OUP user menu

Genotype-specific patterns of atrophy progression are more sensitive than clinical decline in SCA1, SCA3 and SCA6

Kathrin Reetz, Ana S. Costa, Shahram Mirzazade, Anna Lehmann, Agnes Juzek, Maria Rakowicz, Romana Boguslawska, Ludger Schöls, Christoph Linnemann, Caterina Mariotti, Marina Grisoli, Alexandra Dürr, Bart P. van de Warrenburg, Dagmar Timmann, Massimo Pandolfo, Peter Bauer, Heike Jacobi, Till-Karsten Hauser, Thomas Klockgether, Jörg B. Schulz
DOI: http://dx.doi.org/10.1093/brain/aws369 905-917 First published online: 18 February 2013


Spinocerebellar ataxias are dominantly inherited disorders that are associated with progressive brain degeneration, mainly affecting the cerebellum and brainstem. As part of the multicentre European integrated project on spinocerebellar ataxias study, 37 patients with spinocerebellar ataxia-1, 19 with spinocerebellar ataxia-3 and seven with spinocerebellar ataxia-6 were clinically examined and underwent magnetic resonance imaging at baseline and after a 2-year follow-up. All patients were compared with age-matched and gender-matched healthy control subjects. Magnetic resonance imaging analysis included three-dimensional volumetry and observer-independent longitudinal voxel-based morphometry. Volumetry revealed loss of brainstem, cerebellar and basal ganglia volume in all genotypes. Most sensitive to change was the pontine volume in spinocerebellar ataxia-1, striatal volume in spinocerebellar ataxia-3 and caudate volume in spinocerebellar ataxia-6. Sensitivity to change, as measured by standard response mean, of the respective MRI measures was greater than that of the most sensitive clinical measure, the Scale for the Assessment and Rating of Ataxia. Longitudinal voxel-based morphometry revealed greatest grey matter loss in the cerebellum and brainstem in spinocerebellar ataxia-1, in the putamen and pallidum in spinocerebellar ataxia-3 and in the cerebellum, thalamus, putamen and pallidum in spinocerebellar ataxia-6. There was a mild correlation between CAG repeat length and volume loss of the bilateral cerebellum and the pons in spinocerebellar ataxia-1. Quantitative volumetry and voxel-based morphometry imaging demonstrated genotype-specific patterns of atrophy progression in spinocerebellar ataxias-1, 3 and 6, and they showed a high sensitivity to detect change that was superior to clinical scales. These structural magnetic resonance imaging findings have the potential to serve as surrogate markers, which might help to delineate quantifiable endpoints and non-invasive methods for rapid and reliable data acquisition, encouraging their use in clinical trials.

  • atrophy
  • neurodegeneration
  • spinocerebellar ataxia
  • volumetry
  • voxel-based morphometry


Spinocerebellar ataxias (SCAs) are a heterogeneous group of autosomal dominantly inherited and progressive neurodegenerative disorders. Their prevalence in Europe is estimated at ∼3:100 000, and SCA1, SCA2, SCA3 (Machado–Joseph disease) and SCA6 represent the most frequent genotypes among the >30 that are known (Kawaguchi et al., 1994; Riess et al., 1997; Schols et al., 2004; Paulson, 2009; Durr, 2010).

SCA1, SCA3 and SCA6 are caused by translated CAG repeat expansion mutations (Chung et al., 1993; Zhuchenko et al., 1997; Orr and Zoghbi, 2007) and are clinically characterized by symptoms resulting from damage to the cerebellum and its interconnections. Ataxia, the defining symptom, is often accompanied by extracerebellar symptoms. In particular, SCA1 and SCA3 show additional non-ataxia symptoms (Orozco Diaz et al., 1990; Dubourg et al., 1995; Durr et al., 1996; Filla et al., 1996; Schols et al., 1997; Schmitz-Hubsch et al., 2008a). The European Integrated Project on Spinocerebellar Ataxias (EUROSCA) natural history study, a multicentre longitudinal cohort study, revealed that progression, as assessed by the Scale for the Assessment and Rating of Ataxia (SARA) and the Inventory of Non-Ataxia Symptoms (INAS), was fastest in SCA1, followed by SCA3 and was lowest in SCA6 (Jacobi et al., 2011).

Evaluation of the baseline MRI data of the EUROSCA cohort has already allowed a better characterization of the structural brain changes associated with SCA1, SCA3 and SCA6. A complementary approach of quantitative 3D volumetry and voxel-based morphometry revealed severe brainstem, cerebellar and basal ganglia atrophy in SCA1 and SCA3, and a predominant cerebellar atrophy pattern in SCA6. Atrophy of the cerebellar hemispheres was less severe in SCA3 than in SCA1 and SCA6 (Schulz et al., 2010).

The purpose of the present study was to use structural imaging methods to capture the dynamics of neurodegeneration in relation to the clinical course in SCA1, SCA3 and SCA6. In this longitudinal MRI study of SCA1, SCA3 and SCA6, combining quantitative 3D volumetry and observer-independent voxel-based morphometry, we aimed to (i) distinguish whether there are specific SCA genotype-related atrophy progression patterns; (ii) identify clinicostructural correlations with disease duration, severity of symptoms and molecular markers; and (iii) compare sensitivity to detect longitudinal changes of MRI versus clinical parameters.

Materials and methods

Subjects and procedures

Within the framework of EUROSCA (www.eurosca.org) of the Ataxia Study Group (www.ataxia-study-group.net), patients with SCA included in the current study were recruited from eight of the participating centres (Departments of Neurology in Bonn, Brussels, Essen, Milan, Nijmegen, Paris, Tübingen and Warsaw). Baseline data were collected between November 2005 and March 2007, and follow-up data between January 2007 and December 2008. Sixty-three (SCA1 n = 37, SCA3 n = 19, SCA6 n = 7) (Table 1) of the 82 patients who were studied at baseline (Schulz et al., 2010) were available for follow-up, whereas 19 (23%) dropped out. Reasons for drop-out included progression to advanced disease stages in which MRI scanning was no longer tolerated, withdrawal of consent, MRI incompatibility and technical reasons.

View this table:
Table 1

Demographic and clinical characteristics of the study population at baseline and follow-up

    Age (years)43.32 ± 11.1445.91 ± 11.1746.79 ± 10.1549.05 ± 10.9566.86 ± 3.7468.75 ± 3.84
    Gender (F/M)16/2116/2110/910/91/61/6
    Visit interval (months)25 ± 1.6728.6 ± 6.4224.1 ± 1.1
    CAG repeat lengtha48.9 ± 6.0671.1 ± 2.4921.1 ± 0.38
    Disease stage1.23 ± 0.491.42 ± 0.55b1.50 ± 0.621.65 ± 0.741.57 ± 0.531.59 ± 0.64
    Disease duration (years)6.97 ± 3.199.11 ± 6.049.43 ± 55.56
    SARA11.63 ± 4.5617.35 ± 6.44b12.08 ± 5.6814.94 ± 6.63b13.57 ± 4.1214.43 ± 5.36
    INAS5.43 ± 2.126.31 ± 2.515.44 ± 2.075.33 ± 1.793.14 ± 1.953.83 ± 2.93
    SCAFI (Z-score)−0.18 ± 0.71−0.25 ± 0.790.46 ± 0.810.39 ± 1.08−0.18 ± 0.81−0.10 ± 0.90
    UHDRS-IV21.33 ± 4.5818.81 ± 5.09b20.61 ± 4.0519.56 ± 5.18b20.14 ± 6.4417.86 ± 5.87b
Control subjects
    Age (years)43.43 ± 11.1746.81 ± 11.9146.35 ± 10.5848.85 ± 10.6466.14 ± 3.9368.57 ± 3.55
    Gender (F/M)16/2116/2110/910/91/61/6
  • Values are mean ± SD.

  • a Expanded allele.

  • b Significant difference from baseline to follow-up (P < 0.05).

The CAG repeat length of the expanded and normal alleles of the affected genes was determined at the Department of Medical Genetics of the University of Tübingen by PCR-based fragment length analysis, as described previously (Jacobi et al., 2011).

We also included a subject-to-subject age-matched and gender-matched healthy control participant for each of the SCA genotypes, both at baseline and at follow-up (Table 1). All recruited healthy control subjects had no history of neurological or psychiatric disease. Informed consent was obtained from all participants. All local ethics committees of the participating recruiting centres approved the study.

Clinical parameters

Disease stage

Disease stages were defined as follows: Stage 0, no gait difficulties; Stage 1, disease onset, as defined by onset of gait difficulties; Stage 2, loss of independent gait, as defined by permanent use of a walking aid or reliance on a supporting arm; and Stage 3, confinement to wheelchair, as defined by permanent use of a wheelchair (Klockgether et al., 1998a).

Scale for the assessment and rating of ataxia

The severity of cerebellar ataxia was determined with the previously validated SARA (Schmitz-Hubsch et al., 2006, 2010; Weyer et al., 2007). SARA is a semi-quantitative clinical scale, which includes eight items assessing gait, stance, sitting, speech, finger-chase test, nose–finger test, fast alternating movements and heel–shin test. The SARA sum score ranges from 0 to 40, with 0 indicating absence of ataxia and 40 the most severe degree of ataxia.

Inventory of non-ataxia symptoms

INAS consists of 30 items combined in 16 extracerebellar symptoms or syndromes: areflexia, hyper-reflexia, extensor plantar response, spasticity, paresis, amyotrophy, fasciculations, myoclonus, rigidity, chorea, dystonia, resting tremor, sensory symptoms, brainstem oculomotor signs (horizontal and vertical ophthalmoparesis, slowing of saccades), urinary dysfunction and cognitive impairment. For a semi-quantitative assessment of non-ataxia symptoms, the number of non-ataxia symptoms is counted, yielding a dimensionless value with a range from 0 to 16. These 16 binary variables can be summed up to a simple sum score, the INAS count. Reliability of INAS ratings was tested in two large multicentre trials that served to validate SARA (Schmitz-Hubsch et al., 2008a, 2010).

Spinocerebellar ataxia functional index

The validated spinocerebellar ataxia functional index (SCAFI), consisting of quantitative performance measures or timed tests, is composed of a timed 8 m walk at maximum speed, the nine-hole peg test and the rate of PATA repetition, a measure of speech performance (Schmitz-Hubsch et al., 2008b).

Unified Huntington’s Disease Rating Scale

The 25-point checklist of daily tasks, as part of the functional assessment of the Unified Huntington’s Disease Rating Scale (UHDRS), was applied to assess the functional level in everyday life in patients with movement disorder (Huntington Study Group, 1996).

Magnetic resonance imaging

MRI data (1.5 T) were acquired with standardized acquisition protocols suitable for multicentre studies, as previously reported (Schulz et al., 2010). The same MRI model-dependent settings for the acquisition of magnetic resonance images, as reported in detail earlier (Schulz et al., 2010), were applied. A rigorous quality control to ensure stability of scan acquisition over time was performed. First, this includes not only non-replacement of MRI scanners during the execution of a longitudinal study and no relevant hardware or software changes but also quality control standards, such as the examination of image quality on a regular basis and routine tests of hardware performance. Moreover, to avoid a possible bias resulting from the use of different scanners and protocols, each site was required to contribute magnetic resonance data sets from normal control subjects, so that patients and control subjects were balanced between centres. For each scanner model, a custom-tailored imaging protocol was used to make data sets as similar as possible without losing image quality.

Quantitative 3D-volumetry

The semi-automated, quantitative 3D MRI-based volumetric technique used in this study was performed as described in the EUROSCA cross-sectional study (Schulz et al., 2010). This method has also been validated for its reproducibility in various previous studies (Luft et al., 1996, 1998; Schulz et al., 1999; Groschel et al., 2004; Hauser et al., 2006). In brief, volumetric processing, blinded to the diagnosis, included semi-automated removal of the skull, manual pre-segmentation set-up on predefined anatomical landmarks and semi-automated region-growing segmentation. The following regions of interest were used for volume analyses: (i) the brainstem subdivided into the medulla oblongata, the pons and mesencephalon; (ii) the cerebellum subdivided into the right and left cerebellar hemispheres, as well as the vermis; (iii) the caudate and putamen; and (iv) the cerebrum. Additionally, the total intracranial volume was estimated by using the optimized method (Eritaia et al., 2000).

First, brainstem and cerebellum were segmented and then subtracted from the original images. Anatomically, the cranial border of the brainstem was defined by a plane through the mammillary body and posterior commissure shifted caudally by one-third the distance to the cranial border of the pons. A parallel plane to the latter intersecting with the posterior rim of the foramen magnum represented the caudal border. Planes parallel to the mammillary body-posterior commissure line passing through the cranial and caudal notch of the pons were used to further divide the brainstem into midbrain, pons and medulla oblongata. The cerebellum was separated from the brainstem by a plane through the obex and posterior commissure shifted posteriorly to include the inferior colliculus. For the subdivision of the cerebellum, the cerebellar vermis was separated from the right and left hemispheres by sagittal planes. The volumes of caudate nucleus and putamen were calculated after manual pre-segmentation. The cerebrum was then divided into left and right hemispheres and fitted to the Talairach grid (Talairach and Tournoux, 1988) by manually defining anterior commissure, posterior commissure and the outermost extensions of the hemispheres. The cerebral lobes were segmented by manually tracing the central lateral and parietooccipital sulci on a sagittal view of the hemispheres. The lobes were further subdivided into grey and white matter using a rater-defined intensity threshold.

The volumes were calculated by multiplying the number of voxels per region of interest and the voxel size. The last step included normalization using the total intracranial volume.

Voxel-based morphometry

Magnetic resonance images of all subjects were analysed on a commercially available Unix machine using a voxel-wise statistical approach. Images were processed and analysed with Statistical Parametric Mapping (SPM) 8 (www.fil.ion.ucl.ac.uk/spm) and the voxel-based-morphometry toolbox (VBM8, http://dbm.neuro.uni-jena.de/vbm). For whole-brain analysis, we used the VBM8 option for processing longitudinal data. The following default preprocessing steps incorporated registration of follow-up measurement scan to baseline measurement scan, intra-subject bias correction, segmentation and spatial normalization as well as linear and non-linear registration. Applying a probabilistic framework, images were registered using linear and non-linear transformations, segmented and bias corrected within the same generative model (Ashburner and Friston, 2005). The following analyses were performed on grey matter segments that were multiplied by the Jakobian of the respective warp-fields (modulated grey matter volumes). Modulated grey matter images were smoothed with a Gaussian kernel of 6 mm full-width at half-maximum, which was also used in the previous cross-sectional analysis (Schulz et al., 2010). Additionally, we applied small volume correction using a cerebellar, medulla, midbrain or pons mask obtained from the Wake Forest University PickAtlas (Maldjian et al., 2003).

Voxel-wise grey matter differences were examined using a flexible factorial design assessing time (baseline and follow-up) × groups (SCA1, SCA3, SCA6 and their respective control groups), as well as including age, gender and study site as a covariate. To avoid possible edge effects around the border between grey and white matter or CSF, an absolute grey matter threshold of 0.01 (absolute threshold masking) was used. For the statistical analysis, we used an uncorrected height threshold of P < 0.001 across the whole brain and family wise error corrected on the cluster level P < 0.05 and minimum cluster size (k) of 20 voxels. Coordinates are reported in the standard anatomical space developed at the Montreal Neurological Institute (MNI). Statistical parametric mapping maps are superimposed for the cerebellum and brainstem on the SUITE template (Diedrichsen, 2006, 2009) and for the cerebrum on a single-subject MRI baseline template. Extracted grey matter values from regions of interest were used for post hoc analyses to contrast regional and genotype-related grey matter changes over time.

Statistical analysis

Group differences regarding male–female ratio were calculated with the Fisher’s exact test. Age differences between the groups were calculated with a one-way ANOVA. Univariate ANOVAs (with Bonferroni test for post hoc analyses), Kruskal–Wallis tests (Mann–Whitney U-test for post hoc analyses) or Wilcoxon signed ranks tests were used to determine group differences in clinical variables, depending on comparison and whether the assumptions for parametric approaches were met.

Total SCAFI Z-scores [ = (individual’s average of both PATA, nine-hole peg test and 8 m walk trials − mean of study population) standard deviation of study population] were calculated for our study population as previously reported (Schmitz-Hubsch et al., 2008b). Longitudinal changes on the SARA, UHDRS, INAS and SCAFI were determined using repeated-measures ANOVA with time (baseline and follow-up) as between-subjects factor, group (SCA1, SCA3 and SCA6) as within-subjects factor and age as covariate. The same factorial ANOVA model was used to establish longitudinal changes of normalized regional brain volumes within the genotype groups. Bonferroni tests were used to calculate pairwise comparisons for each group pair. Independent sample t-tests were used to establish cross-sectional comparisons in regional brain volumes between each of the genotype groups and their matched control group.

Per cent change of volume loss for each participant was calculated based on the equation reported by Risacher et al. (2010), using the baseline and follow-up normalized values for each region of interest, corrected for visit interval. Embedded Image

We used the same equation to calculate mean per cent change of the following clinical variables: SARA, UHDRS, INAS and SCAFI. We explored group differences in the per cent change of regional brain volumes and clinical variables using univariate ANOVAs with age as covariate (with Bonferroni test for post hoc analyses).

The standardized response mean as mean score change/standard deviation (SD) of score change is reported as effect size index to enable a comparison between scales. Values of 0.20, 0.50 and 0.80 were considered to represent small, moderate and large changes (Schmitz-Hubsch et al., 2010).

The magnitude of the association between regional brain volumes and clinical rate of change on the different measures used, disease duration and length of the expanded CAG repeat was determined through Pearson’s r for each genotype. The statistical analyses were performed using SPSS version 20 (IBM, 2011) with an alpha set at 0.05 as the statistical threshold for significance.


Group characteristics

Demographic and clinical characteristics at baseline and follow-up are reported in Table 1. Age at baseline [F(5,196) = 17.26, P < 0.0001] and follow-up [F(5,196) = 10.69, P < 0.0001] did not differ significantly between patients with SCA1 or SCA3 and respective control groups, but the patients with SCA6 and their matched control group were significantly older. Male-to-female ratio was not significantly different between groups. Disease duration [F(2,56) = 1.05, not significant] did not differ between genotypes. Visit interval between baseline and follow-up [H(2) = 7.75, P < 0.05] was slightly longer for patients with SCA3 than those with SCA1 (U = 198, Z = −2.28, P < 0.05) and SCA6 (U = 23.5, Z = −2.33, P < 0.05). Patients with SCA1 and SCA6 did not differ regarding disease duration. Disease stage at baseline [H(2) = 3.2, not significant] and follow-up [H(2) = 0.46, not significant] was similar between genotypes, but SCA1 was the only group to show a significant decline over time in this measure (Z = −2.65, P < 0.01).

As shown in Fig. 1, ataxia severity, assessed with the SARA, significantly worsened from baseline to follow-up in patients with SCA1 and SCA3, but not in those with SCA6 [F(2,58) = 8.29, P = 0.001 for interaction Time × Group]. For INAS, we found that, regardless of time point, patients with SCA6 showed a lower total INAS score when compared with patients with SCA1 or SCA3, but these two groups did not differ between one another [F(2,55) = 3.41, P < 0.05 for main effect group]. We did not find significant changes across time, or group differences on the SCAFI total score [F(2,49) = 0.136, not significant, for interaction Time × Group]. Analyses with the UHDRS showed that all three patient groups showed a similar decline in the daily life functional capacity across time [F(1,58) = 27.68, P < 0.0001 for main effect time].

Figure 1

Longitudinal changes of clinical and functional scores: SARA, INAS, SCAFI and UHDRS-IV in spinocerebellar ataxia (SCA) 1, 3 and 6. SCA1 and SCA3 showed deterioration in ataxia symptoms (SARA), but rate of progression was significantly high in SCA1. Patients with SCA6 showed a significant impairment in non-ataxia symptoms at both time points. There were no group, nor time effects in ataxia-related functional measures (SCAFI), but all three genotypes showed a similar decline in overall functional daily life (UHDRS). Asterisk indicates significant time effect for genotype.

Regarding rate of progression, calculated through per cent change of the SARA [F(2,57) = 6.26, P < 0.005], patients with SCA1 showed a greater decline than those with SCA3 or SCA6. The rate of change for patients with SCA3 on the SARA did not differ from those with SCA6. Rates of progression in the INAS [F(2,57) = 0.98, not significant], the SCAFI [F(2,51) = 1.045, not significant] and the UHDRS [F(2,57) = 0.08, not significant] were not significantly different between genotypes.

Longitudinal volume loss in genotypes

Using repeated measures ANOVA, we found a significant interaction between group and time for the following regions of interest: brainstem, pons, right putamen and left caudate nucleus. As presented in Fig. 2, planned comparisons showed a significant decrease of volume in the brainstem [F(2,56) = 3.5.37, P < 0.05, for interaction Time × Group], in the pons [F(2,56) = 6.834, P < 0.005, for interaction Time × Group], in the right putamen [F(2,55) = 3.938, P < 0.05, for interaction Time × Group] and the left caudate nucleus [F(2,55) = 3.624, P < 0.05, for interaction Time × Group] in the SCA1 and SCA3 groups, but not in the SCA6 group. Patients with SCA1 also showed a trend for progression of volume loss in the left cerebellar hemisphere [F(2,55) = 3.121, P = 0.052, for interaction Time × Group]. A significant decrease of volume in cerebrum [F(2,55) = 5.078, P < 0.05, for interaction Time × Group] was only found in the SCA6 group. All genotypes showed a similar decline in the left putamen and caudate nucleus [F(1,55) = 7.927, P < 0.05 and F(1,55) = 10.215, P < 0.005, for main effect time]. The remaining regions showed no significant interactions or main effects.

Figure 2

Longitudinal changes of brain volume in spinocerebellar ataxia (SCA) 1, 3 and 6. Volume of regions of interest in percent (%), which are normalized to the total intracranial volume (TICV), in each of the SCA genotypes (solid lines) and respective cross-sectional data of matched control subjects (round dotted lines) at baseline and follow-up, showed an increase of atrophy in the brainstem, cerebellum and basal ganglia in all genotypes over time. Asterisk indicates significant time effect for genotype.

Rate of volume loss for different genotypes

As presented in Table 2, all genotypes showed a marked reduction of volume in the basal ganglia (bilateral putamen and bilateral caudate) and in the mesencephalon. Nonetheless, specific patterns in the degree of change for each genotype were identified. In comparison with other genotypes, SCA1 showed an increased rate of volume loss in the brainstem [F(3,57) = 5.34, P < 0.05], left cerebellar hemisphere [F(3,57) = 3.33, P < 0.05] and putamen [F(3,57) = 2.83, P < 0.05]. The degree of change in the pons [F(3,57) = 7, 16, P < 0.0001] was similar in patients with SCA1 and SCA3, but larger than in those with SCA6. Patients with SCA3 also showed an increased per cent change in the brainstem in contrast to those with SCA6, but both groups did not significantly differ regarding rate of change in the left cerebellar hemisphere or putamen. Compared with other genotypes, only patients with SCA6 showed a significantly changed value in the cerebrum [F(3,57) = 1.45, P < 0.05]. Analyses of other regions did not reveal group differences of per cent changes from baseline to follow-up (see Supplementary material for mean and P-values).

View this table:
Table 2

Rate of volume loss of total intracranial volume-normalized volumetric results over time in SCA1, SCA3 and SCA6

Total intracranial volume (mm3)0.08 ± 0.260.03 ± 0.060.03 ± 0.08
Brainstem (%)−0.36 ± 0.22**,***0.20 ± 0.19*−0.05 ± 0.24*
Mesencephalon (%)−0.65 ± 0.70***−0.41 ± 0.60***−0.71 ± 0.86*,**
    Pons (%)−0.34 ± 0.23−0.20 ± 0.240.13 ± 0.33
    Medulla (%)0.12 ± 0.470.17 ± 0.470.30 ± 0.72
Cerebellum (%)−0.27 ± 0.38−0.10 ± 0.22−0.04 ± 0.18
    Cerebellar hemisphere right (%)−0.22 ± 0.44−0.17 ± 0.29−0.09 ± 0.44
    Cerebellar hemisphere left (%)−0.32 ± 0.44**0.01 ± 0.37*0.09 ± 0.44
    Vermis (%)−0.12 ± 1.10−0.10 ± 0.65−0.08 ± 0.86
Putamen (%)−0.84 ± 0.67**−0.48 ± 0.31*−0.40 ± 0.14
    Putamen right (%)−0.86 ± 0.53**−0.38 ± 0.33*−0.46 ± 0.25
    Putamen left (%)−0.81 ± 0.53−0.53 ± 0.36−0.67 ± 0.60
Caudate (%)−0.79 ± 0.65−0.49 ± 0.30−0.69 ± 0.20
    Caudate right (%)−0.78 ± 0.65−0.48 ± 0.41−0.47 ± 0.26
    Caudate left (%)−0.79 ± 0.66−0.48 ± 0.38−0.83 ± 0.30
Cerebrum (%)0.01 ± 0.23***−0.07 ± 0.24***−0.30 ± 0.15*,**
  • Values are mean ± SD.

  • *P < 0.05 compared with SCA1.

  • **P < 0.05 compared with SCA3.

  • ***P < 0.05 compared with SCA6.

Voxel-based morphometry comparisons

Patients with SCA1 showed the greatest decline in grey matter over time when compared with their respective age-matched and gender-matched control groups in the brainstem (pons and mesencephalon), left anterior and posterior cerebellum (lobule XI, lobule VIII) and the right putamen and pallidum [Table 3 and Fig. 3, P < 0.001 (uncorrected), k = voxels per cluster].

Figure 3

Longitudinal voxel-based morphometry changes in SCA1, SCA3 and SCA6 compared with their respective control groups. Group differences in pattern of reduction in grey matter over ∼2 years in SCA1 (red), SCA3 (blue) and SCA6 (green). Time × group interactions demonstrate differences in atrophy progression. Results are adjusted for age, sex and study site. Compared with healthy control subjects longitudinal voxel-based morphometry revealed greatest increase in grey matter atrophy in the cerebellum and brainstem in SCA1 (A), in the putamen and pallidum in SCA3 (middle row in C), and in the cerebellum as well as in the thalamus, putamen and pallidum in SCA6 (B, last row in C). Interaction contrasts are displayed at a threshold P < 0.0001 (uncorrected) with a minimum cluster size (k) >20 voxels. SPM maps are superimposed for the cerebellum and brainstem on the SUITE template and for the cerebrum on a single-subject MRI baseline template. The colour bars represent T-values.

View this table:
Table 3

Longitudinal voxel-based morphometry changes in SCA1, SCA3 and SCA6 compared with their respective control subjects

RegionSideMNI co-ordinatesZ-scorek
    Cerebellum anterior lobe, lobule IXLeft−13−51−384.4653
    Cerebellum posterior lobe, lobule VIIILeft−13−63−373.8428
    Cerebellum posterior lobe, tonsilLeft−22−45−413.6731
    Pons/cerebellum anterior lobeRight20−42−355.63894
    Cerebellum anterior lobe, dentateLeft−13−61−354.46151
    Thalamus, ventral lateral nucleusLeft−16−12134.8737

Whereas the main effect in grey matter changes in the SCA3 group was restricted to the bilateral putamen and pallidum (Table 3 and Fig. 3), in the SCA6 group, the most grey matter reduction was located in the cerebellum extending to the pons, in the left thalamus and bilateral pallidum/putamen (Table 3 and Fig. 3).

Clinical correlations

Association measures between localized volume reduction and the change on the various clinical variables revealed no significant findings. The length of the extended CAG repeat allele showed a moderate negative correlation with change on the left (r = −0.37, P < 0.05) and right cerebellum (r = −0.38, P < 0.05) in the SCA1 group. The same pattern was found when using the voxel-based morphometry data for cerebellum (r = −0.48, P < 0.005) and pons at the cluster level (r = −0.47, P < 0.005). Remaining correlation coefficients for each genotype were not significant in both data sets. Disease duration showed no significant association with changes on the clinical variables nor with changes on brain volume loss.

Standardized response mean of clinical and magnetic resonance imaging parameters

Calculation of standardized response means for all genotypes together showed that SARA (standardized response mean: 0.9) was the only clinical scale that reached a large effect size of change. The standardized response means for INAS, SCAFI and UHDRS (0.3, 0.1 and −0.2, respectively) indicated only weak to moderate effect sizes. When the SCA1 (standardized response mean: 1.2) and SCA3 (standardized response mean: 1.4) groups were analysed separately, SARA also had the largest effect size among all clinical scales (Table 4). In patients with SCA6, the UHDRS was the most sensitive scale (standardized response mean: −1.6; Table 4).

View this table:
Table 4

Standardized response means of clinical change for SCA1, SCA3 and SCA6

Clinical scoreSCA1SCA3SCA6All genotypes

Compared with clinical scales, almost all MRI parameters had larger effect sizes. When all genotypes were considered, the effect sizes of brainstem (standardized response mean: −1.2), putaminal (standardized response mean: −1.2) and caudate volumes (standardized response mean: −1.3) were largest followed by cerebellar volume (standardized response mean: −0.6) (Table 5). At the level of single genotypes, brainstem volume (pons > mesencephalon > medulla) showed the strongest effect of change in the SCA1 group (standardized response mean: −1.6) followed by putaminal and caudate volume (both standardized response means: −1.3 and −1.2). In patients with SCA3, the largest effect sizes were found for the putamen (standardized response mean: −1.5) and caudate volume (standardized response mean: −1.6) followed by brainstem volume (standardized response mean: −1.1) (Table 5). In patients with SCA6, the largest effect size index was found for the caudate (standardized response mean: −3.3), followed by the putamen (standardized response mean: −2.7), and the brainstem (standardized response mean: −0.2).

View this table:
Table 5

Standardized response means of volume loss rate for SCA1, SCA3 and SCA6

RegionSCA1SCA3SCA6All genotypes


This study provides a quantitative account of the progression of brain atrophy of three common spinocerebellar ataxias: SCA1, SCA3 and SCA6. It is based on an analysis of the MRI data obtained at baseline and after a follow-up period of 2 years from the ongoing EUROSCA natural history study, a multicentre longitudinal cohort study of 526 patients (Jacobi et al., 2011). Using a complementary volumetric and voxel-based morphometry approach, we found genotype-specific patterns of brain volume changes. In addition, we determined the rate of volume loss and compared morphometric measures and clinical scales with respect to their sensitivity to change. Specifically, volumetry and voxel-based morphometry provided converging evidence of genotype-specific progression of brain atrophy that particularly affected the brainstem and cerebellum in SCA1, the putamen and pallidum in SCA3 and cerebellum as well as the thalamus, putamen, caudate and pallidum in SCA6. Effect sizes of the most sensitive MRI parameters in each genotype were larger than those of the clinical measurements. Finally, in SCA1, CAG repeat length correlated with atrophy progression of the cerebellum and pons.

Previous structural neuroimaging studies in SCA1 mainly revealed atrophy in the brainstem, cerebellum and basal ganglia (Klockgether et al., 1998a; Guerrini et al., 2004; Della Nave et al., 2008; Ginestroni et al., 2008; Schulz et al., 2010; Goel et al., 2011). Compared with other SCAs, the typical olivopontocerebellar atrophy has been described as similar to, but not as severe as in SCA2 (Burk et al., 1996; Klockgether et al., 1998b; Guerrini et al., 2004), and more prominent than in SCA3 regarding the cerebellar hemispheres (Schulz et al., 2010). Extending these neuroimaging findings and current neuropathological knowledge of SCA1, a recent pathoanatomical study showed that brain damage in patients in the advanced clinical stages of SCA1 may go beyond the well-known macroscopic brain predilection sites (for review see Seidel et al., 2012), involving the motor cerebellothalamocortical and basal gangliathalamocortical circuits, the visual, auditory, somatosensory, oculomotor, vestibular, ingestion-related, precerebellar, basal forebrain cholinergic and midbrain dopaminergic systems (Rub et al., 2012). Our study confirms well-acknowledged atrophy patterns in the brainstem, the pons, the medulla, bilateral cerebellum, the vermis and the left putamen at baseline, and provides for the first time, data on the longitudinal atrophy patterns, which showed a significant decline in the mesencephalon, bilateral putamen and bilateral caudate nuclei within patients with SCA1. Contrary to SCA3 and SCA6, brainstem volume showed the greatest sensitivity to change in SCA1. Specifically, the largest effect size was found for the pons, followed by the mesencephalon, and only small responsiveness for the medulla. Likewise, large effect sizes were found in the putamen and caudate and moderate ones for the cerebellum. These results were further underpinned by the voxel-based morphometry analysis including comparison with their respective control groups, which revealed the greatest decline in grey matter in the brainstem (pons and mesencephalon), the right putamen and pallidum and left anterior and posterior cerebellum (lobule XI, lobule VIII). The important role of the pons in SCA1 was additionally supported by the mild negative correlation between the length of the extended CAG repeat allele and structural changes in pons (and also the cerebellum). Notably, the rare literature on correlations between molecular, clinical variables and structural findings in SCA is highly heterogeneous (Netravathi et al., 2009; Schulz et al., 2010; Goel et al., 2011). Further association measures between the rate of progression of atrophy and the change on the diverse clinical variables revealed no significant findings. For our longitudinal study, it might be argued that the impact of clinical decline in this slowly progressive disorder was not sufficient in the small 2-year interval to unravel significant associations between functional measures and atrophy.

Overall, comparing sensitivity with change of clinical and MRI parameters in SCA1, our data demonstrate that volume loss in the brainstem and in particular of the pons was more sensitive than the most sensitive clinical scale in this genotype, SARA. Thus, MRI of the pons was the most sensitive parameter to reveal changes over time, mirroring the fact that the pons seems to be one of the key regions in the neurodegenerative process in our SCA1 group.

Patterns of atrophy in the brainstem and cerebellum in SCA3 have been reported in previous neuroimaging studies (Burk et al., 1996; Klockgether et al., 1998b; Murata et al., 1998b; Onodera et al., 1998; Lukas et al., 2006; D’Abreu et al., 2012), which is well in consonance with autopsy findings that showed additional pallor of the substantia nigra. Microscopic analyses reveal widespread neuronal loss in the cerebral cortex, basal ganglia, thalamus, midbrain, pons, medulla oblongata and cerebellum (for review see Seidel et al., 2012). In previous studies, pontocerebellar atrophy was shown to be similar to SCA1 (Burk et al., 1996; Klockgether et al., 1998b). An early quantitative MRI study in patients with SCA3 indicated an effect of age and the expanded CAG repeat size on the neurodegenerative progress based on an observed closely correlation between the atrophy of brainstem and cerebellar vermis with these two factors (Onodera et al., 1998). A recent longitudinal neuroimaging study, using voxel-based morphometry in SCA3, failed to demonstrate progression of atrophy within 1 year (D’Abreu et al., 2012). However, the authors reported cortical involvement affecting the frontal, parietal, temporal and occipital lobes (D’Abreu et al., 2012) similar to another study that also reported a diminished transverse diameter of the pallidum (Murata et al., 1998b). The involvement of the basal ganglia in SCA3 has been further described in previous volumetric measurements in SCA3, but not in SCA1 and SCA2 (Klockgether et al., 1998b). Regarding our neuroimaging findings in SCA3, the volumetric results are congruent with the known involved regions. Specifically, the strongest effect sizes of volume loss change were found in the caudate and putamen, followed by the brainstem and only moderate effects regarding the cerebellum. Similarly, in the voxel-based morphometry interaction analysis, grey matter changes in the bilateral putamen and pallidum were the only significant change in the SCA3 group when compared with control subjects over time. Why are the basal ganglia so prominently presented in our longitudinal analysis in SCA3? One reason is the finding of an already severe atrophy of the cerebellum and brainstem at baseline (Schulz et al., 2010), leading to a ceiling effect in the longitudinal approach. Apart from that, our data reflect the strong basal ganglia involvement that is typical for SCA3. One of the first SCA3 families reported was initially characterized as autosomal dominant striatonigral degeneration (Rosenberg et al., 1976). Various basal ganglia symptoms, in particular, dystonia and parkinsonism have been found in SCA3 families (van Gaalen et al., 2011). Dystonia was found to be associated with decreased thalamic grey matter (D’Abreu et al., 2011). Nevertheless, in our SCA3 group, only three patients showed an upper limb rigidity at baseline and follow-up. One reason for the lack of correlation between clinical and imaging findings may be that basal ganglia symptoms in SCA3 may be masked by clinically severe cerebellar ataxia. In conclusion, in our SCA3 group, the most sensitive parameter to detect change over time was the MRI of the striatum, which was greater than the most sensitive clinical scale, SARA.

As SCA6 is considered the prototype of a pure cerebellar ataxia, our findings of clear extracerebellar abnormalities challenge this view of a purely cerebellar disorder. Although the voxel-based morphometry also revealed atrophy over time in the cerebellum and pons, the main and surprising finding was the impact on the basal ganglia with both methods of analysis (volumetry and voxel-based morphometry) and thalamus (voxel-based morphometry only). At this, the greatest effect size was found for the caudate. However, although we found large effect sizes of change in the cerebrum in SCA6, compared with control subjects, these effects did not reach significance. Our earlier cross-sectional study revealed atrophy in SCA6 mostly in the cerebellum, but also in total brainstem and pons (Schulz et al., 2010). With the current data, we were now also able to identify subcortical changes in the thalamus and parts of the basal ganglia. In the literature, brain pathology studies give evidence for microscopic degeneration in the cerebral cortex, thalamus, midbrain, pons, medulla oblongata and cerebellum (Gomez et al., 1997; Gierga et al., 2009). Neuroimaging studies mainly report moderate to severe atrophy of the vermis, mild atrophy of the cerebellar hemispheres and no atrophy of the middle cerebellar peduncles, pons or other structures of the posterior fossa (Murata et al., 1998a; Satoh et al., 1998; Schols et al., 1998; Butteriss et al., 2005; Lukas et al., 2006). Some studies documented mild atrophy in the anteroposterior diameter of the pons and the diameter of the middle cerebellar peduncle (Murata et al., 1998a; Eichler et al., 2011), which, nonetheless, was never as severe as in SCA1 or SCA3. Mild diffuse atrophy of cortical areas have been reported in some patients with SCA6 (Schols et al., 1998). The view of a more multiregional brain disorder is supported by metabolic studies. Cerebral glucose positron emission tomography (FDG-PET) demonstrated hypometabolism not only in the cerebellum and brainstem but also widespread in cortical regions and basal ganglia (Soong et al., 2001). In addition, some SCA6 cases with parkinsonism show evidence of nigrostriatal dopaminergic dysfunction (Khan et al., 2005; Kim et al., 2010).

There might be also disease-related underlying different stages of neurodegenerative processes, requiring years to reach its full extent. For SCA6, our baseline data imply severe cerebellar (and brainstem) atrophy in early phases (Schulz et al., 2010), and the longitudinal approach demonstrates the primary or secondary impact on parts of the basal ganglia and the thalamus in later phases.

Clinically, the most sensitive scale was the assessment of the functional level in everyday life (UHDRS-IV) in our SCA6 sample, which is not symptom-specific and may also be because of the older age of the SCA6 cohort and/or more severe functional difficulties. As these findings are based on a small sample and particularly an older cohort of patients with SCA6, further studies are needed to confirm these newer aspects. Nevertheless, when taken together, the most sensitive MRI parameter in SCA6 was found for the caudate, which was better than the most sensitive clinical scale, the UHDRS-IV, and indicates extra-cerebellar involvement in SCA6.

In SCAs, understanding the genotype-specific pathoanatomical profile is important for the establishment of specific outcome measures and their use in clinical trials. In this study, we demonstrated the feasibility of detectible changes of genotype-related atrophy patterns. Our results indicate that for SCA1, the most sensitive and key primary outcome measure over a course of 2 years is MRI of the pons and for SCA3 the striatum. Because of limited explanatory power given by the small sample size in SCA6, we are cautious with statements about outcome measures in this sample. However, there is evidence for extracerebellar involvement in SCA6. Another limitation factor might be the relatively short duration of 2-year intervals in an otherwise slowly progressive neurodegenerative disorder. Nevertheless, it may be argued that (i) despite the short visit interval, we already found genotype-specific changes in SCAs; (ii) it is well acknowledged that neurodegeneration patterns overlap considerably in advanced disease stages; as well as (iii) this time-interval seems to be also appropriate in other longitudinal SCA studies (Reetz et al., 2010).

Overall, this prospective, longitudinal MRI study provides a sensitive indicator of genotype-related brain atrophy over time in SCA1 and SCA3. Similar to the TRACK-HD study in pre-manifest and early Huntington’s disease (Tabrizi et al., 2012), we revealed specific MRI and clinical outcome measures on the basis of longitudinal effect size. We encourage the usage of MRI for longitudinal studies and point towards other promising MRI techniques, such as metabolic (e.g. spectroscopy), diffusion tensor and ultra-highfield magnetic resonance imaging, allowing early detection of change. Nevertheless, future therapeutic trials of disease-modifying agents may incorporate these parameters to gauge the likelihood of future treatment success.


EUROSCA represents the first large multisite study to report changes across magnetic resonance brain imaging in clinical and molecularly characterized SCA1, SCA3 and SCA6 patients over a 2-year time interval. The present study used a complementary approach of structural neuroimaging methods to confirm that MRI-based structural markers are able to detect dynamic genotype-related changes in atrophy, which are superior to the most sensitive clinical scales, encouraging their use in future trials.


EUROSCA/LSHM-CT-2004 –503304 from the European Union (to J.B.S.), GeneMove/01 GM 0602 from the German Ministry of Education and Research (to J.B.S.); Excellence Initiative of the German federal and state governments (DFG ZUK32/1, to K.R.); PhD fellowship (SFRH/BD/65743/2009) from the Fundação para a Ciência e Tecnologia (Portugal), financed by the POPH – QREN Program (European Social Fund) (to A.S.C.).

Supplementary material

Supplementary material is available at Brain online.


The authors would like to thank all participants for their enduring collaboration and interest in this research.

European Integrated Project on Spinocerebellar Ataxias
Inventory of Non-Ataxia Symptoms
Scale for the Assessment and Rating of Ataxia
spinocerebellar ataxia
spinocerebellar ataxia functional index
Unified Huntington’s Disease Rating Scale


View Abstract