Neocortical Atrophy in Machado-Joseph Disease: A Longitudinal Neuroimaging Study


  • Anelyssa D’Abreu MD, PhD,

    1. From the Neuroimaging Laboratory (AD, CLY, BAGC, FC); Department of Neurology (AD, MCF, FC); and Department of Medical Genetics (ILC)—University of Campinas-UNICAMP (ILC).
    Search for more papers by this author
  • Marcondes C. França Jr. MD, PhD,

    1. From the Neuroimaging Laboratory (AD, CLY, BAGC, FC); Department of Neurology (AD, MCF, FC); and Department of Medical Genetics (ILC)—University of Campinas-UNICAMP (ILC).
    Search for more papers by this author
  • Clarissa L. Yasuda MD, PhD,

    1. From the Neuroimaging Laboratory (AD, CLY, BAGC, FC); Department of Neurology (AD, MCF, FC); and Department of Medical Genetics (ILC)—University of Campinas-UNICAMP (ILC).
    Search for more papers by this author
  • Bruno A. G. Campos MD,

    1. From the Neuroimaging Laboratory (AD, CLY, BAGC, FC); Department of Neurology (AD, MCF, FC); and Department of Medical Genetics (ILC)—University of Campinas-UNICAMP (ILC).
    Search for more papers by this author
  • Iscia Lopes-Cendes MD, PhD,

    1. From the Neuroimaging Laboratory (AD, CLY, BAGC, FC); Department of Neurology (AD, MCF, FC); and Department of Medical Genetics (ILC)—University of Campinas-UNICAMP (ILC).
    Search for more papers by this author
  • Fernando Cendes MD, PhD

    1. From the Neuroimaging Laboratory (AD, CLY, BAGC, FC); Department of Neurology (AD, MCF, FC); and Department of Medical Genetics (ILC)—University of Campinas-UNICAMP (ILC).
    Search for more papers by this author

  • The authors report no conflict of interests regarding this research. This work was supported by FAPESP.

  • J Neuroimaging 2012;22:285-291.

Dr Fernando Cendes, Department of Neurology-UNICAMP, Cidade Universitária Zeferino Vaz, CEP 13083970 Campinas-SP-Brazil. E-mail:




Previous imaging studies in the Machado-Joseph disease (MJD/SCA3) have mostly concentrated on the cerebellum and brainstem. Our goal was to perform a whole brain longitudinal evaluation.


We included 45 patients and 51 controls, who underwent two brain magnetic resonance imaging and magnetic resonance spectroscopy (mean interval of 12.5 ± 1.5 months). We used voxel-based morphometry (VBM) and the MarsBar analysis toolbox to extract grey matter density (GMD) values from regions of interest. We used a linear regression model and a general linear model to correlate GMD with clinical markers, and paired t-test for the longitudinal evaluation.


We observed decreased GMD (P < .01) at frontal, parietal, temporal and occipital lobes, subcortical grey matter, cerebellum, and brainstem. White matter atrophy was restricted to the cerebellum. Age, CAG, and disease duration predicted GMD in different areas, but age and CAG were the most important predictors. The longitudinal analysis failed to demonstrate changes. Changes in regions other than the cerebellum appeared to contribute significantly to the final International Cooperative Ataxia Rating Scale score.


We confirmed cortical involvement in MJD/SCA3. The most important factors in predicting GMD were age and CAG. The lack of progression of atrophy may indicate floor effect and/or short duration of follow-up.


The Machado-Joseph disease (MJD/SCA3) is an autosomal-dominant spinocerebellar ataxia resulting from a mutation at the ATXN3 (MJD1) gene at the chromosome 14q, involving CAG repeat expansion.1 Clinical variability is a major characteristic in MJD/SCA3.2-5 MJD/SCA3 is characterized by variable degrees of ataxia, peripheral neuropathy, opthalmoparesis, pyramidal signs, dystonia, sleep disorders, or parkinsonism.6-8 Cognitive changes, psychiatric symptoms, and motor neuron degeneration have been reported,9-11 even though dementia is usually not regarded as part of the classical clinical spectrum.12 CAG repeat length inversely correlates with age of onset.13

Neuropathological and neuroimaging studies have mostly concentrated on the cerebellum, brainstem, spinal cord, and basal ganglia.4,5,7,14-16 Neuroimaging studies used visual analysis or manual quantification, most frequently quantifying cross-sectional area and only a few have used volumetric measurements.17-19 The cortex has rarely been analyzed.20 There is growing evidence, however, of structural and functional abnormalities in regions other than the cerebellum and its connections. Descriptive neuroimaging studies in small series showed atrophy of the frontal and temporal lobes.14,21 Positron emission tomography and single photon emission tomography also suggested involvement of cortical areas.22,23 A spectroscopy (MRS) study in the deep cerebral white matter demonstrated decreased N-acetyl-aspartate over creatine (NAA/Cr) ratios suggesting axonal dysfunction in a region not known to be anatomically affected.24

Our main purpose was to study the whole brain of a larger series of MJD/SCA3 patients. Voxel-based morphometry (VBM) and MarsBar allow for an unbiased whole brain study.25 Our secondary goal was to track changes over time.



All participants signed a written informed consent form, approved by our Institutional Review Board. Forty-five consecutive patients with a clinical and molecular diagnosis of MJD/SCA3 underwent a review of their clinical history, a neurological examination, and assessment by the International Cooperative Ataxia Scale (ICARS).26 Mean age was 47.02 years (21-73; SD = 12.23), age of onset was 37.04 years (13-55; SD = 11.05), and duration of disease was 9.97 years (2-29; SD = 6.13). Age of onset was defined as the first appearance of motor symptoms as reported by patients. Mean CAG repeat number was 72 (65-81). Thirty subjects were male (65%). Mean ICARS score was 36.36 (4-80; SD = 18.5).

The control group consisted of 51 healthy controls with no history of neurological or psychiatric illness. Mean age of this group was 44.08 (20-68; SD = 11.78), which was not different from the patient sample (P= .25).

Structural MRI and MRS Scanning Protocol

For a detailed account of the imaging methodology (acquisition and processing), please refer to the supplementary data. MRI and MRS were performed on a 2-Tesla scanner (Elscint Prestige, Haifa, Israel). MRI acquisition protocol for T1-weighted 3-dimensional gradient echo was with 1-mm isotropic voxels, acquired in the sagittal plane (1 mm thick; flip angle, 35°; TR, 22; TE, 9; matrix, 256 × 220; and FOV, 25 × 22 cm). We acquired single voxel1 H-MRS using point-resolved spectroscopy sequence (PRESS; TR = 1500 ms, TE = 135 ms, NEX = 200). We placed one single voxel (2 × 5 × 1 cm) over the superior-posterior region of the left hemisphere at the level of corpus callosum after the acquisition of scout anatomical images in sagittal and axial planes, as described previously.24 The purpose of the ROI in this location was to sample the largest possible volume of white matter with the least interference from CSF or gray matter. Prior to the acquisition, we performed a localized shimming at the ROI to ensure adequate field homogeneity followed by water suppression adjustment.

All images were submitted to a visual analysis by the investigators and an independent neuroradiologist. The main abnormalities observed in the visual MRI analysis of the MJD population are consistent with those described in the literature.21,24 We also verified all images after automatic segmentation to ensure that images were properly aligned and segmented without errors.

Data Pre-processing and VBM Analysis

Data were analyzed using SPM 2 ( Images were then transformed from Dicom to Analyze® format using MRIcro ( The VBM analysis involves several fully automated pre-processing steps. We investigated differences in GM and WM density between patients and controls. Statistical voxel wise analyses were performed in SPM2. Contrasts were defined in order to estimate the probability of each voxel being grey matter (GM) or white matter (WM). Secondly, we used the MarsBar software ( on SPM2 to extract the grey matter density (GMD) in selected ROIs which were defined according to the freeware library Anatomical Automatic Labeling (,28 For each ROI we at first performed a multivariate linear regression and then a linear regression with stepwise backward regression for the ROIs that did not reach significance. More details about pre-processing of MRI and ROI analyses are provided in supplementary data.

We also sought to correlate ICARS scores and disease severity as measured by the degree of imaging change. For this purpose we performed a general linear model with stepwise backward regression. The dependent variable was the ICARS score, while the independent variables were the ROI densities. As we could not perform this analysis with every single ROI at once, we subdivided the ROIs into lobes (frontal, parietal, temporal, occipital, limbic), subcortical GM (thalamus, caudate nucleus, putamen, and pallidum) and cerebellum+vermis. Significance was set at P < .05.

Longitudinal Analysis

After an interval of 12.5 ± 1.5 months, 30 patients had a second MRI. We applied the same techniques described above to process those images for the VBM analysis. We used a paired t-test with 5% FDR to compare the GMD and white matter density (WMD) between the first and second MRI.

We also performed longitudinal MRS studies. Data were converted to java language and analyzed by software Java Magnetic Resonance User Interface (jMRUI version 3.0). N-acetyl-aspartate (NAA), Creatine (Cre), and Choline (Cho) were quantified and NAA/Cre ratios calculated. Out of the 30 patients with repeat MRS, we were able to acquire good quality spectra in 18 pairs. Exclusion criteria were basically quality of the spectra. Differences in NAA/Cr between examinations were examined by paired t-test with significance at P < .05.


We observed significant decreases in GMD (P < .01) in the: cerebellum and vermis, brainstem (medulla, pons, and midbrain), lentiform nucleus, caudate nucleus, claustrum; frontal lobes (precentral; inferior, superior and middle frontal gyrus; paracentral lobule); parietal lobes (postcentral gyrus; precuneus; inferior, superior, angular, and supramarginal gyrus); temporal lobes (fusiform gyrus; insula; middle, superior temporal gyrus); occipital gyrus (cuneus, inferior occipital gyrus, lingual gyrus, middle occipital gyrus, superior occipital gyrus); limbic lobe (cingulate cortex; parahippocampal gyrus); and thalami (Fig 1 and Table 1).

Figure 1.

Areas of decreased GM density (hot colors) and WM density (cold colors) when comparing MJD/SCA3 and controls (P < .001).

Table 1.  Areas of Significant Atrophy by VBM Analysis. R = Right Cerebrum; L = Left Cerebrum; (R/L) = Right and Left Cerebrum. T =t-score; Z = z-score; FDR-COR = False Discovery Rate Corrected
ClusterTZVoxel P(FDR-COR)X,Y,Z {MM}Region
28117213.94Inf.000–3 –37 –17Cerebrum R/L (posterior cingulate, anterior cingulate, subcallosal gyrus, sub-gyral, fusiform Gyrus, inferior occipital gyrus, insula (R), parahippocampal gyrus, lingual gyrus, middle occipital gyrus, middle temporal gyrus (R), superior occipital gyrus (R), precentral gyrus, inferior frontal gyrus (R),
 11.49Inf.000–5 –46 –51Cuneus, angular gyrus, supramarginal gyrus, cingulate gyrus, inferior parietal lobule, precuneus, superior parietal lobule, paracentral lobule, postcentral gyrus); R/L cerebellum; vermis; medulla, pons, midbrain;
 11.40Inf.0005 –49 –52lentiform nucleus; claustrum (R); caudate nucleus; thalamus
115474.954.66.00045 40 –9(R) frontal lobe (sub-gyral; inferior/middle/superior/medial frontal gyrus)
 4.504.28.00051 32 3 37 –12 
207124.854.57.000–45 40 –5(L) limbic lobe (anterior cingulate, sub-gyral); frontal lobe (precentral; inferior/middle/superior/medial frontal gyrus)
 4.474.24.000–26 52 9–19 47 25 
66646.485.89.000–27 –9 10(L) sub-lobar; lentiform nucleus; claustrum
15944.414.20.000–27 –53 1(L) limbic lobe (posterior cingulate, parahippocampal gyrus); sub-lobar (caudate); occipital lobe (lingual gyrus, middle occipital gyrus)–25 –58 8 
 3.743.60.001–24 –44 2 
923.213.12.005–67 –55 21(L) temporal lobe (superior temporal gyrus); (L) parietal lobe
     (Supramarginal Gyrus)
5703.573.45.00232 42 18(R) frontal lobe (sub-gyral, middle/superior frontal gyrus)
373.143.05.00618 50 19(R) frontal lobe (superior/medial frontal gyrus)
6204.344.13.00059 –67 30(R) temporal lobe (middle/ superior temporal gyrus); parietal lobe
 3.323.22.00461 –65 20(angular gyrus, supramarginal gyrus, inferior parietal lobule)
24364.694.44.000–5 –7 28(R/L) limbic lobe (anterior cingulate; cingulate gyrus); frontal lobe (sub-gyral)
 4.354.15.000–3 1 26 
22603.543.42.00259 4 31(R) parietal lobe (postcentral gyrus); frontal lobe (precentral gyrus inferior/middle frontal gyrus)
 3.493.38.00357 –4 26 
 3.383.27.00453 2 37 
413.012.93.008–26 –93 26(L); occipital lobe (cuneus)
135484.484.26.00015 22 54(R/L) frontal lobe (sub-gyral; middle/superior medial frontal gyrus)
 4.414.19.000–1 35 46 
 3.773.63.00113 40 46 
5353.413.30.003–41 23 46(L) frontal lobe (precentral gyrus, middle frontal gyrus)
703.062.99.007–26 –85 38(L) occipital lobe (cuneus); parietal lobe (precuneus)
3233.243.14.00510 –11 39(R) limbic lobe (cingulate gyrus)
1113.463.35.00337 –15 40(R) frontal lobe (sub-gyral, precentral gyrus)
3693.763.62.001–34 –48 38(L) parietal lobe (sub-gyral, supramarginal gyrus, inferior parietal lobule)
2273.723.59.00167 –26 43(R) frontal lobe (precentral gyrus)
1383.213.12.005–10 –21 40(L) limbic lobe (cingulate gyrus)
3043.413.31.00342 18 49(R) frontal lobe (middle frontal gyrus)
3963.213.12.005–24 –4 60(L) frontal lobe (sub-gyral, middle frontal gyrus)–32 –3 53 
362.962.89.00933 –4 52(R) frontal lobe (middle frontal gyrus)
4813.223.13.005–31 –27 64(L) parietal lobe (postcentral gyrus); frontal lobe (precentral gyrus)
853.093.01.0072 –15 62(R) frontal lobe (medial frontal gyrus)
873.133.05.00618 –13 64(R) frontal lobe (superior/middle frontal gyrus)
2043.103.02.00722 –31 68(R) parietal lobe (postcentral gyrus); frontal lobe (precentral gyrus)

We performed the multivariate analysis for each of 116 ROIs and CAG, age, and duration of disease were the independent variables (Supplementary Table 2). In summary, the CAG number, disease duration, and age were significant factors in the determination of the GMD in the following areas: areas 3, 4_5, and 6 of the cerebellum; left anterior cyngulate gyrus; inferior (bilateral), middle (left), and superior (left) frontal lobes; fusiform gyrus (bilateral); heschl gyrus (bilateral); insula (bilateral); lingual gyrus (bilateral); paracentral lobule (bilateral); postcentral gyrus (left); precentral gyrus (bilateral); precuneus (left); rolandic area (bilateral); supplementary motor area (left); inferior (left), middle (left), and superior temporal gyus (bilateral); all vermian areas.

We performed a further linear regression with stepwise backward regression in the areas that did not reach significance in the multivariate analysis, to evaluate if one of these variables was an independent factor in the determination of the ROI density. Age and CAG length were the most frequent factors in the determination of ROI densities while duration of disease was only an independent factor in the middle frontal gyrus-orbicular part (R) (Supplementary Table 3). Age was by far the most frequent independent determinant of ROI density.

Several ROIs were important determinants of the final ICARS score. They were localized in the cerebellum, vermis, frontal, parietal, temporal, and occipital lobes (Supplementary Table 4).

WM abnormalities were observed only in the deep cerebellar WM (Fig 1).

There were no changes in GMD and WMD between MRIs. The 19 patients, who also had longitudinal MRS studies, did not present significant differences in NAA/Cr values between studies (1.448 ± .299 versus 1.414 ± .325; P= .77). The NAA/Cr changes observed in each time point were similar to that of the previously described24 (Supplementary Table 5).


The main purpose of our study was to perform a comprehensive whole brain study in a large MJD/SCA3 series and to correlate the findings with clinical markers of the disease. Our second objective was to evaluate those changes longitudinally.

In addition to areas classically affected in MJD/SCA3, we were able to demonstrate by VBM, decreased GMD in the frontal lobes, parietal lobes, temporal lobes, occipital lobes, and limbic lobe. These findings corroborate previous neuroimaging and neuropathological studies, as well as confirm previous reports of cortical dysfunction in MJD/SCA3.13-15,18,21-23

A previous VBM study which contrasted the patterns of atrophy in SCA3 and SCA6 demonstrated a decrease in GM in the pons and the vermis, and WM reduction in the pons, surrounding the dentate nucleus, and the cerebellar peduncles in 9 patients with MJD; however these findings were not correlated to clinical findings.29 Another one aimed to identify potential biomarkers in the differentiation of SCA1, SCA3, and SCA6 by volumetric measurement and VBM.30 A reason for the discrepancy observed may be the smaller number of patients included in those studies, compared to our larger sample.

Neuropathological studies have probably underestimated the degree of cerebral involvement in MJD.21,31 One study demonstrates that the cerebral hemispheres are far more involved in MJD/SCA3 than generally acknowledged, including the: cerebellothalamocortical motor loop; basal ganglia-thalamocortical loop; visual, auditory, vestibular, oculomotor and sensory systems; precerebellar brainstem system; midbrain dopaminergic and cholinergic system; and pontine noradrenergic system. However, the neocortex was spared, as well as the striatum, major components of the limbic cholinergic and serotonergic systems and high-order autonomic brain centers.20 Conversely, we know both normal and mutant ataxin-3 expression is not confined to the cerebellum, spinal cord, and basal ganglia, but rather is present in all brain areas, including the cerebral cortex.32 We were unable to perform postmortem evaluation in any of our subjects to corroborate our findings. Two published studies using VBM in MJD/SCA3 did not describe cortical involvement in MJD/SCA3.29,30

Although classically MJD/SCA3 patients have been considered to have normal cognition, several studies have demonstrated either subclinical cognitive dysfunction or overt dementia.9,10,33, 34 The cerebellar cognitive affective syndrome is characterized by impairment of executive functions, shifting, verbal fluency, abstract reasoning, and working memory, associated with visuospatial deficits, personality changes, and language deficits.35 It is thought to represent dysfunction of the cerebellar circuit with the prefrontal, superior parietal, superior temporal, and limbic cortices. In our study, the cortical areas affected are mostly primary and association areas involved in the higher processing of all the systems mentioned above. The lack of cognitive measures in our study, however, precludes any further information on this matter. It is possible that cortical areas are only secondarily affected, or that effects vary among patients (producing a cancellation effect when averaging MRIs on smaller number of patients); thus atrophy of these regions would only be noticeable in a larger sample of patients or with a longer duration of disease.

Our findings in the WM are in agreement with neuropathological studies, since WM atrophy has only been reported in the cerebellum and brainstem.21 We previously reported axonal dysfunction in the deep WM of the cerebral hemispheres by means of MRS in areas of normal appearing WM.24 However, we did not find progression of these MRS abnormalities in the short-term follow up.

Previous studies suggested the atrophy to be heterogeneous in different areas of the brain as disease progresses. Horimoto et al36 studied 7 patients longitudinally through measurement of the midsagittal areas of infratentorial structures. While the atrophy of the pontine base and cerebellum significantly correlated with age, the atrophy in the midbrain and the pontine tegmentum showed no progression. In another study, the pontine tegmentum atrophy was present early in the clinical onset of the disease.18 This study also found that the quotient of atrophy of the pontine tegmentum divided by age correlated well with the CAG repeat number, while the area of the pontine base correlated negatively with disease duration. Particularly, the size of the pontine base remained in the range of controls for a relatively long time after the onset of symptoms.

Disease duration, CAG repeat length, and age were important factors in the determination of ROIs’ GMD. However, in several other areas only CAG and age played a role, while in others age was the only significant factor. Although it is possible that age is a factor in GM density, independent of MJD, this suggests that age is probably the most important determinant of progression of brain atrophy, while CAG would be the second most important. A previous neurophysiological study showed age-related decrease of CMAP and SNAP amplitudes in SCA3/MJD greater than in normal subjects.37 These data suggested that the degree of peripheral damage in SCA3/MJD was mainly associated with age, rather than disease duration and CAG length. Disease duration probably plays a more important role in younger patients, with larger CAG numbers. The importance of CAG is well-known since a correlation between CAG length and disease severity and an inverse correlation between age of onset has been demonstrated in several studies and was also observable in our population (data not presented).

Even though ICARS has major disadvantages as a scale38,39 it seems to be a reliable measurement of disease severity and possibly brain pathology, since multiple areas, including cortical ones, were strong determinants of the final ICARS score. Prior studies have clearly demonstrated the interference of non-ataxic symptoms in the total final score, suggesting ICARS may in fact be a measure of compromise not restricted to ataxia.39 Another possibility is that the atrophy of these cortical areas may occur concomitantly with the cerebellar pathology, resulting in an indirect finding. SARA was not published when we first started collecting our data.40

The large clusters observed may raise concerns. However, those were specifically observed in the subcortical GM nuclei, cerebellum and brainstem, areas clearly atrophic in these patients, especially if disease duration is taken into consideration. Several patients with MJD also present frontal atrophy in routine MRI examination, a finding previously published in our own sample.24

We were unable to identify areas of progression of the atrophy after an interval of approximately 1 year. A similar longitudinal study in Huntington's disease performed at our laboratory with a similar interval demonstrated clear progression of the GM atrophy.41 The most likely reason for our results is floor effect. As the mean duration of disease in our patients was 10 years, the degree of atrophy was already so pronounced that minor progressive changes would not be recognizable. Also, the interval between MRIs was probably short. Although there was clinical progression within this period, as demonstrated by higher ICARS scores, the clinical deterioration was never marked and not universally present in our series.

Intuitively, one may not find the results of our longitudinal study surprising. However, a longitudinal study of this magnitude in MJD/SCA3 had never been performed. Future longitudinal neuroimaging studies in MJD/SCA3 should concentrate on pre-symptomatic or early-symptomatic patients, or include a much longer follow-up. The long duration of disease in our patients may have led to an overestimation of the degree of compromise in some areas, which are probably affected only as disease progresses, and are not necessarily the primary target of pathology. We were unable to find differences within different subtypes (data not shown), even though we had a reasonable number of subjects and representation of all major subtypes. This was probably due to the small number of subjects within each subgroup.

In summary, we were able to show a widespread decrease in GMD confirming multiple smaller reports of cortical involvement and dysfunction. MJD/SCA3 should no longer be considered a disease limited to the cerebellum and its connections, but rather a pathology involving all levels of the brain.


The authors acknowledge Dr Joseph Friedman for reviewing this manuscript. This project was funded by FAPESP.