Background Cerebellar damage has been implicated in information processing speed (IPS) impairment associated with multiple sclerosis (MS) that might result from functional disconnection in the frontocerebellar loop. Structural alterations in individual posterior lobules, in which cognitive functioning seems preponderant, are still unknown. Our aim was to investigate the impact of grey matter (GM) volume alterations in lobules VI to VIIIb on IPS in persons with clinically isolated syndrome (PwCIS), MS (PwMS) and healthy subjects (HS).
Methods 69 patients (37 PwCIS, 32 PwMS) and 36 HS underwent 3 T MRI including 3-dimensional T1-weighted MRIs. Cerebellum lobules were segmented using SUIT V.3.0 to estimate their normalised GM volume. Neuropsychological testing was performed to assess IPS and main cognitive functions.
Results Normalised GM volumes were significantly different between PwMS and HS for the right (p<0.001) and left lobule VI (p<0.01), left crus I, right VIIb and entire cerebellum (p<0.05 for each comparison) and between PwMS and PwCIS for all lobules in subregions VI and left crus I (p<0.05). IPS, attention and working memory were impaired in PwMS compared with PwCIS. In the whole population of patients (PwMS and PwCIS), GM loss in vermis VI (R2=0.36; p<0.05 when considering age and T2 lesion volume as covariates) were associated with IPS impairment.
Conclusions GM volume decrease in posterior lobules (especially vermis VI) was associated with reduced IPS. Our results suggest a significant impact of posterior lobules pathology in corticocerebellar loop disruption resulting in automation and cognitive optimisation lack in MS.
Statistics from Altmetric.com
The mechanisms involved in cognitive impairment (CI) of persons with multiple sclerosis (PwMS) are not fully understood.1 CI affects several cognitive domains, including episodic memory, attention, working memory and executive functions.1 Slowing of information processing speed (IPS) is the main cognitive dysfunction observed in multiple sclerosis (MS) even at early stages.2 MRI studies suggest that it may be related to distributed demyelinating and neurodegenerative alterations of brain networks.3 ,4 The cerebellum could be part of these networks and the involvement of cerebellar dysfunction in IPS impairment has been suggested.5 A role for the cerebellum in cognition in healthy subjects (HS) has been suggested by anatomical, clinical and imaging studies.6–10 The corticocerebellar loop connects posterior cerebellar hemispheres and dentate nuclei to prefrontal, superior temporal and lateral parietal cortices, via the thalamus and the pons.6 A precise cerebellar functional map has emerged from different functional MRI (fMRI) studies leading to the idea of dedicated cognitive lobules in the most posterior parts of the cerebellum.11 Cerebellar damage is common in MS and extensive demyelination has been described in the cerebellar cortex on postmortem samples.12 Some studies showed an association between CI, including IPS, and cerebellar symptoms in PwMS.5 ,13 An association between cerebellar atrophy, especially grey matter (GM) global atrophy, and IPS and working memory impairment has been shown in MS.14–18 Several fMRI studies in HS and in PwMS have shown that greater cerebellar activation correlates with faster cognitive performances in HS but not in PwMS and that the functional link between the cerebellum and frontal areas was impaired in patients with MS.19–21 The decrease in cerebellar activation in PwMS has been interpreted as a failure of the cerebellum in facilitating rapid cognitive performances and may play a role in diminished IPS in MS.
Since the posterior cerebellar lobules that have been specifically implicated in cognition are anatomically connected to the frontal areas implicated in cognitive networks, we hypothesised that posterior cerebellar GM damage could be responsible for corticocerebellar disconnection and thus lack of cognitive optimisation and automation resulting in IPS impairment. In paediatric MS, the global volume of posterior lobules of the cerebellum has been correlated with cognition including IPS.18 Our study aimed to investigate the impact of GM volume alterations in individual lobules VI to VIIIb on cognitive outcome, especially IPS, in persons with clinically isolated syndrome (PwCIS), PwMS and HS.
Sixty-nine patients (PwCIS or PwMS) and 69 HS, matched for age, sex and educational level, were recruited between June 2010 and December 2014 at the Bordeaux University Hospital Center, France. All patients, as well as 36 out of 69 HS, underwent an MRI scan and all study participants were evaluated with cognitive testing.
All PwCIS (n=37) were included within 6 months after their first neurological episode and presented with at least two asymptomatic cerebral lesions on fast fluid-attenuated inversion recovery (FLAIR) images. For PwMS (n=32), the inclusion criteria were as follows: MS diagnosis according to McDonald's criteria,22 disease duration >6 months and ≤15 years, and mild CI defined as two scores beyond one SD among a large neuropsychological battery. Patients with MS were treated according to current standards of clinical care.
Exclusion criteria were: age under 18 or over 55 years, history of other neurological or psychiatric disorder, inability to perform computerised tasks or MRI, MS attack in the 2 months preceding the screening, corticosteroid pulse therapy within 2 months preceding screening, severe cognitive deficits (Mini-Mental State Examination <27), depression (Beck Depression Inventory score (BDI) >27).
Clinical assessment, the French version of Expanded Disability Status Scale (EDSS) and the Multiple Sclerosis Severity Score (MSSS)23 were determined by expert neurologists.
Standard protocol, approvals, registration and patient consents
Each participant gave written informed consent. Patients were included from two different studies (REACTIV, ClinicalTrials.gov Identifier: NCT01207856, study concerning cognitively impaired PwMS, and SCI-COG, ClinicalTrials.gov Identifier: NCT01865357, analysing CI in PwCIS). Both studies were approved by the local ethics committee.
Each cognitive domain was evaluated with the following tests (clinically isolated syndrome (CIS) and paired HS: *; MS and paired HS: **). Since PwCIS and PwMS were included in two different studies, a few tests, only for working memory and verbal fluency, differed between the two samples, and z scores for cognitive domains were calculated by comparisons with scores obtained in the control group. IPS tests were the same for all participants. Tests used for neuropsychological assessment are described in a previous work.24
Attention: Test of Attentional Performance (TAP)*/** consisting of subtests for accurate answers of visual scanning, and visual and auditory divided attention. For divided attention, the number of accurate answers and reaction time ratios of the double task (auditory and visual divided attention) to the simple task (auditory or visual divided attention) was considered,
Working memory: numerical span test (forward*/** and backward*/**) and Paced-Auditory Serial Addition Test–3 s (PASAT)* or working memory subtest of the TAP,**
Executive functions: Stroop test*/** (using the difference between the denomination part and the inhibition task scores) and Word List Generation test (verbal fluency assessment)* or semantic verbal fluency (using animal category),**
IPS: Symbol Digit Modalities Test (SDMT).*/**
Depression and anxiety levels were assessed using the BDI and State-Trait Anxiety Inventory for adults, subtest State (STAI-S), respectively.
MRIs were performed on a 3 T Achieva TX system (Philips Healthcare, Best, The Netherlands) with an eight-channel phased array head coil. A morphological protocol consisted of: three-dimensional (3D) T1-weighted MRIs acquired using magnetisation prepared rapid gradient echo (MPRAGE) imaging (TR=8.2 ms, TE=3.5 ms, TI=982 ms, α=7°, FOV=256 mm, voxel size=1 mm3, 180 slices) and 2D multislice FLAIR images (TR=11 000 ms, TE=140 ms, TI=2800 ms, FOV=230 mm, 45 axial slices, 3 mm thick).
Postprocessing and image analysis
Total brain, GM and white matter (WM) volumes, normalised for skull size in the Montreal Neurological Institute (MNI) space, were estimated from 3D-T1 MPRAGE images using SIENAX25 (part of FSL V.5.0, with brain extraction tool (BET options: ‘-B -f 0.3’)), taking into account FLAIR lesion masks for patients. A custom image processing pipeline was applied to automatically segment the WM lesions. A corrected Freesurfer parcellation was used on the T1-weighted MRI to obtain spatial priors for WM, GM and cerebrospinal fluid.26 A two-channel expectation-maximisation algorithm produced the final segmentation of WM lesions as outliers of the WM intensity distributions.27 The output of the pipeline is a WM lesion probability map from which an expert user derived the final binary map by a lesion probability thresholding step using 3D Slicer V.4.0.
For cerebellar volume estimations, cerebellar GM and WM probability maps were obtained as well, using BET and FAST (FSL V.5.0). SUIT toolbox V.3.0 (SPM 8)28 was used for cerebellar lobule segmentation. The software allows standardising lobule size to drive a reliable segmentation independent of interindividual variability, and set back the images in each patient’s native space in order to calculate lobular volumes. Steps are described below (figure 1):
The cerebellum was identified on the brain on 3D-T1 MPRAGE images,
Cropped T1-weighted images underwent a non-linear registration to the SUIT template,
The same non-linear registration was performed on WM/GM probability maps described above, using 7° B-spline interpolation in order for lobules of all participants to be in a normalised coordinate space,
A reverse non-linear registration was performed for probability maps, a cerebellar lobule atlas (CerebellumSUIT) and a whole cerebellum atlas (Buckner) into each patient’s native space in order to obtain total cerebellar and lobular volumes, as well as the volumes for their WM and GM compartments for each participant.
Using FSL V.5.0, GM and WM probability maps were thresholded in order to include the whole cerebellar volume without any overlap between GM and WM and to create binarised label maps. Atlases previously registered in patients' native space were also binarised in order to differentially label each lobule. The total, GM and WM volumes were then calculated for the whole cerebellum, as well as for each lobule using MATLAB V.2011b. All volumes were normalised for skull size, using the SIENAX scaling factor.
Lobules VI, crus I, crus II, VIIb, VIIIa and b were studied because of their known involvement in cognition. Vermis crus I was excluded from analyses as some participants were lacking reliable segmentation maps due to this structure's small size.
All data were analysed with R package ‘stats’ (V.3.1.3). Normal distribution was tested for all variables with the Shapiro-Wilk test.
Sex and educational level were compared using the χ2 test. Quantitative clinical and imaging data were compared between PwCIS, PwMS and HS with the analysis of variance (ANOVA) or Kruskal-Wallis test depending on their distributions. For post hoc analyses, Tukey or Nemenyi tests were used to compare two subgroups when ANOVA or Kruskal-Wallis tests, respectively, showed significant results.
Z scores were calculated according to the formula:using our population of 69 HS. Cognitive domain impairment was defined by a z score below −1.5 in the given domain. Z score comparisons between PwMS and PwCIS were obtained with paired t-test (or Mann-Whitney). A significance threshold of 0.05 was applied.
According to the variables' distribution, Spearman or Pearson correlations between imaging and cognitive outcome in all patients (CIS and MS), and PwCIS and PwMS independently, were used and then adjusted for multiple comparisons using a Bonferroni-adjusted significance threshold of p<0.002.
Linear regression analysis was used to predict cognitive outcome, including three hierarchical blocks: (1) clinical data (age and MSSS), (2) cerebellar volumes (lobule GM volumes and whole infratentorial T2 lesion volume), (3) cerebral volumes (cerebral WM, cerebral GM and cerebral T2 lesion volume). Dependent variable and residual normal distributions were checked using the Shapiro test and histogram analyses. Independent variables were entered in the model only if the related p value was below 0.10 in univariate analyses.
Demographics, clinical data and cognitive assessment
We included 37 PwCIS, 32 PwMS and 69 HS. There were no significant differences in sex, median age and educational level between groups, either when considering the whole HS group or only the HS subgroup that underwent MRI. Table 1 describes population demographics and clinical characteristics.
No PwCIS and 6.25% of PwMS were impaired on attention compared with HS. Working memory impairment occurred in 5.41% of PwCIS and 25% of PwMS, executive functions impairment in 8.11% of PwCIS and 15.63% of PwMS, and IPS slowness occurred in 13.51% of PwCIS and 71.77% of PwMS. Attention, working memory and IPS mean z scores were significantly decreased in PwMS versus PwCIS. No differences were detected for executive functions.
No correlation was found between cognitive assessment and anxiety, depression scales or MSSS.
Table 2 shows that normalised cerebral and cerebellar volumes were significantly different between groups. The PwMS group developed significant GM atrophy in the whole cerebellum, as well as the cerebrum. Interestingly, within the cerebellum, atrophy was observed in several substructures, that is, lobules VI, left crus I and the right VIIb. The PwCIS group had preserved GM volumes compared with HS.
Correlations between cognitive outcomes and imaging data
Results of univariate correlations between imaging data and cognitive outcomes and multiple linear regressions are presented in table 3. In the whole group of patients (MS and CIS, n=69), normalised GM matter volumes in the whole cerebellum, vermis VI and bilateral crus II were significantly associated with IPS z scores, and a trend was also observed for all posterior lobules. No significant correlation was found between other cognitive domains and GM volume in posterior lobules.
According to multivariate analysis, the right VIIb volume (R2=0.09; F=7.98; p<0.01) was associated with working memory, while age, vermis VI GM volume and cerebral T2 lesion volume (R2=0.36; F=13.30; p<0.001) were associated with IPS z scores. Infratentorial T2 lesion volume was correlated with executive functions. No model was significant to predict attentional outcome.
We found evidence that cerebellar atrophy in specific posterior cerebellar lobules had an impact on cognitive outcome in MS, specifically on IPS, even though global cerebellar atrophy did not show a significant independent correlation with cognition. Indeed, GM loss in vermis VI was associated with IPS impairment. This result supports our hypothesis that a lack of optimisation and automation role of the posterior substructures of the cerebellum, and especially vermis VI, may induce IPS slowness. The correlation observed between the right VIIb GM volume decrease and working memory impairment was probably not clinically relevant due to the very low value of the R2 coefficient, but could reflect an effect of IPS impairment on working memory which has been previously discussed by others.29
IPS is strongly related to other cognitive domains and particularly to executive functioning. Many studies have related the impact of cerebellar posterior lesions on executive functioning such as strokes, tumours, postinfectious cerebellitis, neurogenetic cerebellar syndromes or cerebellar congenital malformations.10 ,30 In their first description of the cerebellar cognitive affective syndrome (CCAS), Schmahmann and colleagues have found that among 20 patients suffering from cerebellar alterations, 18 had executive impairment. In this study, patients free from such processing troubles had anterior and no posterior cerebellar damages, leading to motor rather than cognitive disability.30
However, CCAS not only consists in executive dysfunctions but also includes visual spatial, and linguistic impairments, and affective dysregulation, whereas remote episode and semantic memory are relatively preserved.30 ,31 Studies in patients with MS corroborate this description. Weier et al15 demonstrated that cerebellar T1 lesion volume correlated with working memory (PASAT). Patients with cerebellar lesions also displayed attentional, verbal fluency and spatial memory deficits compared with lesion-free patients and HS.32 Cerebellar WM lesions contribute to cognitive disability but did not seem to be the leading cause of CI in MS. In their papers, van de Pavert et al17 and Morgen et al14 correlated working memory (PASAT), executive functions and attention performances17 to GM atrophy in the whole cerebellum. In our study, the whole cerebellar volume correlated with cognitive scores in univariate analyses but not in the multivariate models, suggesting that the association was mainly driven by the volume of the posterior lobules. We observed preponderant GM atrophy in PwMS but not in PwCIS, in agreement with previous studies,33 ,34 whereas another study detected atrophy at this stage only when disease duration was not taken into account.35 We observed a non-significant trend of an increase of GM cerebellar volume in PwCIS, in agreement with a longitudinal study showing a transient paradoxical increase in cerebellar GM volume in PwCIS, possibly due to inflammation preceding atrophy.36 In relapsing-remitting MS (RRMS), a majority of studies tend to show cerebellar GM atrophy.15 ,16 ,33 ,34 In patients with MS with cerebellar symptoms, a specific pattern of frontotemporal cortical atrophy, independent of cerebellar lesions, has been described highlighting corticocerebellar disconnection in MS.37
As aforementioned, this disconnection seems to occur specifically between associative cerebral cortices and posterior cerebellar lobules.8 In HS, although all posterior lobules are engaged, some preferential contribution of specific lobules in cognitive domains have been observed, such as the left superoposterior cerebellum for attention,7 vermis VI and crus I for verbal working memory and lobule VI, crus I/II and VIIb for executive functions.9 ,8 For patients with MS, no such anatomical associations have been investigated previously. Our results highlighted the relationship between specific posterior cerebellar GM alterations and IPS rather than a specific cognitive domain. It is, however, possible that at later stages of the disease cerebellar pathology contributes more significantly to CI in other domains. We found that GM loss in vermis VI was more specifically associated with IPS impairment. Vermis VI involvement has been documented in implicit memory and sequence learning reaction time, supporting its role in IPS.38 Moreover, lobule VIIb volume, associated with working memory in our results, has been previously linked to executive functions and language, notably concerning the right side, which is connected to the left cerebral associative cortices.8 ,9 Finally, crus II was associated with IPS in univariate analysis, but not in multivariate models. Crus II is less involved in cognitive processing in the literature, consistent with our findings.8
Associative cerebellar processes are involved in selecting, optimising and automating cognitive processes rather than integrating cognitive tasks, which is the prefrontal cortex’s main role. In HS, fMRI studies suggested that speeder cognitive performances are associated with greater activation in the cerebellum and disclosed functional connectivity between the cerebellum and the dorsolateral prefrontal areas during cognitive tasks.39 The cerebellum may assume the most automatised parts of attentional requests which are needed to perform cognitive tasks in order to spare cortical regions. Cerebellar processes leading to cognitive automation and optimisation are mostly unknown. However, analogous to motor skills, theories have emerged considering a transition from ‘controlled’ to ‘automatic’ cognitive functioning throughout learning processing. Hence, owing to cerebellar modulation, tasks that require high cognitive load and complex attention become stereotyped, independent from distraction and from high-level cerebral associative oversight. Posterior cerebellar lobules seem to integrate internal representations with external stimuli in order to generate the proper response, progressing efficiently according to the context. From a physiological point of view, the cerebellar cortex consists of a homogeneous histological organisation independent from sensorimotor or cognitive functioning. Cerebellar inputs are then responsible for functional specialisation. Concerning cognitive adaptation, theoretical models suggest that this phenomenon implies synaptic plasticity between Purkinje cells and prefrontal and parietal projections. Once afferent information from prefrontal and parietal cortices is selected and compared with internal representation, synaptic reinforcement permits strengthening of the most adapted comportment. Associative cerebral cortices are then disengaged from the automatic and harmonious generated response, resulting in faster reaction time and possibility to perform new high competence behaviour.40 According to Schmahmann and colleagues, dysfunctions of this homoeostatic control may lead to ‘dysmetria of thoughts’. The mechanisms by which cerebellar damage could be associated with IPS impairment have been suggested by several fMRI studies.19–21 In early RRMS, when patients showed normal performances in the easiest levels of the tasks and significantly longer reaction times (slow IPS) during the most demanding cognitive conditions, shorter reaction times were not associated with higher cerebellar activation but with an increase in medial prefrontal activation instead and the functional connectivity analysis showed a functional link between the dorsolateral prefrontal cortex and medial prefrontal regions but not with the cerebellum.19 These results suggest that patients with MS are unable to activate the typical cerebellofrontal network associated with the fastest responses in the task and that they activate a substitute compensatory network involving the medial prefrontal cortex.
Our study is not without limitations. Volumetric analysis is challenging because of interindividual variability. Moreover, thresholding at 0.5 was used to discriminate between GM and WM with a possible overestimation of cerebellar WM. However, we checked every segmentation output and, when necessary, manually corrected mismatches in GM and WM in the 3D-T1 sequences and these limitations did not have an impact on our results.
This study reinforces and extends the suggestion made by previous works of an important role for cerebellar damage in CI at different stages of MS. For the first time, alterations in specific posterior lobules (especially vermis VI) were associated with IPS impairment, suggesting an important role for alterations of these structures in this impairment due to a lack of cognitive automation and optimisation through anatomical disconnection in the corticocerebellar loop.
Contributors AM and BB were involved in drafting the manuscript. All authors revised the manuscript for important intellectual content. AM, AR, DL-H, MD and BB were involved in study concept and design. AM, AR, DL-H, FM, MD, PC, NM, DSM, TT, CRGG and BB were involved in analysis and interpretation of the data. AR, MD, DL-H and BB were involved in acquisition of the data. AM and MD were involved in statistical analysis. BB, AR and TT were involved in study supervision and coordination.
Funding AM received a research grant from the Fondation pour la Recherche Médicale (DEA20140630564). This study was supported by ANR-10-LABX-57 Translational Research and Advanced Imaging Laboratory (TRAIL), laboratory of excellence. The SCICOG study was also supported by a grant from Teva, and the REACTIV study by a grant from Merck-Serono. CRGG was supported in part by a grant from the National Multiple Sclerosis Society (grant identifier RG-1501-03141) and by a visiting professorship grant from the Excellence Initiative (IdEx) of the University of Bordeaux.
Competing interests BB, AR and J-CO or their institution received research grants and/or consulting fees from Biogen-Idec, Bayer-Healthcare, Novartis, Genzyme, Roche, Medday, Merck-Serono and Teva. CRGG received a research grant from Sanofi. VP received travel expenses from ARSEP Fondation, Biogen, Teva-Lundbeck and Merk-Serono.
Ethics approval CCP Bordeaux Aquitaine.
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.