Grant (Forefront #1037746). MRT is funded by the Medical Research Council and Motor Neurone Disease Association Lady Edith Wolfson Senior Clinical Fellowship (MR/K01014X/1). The Oxford MND Centre (MRT, KT) receives funding from the Motor Neurone Disease Association.
Competing interests None declared.
Patient consent Not required.
Ethics approval Ethical approval for this study was obtained from South Central Oxford Ethics Committee B (08/H0605/85).
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.