Progression of Cerebellar Atrophy in Spinocerebellar Ataxia Type 2 Gene Carriers: A Longitudinal MRI Study in Preclinical and Early Disease Stages

Spinocerebellar ataxias type 2 (SCA2) is an autosomal dominant inherited disease caused by expanded trinucleotide repeats (≥32 CAG) within the coding region of ATXN2 gene. Age of disease onset primarily depends on the length of the expanded region. The majority of subjects carrying the mutation remain free of clinical signs for few decades (“pre-symptomatic” stage), but in proximity of disease onset subtle neurophysiological, cognitive, and structural brain imaging changes may occur. Aims of the present study are to determine the time-window in which early clinical and neurodegenerative MRI changes may be identified, and to evaluate the rate of the disease progression in both preclinical and early disease phases. We performed a 1-year longitudinal study in 42 subjects: 14 SCA2 patients (mean age 39 years, disease duration 7 years, SARA score 9 points), 13 presymptomatic SCA2 subjects (preSCA2, mean age 39 years, expected time to disease onset 16 years), and 15 gene-negative healthy controls (mean age 33 years). All participants underwent genetic test, neurological examination, cognitive tests, and brain MRI. Evaluations were repeated at 1-year interval. Baseline MRI evaluations in SCA2 patients showed significant atrophy in cerebellum, brainstem, basal ganglia and cortex compared to controls, while preSCA2 subjects had isolated volume loss in the pons, and cortical thinning in specific frontal and parietal areas, namely rostral-middle-frontal and precuneus. One-year longitudinal follow-up demonstrated, in SCA2 patients, volume reduction in cerebellum, pons, superior cerebellar peduncles, and midbrain, and only in the cerebellum in preSCA2 subjects. No progression in clinical or cognitive measures was observed in preSCA2 subjects. The rate of volume loss in the cerebellum and subcortical regions greatly differed between patients and preSCA2. In conclusion, our pilot study demonstrated that MRI measures are highly sensitive to identify longitudinal structural changes in SCA2 patients, and in preSCA2 up to a decade before expected disease onset. These findings may contribute in the understanding of early neurodegenerative processes and may be useful in future therapeutical trials.


INTRODUCTION
Spinocerebellar ataxia type 2 (SCA2, OMIM 183090) is an autosomal dominant inherited disease caused by trinucleotide repeat expansion in the coding region of ATXN2 gene. SCA2 is classified among the group of polyglutamine disorders (1), and accounts for ∼10-25% of dominant familial cases. SCA2 is particularly frequent in Cuba, Mexico, Korea, India, Spain, and Italy (2). The usual clinical presentation is a progressive adult onset cerebellar syndrome associated with saccadic slowing, axonal sensory neuropathy, autonomic dysfunctions, motor neuron involvement, and cognitive abnormalities (1). Less frequently SCA2 patients may present a dopa-responsive Parkinsonism (3).
As in other polyglutamine diseases, the age at onset is inversely related to the number of triplet repeats, and the subjects carrying the mutation may remain free of symptoms and clinical signs for a few decades of their life ("asymptomatic or presymptomatic" stage). Clinical disease onset is characterized by the first appearance of gait difficulties or other obvious cerebellar features ("ataxic" stage). Between the asymptomatic and the overt ataxic stage, gene mutation carriers experience a "preclinical or prodromal" stage in which unspecific neurological symptoms are recognized, and early neurophysiological, cognitive, and structural brain imaging changes may be measured (4)(5)(6)(7). In particular voxel-based morphometry was shown to represent a potential source of biomarkers for disease severity and progression (8,9). Neuroimaging studies in SCA2 demonstrated gray matter (GM) volume loss in the cerebellar hemispheres and vermis, brainstem, and thalamus, and also in the inferior frontal, parietal, and mesial temporal structures, as compared to controls (10)(11)(12). White matter damage was also observed in the corticospinal tracts, superior and inferior longitudinal, uncinated, and inferior occipital fasciculi, besides the cerebellar and brainstem regions compared to controls (13,14). Gray matter loss in brainstem and cerebellum, particularly lobules V and VI, and in frontal cortex in SCA2 are already detectable in prodromal disease stages (4,(15)(16)(17).
At present only a few studies have evaluated preclinical SCA2 subjects (4,5,(16)(17)(18). There is the need for longitudinal clinical and neuroimaging data for the implementation of future clinical trials on disease-modifying interventions (19). In this study, we prospectively followed a cohort of preclinical and early symptomatic SCA2 mutation carriers with the aims of establishing (i) the rate of the disease progression in preclinical and early disease phases; (ii) the time-window in which structural brain changes can be observed; and (iii) the correlations between clinical and neuroimaging outcome measures.

Participant Groups
Between February 2015 and March 2017, we screened 44 adult subjects belonging to SCA2 diagnosed families. Fifteen individuals had overt ataxic signs and confirmed SCA2 genetic diagnosis, 14 were asymptomatic gene-mutations carriers (preSCA2) and 15 were gene-negative healthy controls (CTR). All participants were over 18 years and gave their written informed consent before undergoing study procedures. The study was approved by the local Ethic Committee.
Trinucleotide CAG repeated region in ATXN2 (SCA2, MIM 183090) was analyzed as previously described (20). Participants who did not present ataxic symptoms were informed to have 50% risk of carrying the SCA2 expansion, and they had the choice to know about their genetic status in agreement with the genetic counseling protocol for predictive genetic test.
Presence and severity of ataxia was evaluated using the Scale for Assessment and Rating of Ataxia (SARA) (21). SARA score was used to differentiate patients (score ≥3) from asymptomatic subjects (score <3) (17). To assess neurological signs other than ataxia, we used the inventory of non-ataxia signs (INAS) (22).
Cognitive assessment was performed using a battery of cognitive tests including: Mini Mental State Examination Test (MMSE), digit span forward, copy and delayed recall of the Rey-Osterrieth complex figure (ROCF), phonemic (letters F-P-L) and semantic fluency tests (animals-fruits-car brands), and Symbol Digit Modalities Test (SDMT) (23)(24)(25). Scores were corrected for age, sex and education, according to the normative data. Disease duration was calculated based on subject's age at enrolment and age of ataxia manifestation. For preSCA2, age of expected disease onset was estimated according to the model described by Tezenas du Montcel et al. (26).
Exclusion criteria were the presence of psychiatric diseases, neurologic abnormalities others than spinocerebellar ataxia, claustrophobia, substance abuse, hydrocephalus, traumatic brain injury or intracranial mass. CTR subjects, SCA2 patients, and 9 out of 13 preSCA2 subjects repeated clinical evaluations and brain MRI at 1-year interval (±2 months).
A region of interest (ROI) analysis on volume and thickness for cerebellum, on volume for subcortical regions, and on thickness for cortical regions, was conducted at baseline and follow up 3D T1-weighted images.
For cerebellum assessment, a pre-processing including a de-noising of the images, an intensity inhomogeneity correction, and a cropping step to limit the processing to the cerebellum was performed and a cerebellum lobule segmentation was obtained using CERES software (27) (see Supplementary Material). Volume and mean cortical thickness for each cerebellar lobule and for the whole cerebellum were extracted.
For evaluation of cerebral cortex, subcortical nuclei (i.e., bilateral putamen, caudate, pallidum, thalamus), corpus callosum, and brainstem substructures (i.e., medulla oblongata, pons, midbrain, and superior cerebellar peduncles), FreeSurfer software (http://freesurfer.net, version 6) was used (28,29). White matter and pial surface were generated and manually corrected by an expert operator (C.P.) when errors where identified. Cortical thickness was calculated as the distance between the vertices of white and pial surface. The segmentations were performed according to Fischl et al. (30) for subcortical nuclei and corpus callosum, and according to Iglesias et al. (28) for brainstem (see Supplementary Material for extensive description). The volume of each segmented subcortical region was obtained. Due to no hypothesis on atrophy lateralization, we computed the mean of the left and right measures for the bilateral cerebellar lobules and subcortical nuclei. For the total volume of the corpus callosum, we summed the volumes of genu, body, and splenium segments.
For cortical regions, we assumed that the possible atrophy in preSCA2 subjects begins in the same regions where the atrophy is detectable in SCA2 patients. To this aim, a vertex-wise analysis using a General Linear Model approach was conducted to detect significant cortical thinning in SCA2 in comparison to CTR. To correct for multiple comparisons a Monte Carlo simulation was applied (cluster threshold: p < 0.01). Significant clusters of cortical thinning in the comparison SCA2 were neuroanatomically labeled based on the Desikan-Killiany cortical atlas and identified as cortical ROIs. Mean cortical thickness for each cluster (i.e., cortical ROI) was extracted for each participant (CTR, preSCA2, and SCA2) at baseline and follow-up evaluation. All extracted volumes (i.e., whole cerebellum, lobules of cerebellum, and subcortical regions) were expressed as percentage of total intracranial volume (TIV), to account for inter-subject variability (31).

Statistical Methods
Baseline clinical (SARA score, total INAS count, cognitive scores) and sMRI quantitative data (i.e., volumes of the subcortical regions, mean thickness and volume of total cerebellum, and cerebellar lobules) were compared in preSCA2 and SCA2 groups with CTR using Wilcoxon rank-sum test.
Differences with CTR were expressed as percentage, using the following formula: [(Median Participant groupmedian CTR )/median CTR ]·100, where the participant group was preSCA2 or SCA2.
To detect changes between baseline and 1-year follow-up evaluations within each subject category (i.e., preSCA2, SCA2, CTR), the Wilcoxon signed-rank test was used. The difference with baseline, was calculated as percentage of change using the following formula: [(Median at1−year follow−up -median at baseline )/median at baseline ]·100.
To explore the correlation between baseline measures and disease duration, we calculated Spearman's rank correlation coefficient. Correlations were calculated separately for the group of SCA2 patients and for preSCA2 subjects. For patients we used disease duration calculated as the difference between age at enrollment and age of ataxia manifestation. For pre-manifest subjects, we considered years-to-disease onset, calculated on the basis of CAG repeat length (26). We also analyzed the evolution   Figure. The Wilcoxon rank-sum test for quantitative data, were used to detect differences between SCA2 groups and controls: *p-value <0.05 compared to controls; **p-value <0.001 compared to controls.
of clinical and sMRI outcome measures combining preSCA2 and SCA2 groups and considering, as a unified variable of time, the years of disease duration for the patients and years before onset in premanifest SCA2 gene carriers. As exploratory analyses, we tested for linear, quadratic and cubic effects of time and chose the best fitting model. The standardized response means (SRM), calculated as mean score change/standard deviation of change, was indicated as effect size index to enable comparison between scale and volumetric measures (32).
P-values of <0.05 were considered as statistically significant. Statistical correction for multiple comparisons was not applied owing to the small sample size and the exploratory nature of the study. Statistical analyses were conducted using either SAS software, version 9.4 or JMP R version 11.0.

Baseline Evaluations
Among the 44 screened subjects, 42 were enrolled for the study. One screened SCA2 patient was not enrolled because refused MRI due to claustrophobia, and one preSCA2 subject was excluded because neuroimaging investigations showed the presence of Arnold-Chiari malformation. Baseline clinical characteristics of participants are summarized in Table 1A. At enrollment SCA2 patients had a mean age of 39 years, and a disease duration of 6.8 years. PreSCA2 subjects had a mean age of 39 years, and they were very far from disease onset with a median estimated time before symptoms of 16.6 years (range 10-18). The CTR group had a men age of 33 years.
In preSCA2 subjects, cerebellar volume and cortical thickness did not differed from those of CTR. sMRI of subcortical brain structures showed a decrease in pons volume in comparison with CTR (−9.8%, p = 0.01) ( Table 1B).
Vertex-wise analysis of cortical thickness showed significant differences between SCA2 patients and CTR in bilateral frontal regions comprising rostral middle frontal, caudal middle frontal, superior frontal areas, and in right precuneus and fusiform regions (Figure 1). Mean cortical thickness extracted from these cortical ROIs confirmed the significant differences between SCA2 and CTR, and a gradual reduction from preSCA2 to SCA2. In preSCA2 subjects, we observed a significant difference in bilateral rostral middle frontal regions (−6.5% thickness in the left side, and −4.5% in the right side) and in right precuneus (−6.9%) in comparison with CTR (Table 1C).

Longitudinal Evaluations
At 1-year follow up, SCA2 patients showed a mean increase of 1.0 point in SARA score, and a mean increase of 0.86 point in INAS count, but the differences with baseline evaluations were not significant. Within-group longitudinal analyses of cognitive tests in SCA2 patients showed no effect of time, except for the ROCFcopy score that was significantly decreased at 1-year interval (p = 0.004). PreSCA2 subjects showed no changes in clinical and cognitive quantitative scores ( Table 2).
In preSCA2, sMRI measures did not show significant differences except for total cerebellar volume, that was minimally but significantly reduced in comparison with baseline (p = 0.022).
In both groups, SCA2 and preSCA2, no longitudinal changes were observed for cerebellar cortical thickness (Supplementary Table 2B). For vertex longitudinal analyses, one SCA2 patient was excluded due to fail of white matter/gray matter segmentation in Freesurfer. Comparison between baseline and follow-up showed no significant changes over time in cerebral cortical thickness.

Brain Imaging and Clinical Correlations
To investigate the rate of decline during the symptomatic and pre-manifest phases of the disease, we analyzed the correlations between disease duration and clinical and sMRI measures. We analyzed only the quantitative data that, at baseline, were significantly different in SCA2 patients in respect to CTR. In the group of SCA2 patients, years of disease duration correlate with SARA scores (ρ = 0.90; p < 0.0001), with SDMT scores    *Difference with baseline, the following formula was used: [(Median 1−year follow−up − median baseline )/median baseline ]·100. ∞ p-value: Wilcoxon signed-rank test was used to detect time effect, between baseline and follow-up examinations. In bold, p values < 0.05.
The same measures considered in patients were tested in pre-manifest subjects. The estimated time-to-onset was used for the variable of time. In preSCA2, years-to-onset were significant correlated with cerebellar cortical thickness (ρ = −0.61; p = 0.027). Trends toward pathological changes in clinical scores (SARA and SDMT), and volume changes (pons and cerebellum) were observed in presymptomatic subjects with proximity to disease onset (Figure 2).
Changes in outcome measures showed different rates of progression in the symptomatic and in the presymptomatic phases, with very low progression of cerebellar atrophy and ataxia clinical score in the preclinical phase, and with more rapid decline after the full manifestation of the disease (Figure 3) (Supplementary Table 3).
The evaluations of SRM for clinical and sMRI measures showed that changes of SARA and INAS scores over time had small effect size in both SCA2 patients (SRM: 0.39 and 0.33), and in preSCA2 subjects (SRM: 0.33) ( Table 3). sMRI metrics demonstrate large effect sizes for total cerebellar volume in both preSCA2 (SRM: −0.97) and in affected SCA2 patients (SRM: −1.12), while brainstem volumes, and in particular pons volume, had a large effect size in patients (SRM = −2.16), but not in the preclinical subjects (SRM = 0.18).

DISCUSSION
In this study, we prospectively followed a cohort of preclinical and early symptomatic SCA2 mutation carriers, over 1-year period, using quantitative clinical and brain MRI measures. SARA score efficiently followed the progression of ataxia in SCA2 patients, with a mean increase of 1.0 point at 1-year longitudinal evaluation, but the scale was not effective in detecting abnormalities or progression in preSCA2 subjects. These figures are in agreement with previous observational studies showing a yearly increase of 1.0-1.75 point of the SARA scores in patients. The progression rates were shown to be different in the early and in the late stages of the disease (33,34).
The inventory of non-ataxia signs (i.e., INAS total count), differently than SARA, revealed a small but noticeable difference between preSCA individuals and healthy controls. In agreement with previous observations, we found that muscle cramps, areflexia, and extensor plantar reflex are present in preSCA2 subjects many years before the ataxic signs (4,17,22).
Cognitive evaluations showed that both SCA2 patients and preSCA subjects had significant lower scores than CTR in SDMT. SDMT has not been extensively used in SCAs, but is has been proposed as a cognitive outcome measures in several neurological disorders, including Huntington disease, Friedreich ataxia and Fragile X-Associated Tremor/Ataxia syndrome (40)(41)(42). SDMT is a measure of speed and efficiency in multiple cognitive processes involving memory, word retrieval, and executive function. We hypothesize that this test may be a sensitive measure to detect very early deficits of executive functions also in SCA patients, but further studies are needed to confirm this preliminary observation. In the battery of cognitive tests, a time-effect in the was identified only for the ROCF-copy score for SCA2 patients. Previous longitudinal studies in symptomatic patients have also shown very slow progression of cognitive decline (24,43).
In preclinical SCA2 mutation carriers, cognitive characteristics have been extensively studied only in a large cohort from Cuba (4). This latter study showed executive function impairment in preSCA2, however, due to different inclusion criteria, the enrolled cohort included a percentage of subjects with overt ataxic signs (SARA up to 4 points).
The neuroimaging data of our study largely confirmed previous finding showing that cerebellum, brain medulla, pons, and superior cerebellar peduncles are consistently affected in SCA2 patients, giving the characteristic imaging of olivoponto-cerebellar atrophy (OPCA) (10-13, 16, 44-48). We also confirmed that atrophy was more severe in cerebellar lobules of the anterior lobe and superior posterior lobe, predominantly associated with motor control (17). Our study also confirmed, in symptomatic patients, volume loss in basal ganglia (thalamus, putamen, and pallidum) and corpus callosum (12,13,47). At 12-months-longitudinal evaluations, we observed a significant increment of the atrophy both in cerebellar and pontine regions, and in subcortical structures. Similar findings were previously reported in longitudinal studies having much longer follow-up intervals (range 2-6 years) (49,50).
Only few MRI reports are available for the presymptomatic phase (4,16,17). These observational studies revealed mild cerebellar and brainstem volume loss in subjects carrying the disease mutation, with volume loss being more pronounced in the individuals closest to the expected onset of ataxia (16). One longitudinal MRI study, performed in a subgroup of 5 SCA2 gene carriers, showed significant decrease in cerebellar and pons volumes at 6-years interval (18). In our cohort of preSCA2 subjects, we found that the pons was the brain structure primarily affected by volume decrease in respect to healthy CTR, but at 12-months interval no significant progression of the atrophy could be detected. On the other hand, total cerebellum volume did not differ from healthy controls, both at baseline and at follow-up, but within the preSCA2 group, we observed a mild, but significant, decrease in total cerebellar volume at longitudinal analysis.
On the contrary, analyses of cerebellar cortical thickness failed to show significant effect of time within SCA2 groups. This is a major novelty of our study, since there are no previous data reporting changes in cerebellar cortical thickness for premanifest or SCA2 patients. Our findings are analogous to the data reported in SCA1 and SCA3 diseases, in which cerebellar cortical thinning was detected only in affected patients with a disease duration of at least 5 years (51, 52). Our SCA2 patients had a mean disease duration of 6.8 years, and they present significant reduction of cortical cerebellar thickness both at baseline and follow-up measures in comparison with controls, but no measurable change could be detected at 12-months evaluation. These data suggest that: (1) cerebellar cortex is part of the neurodegenerative pattern in SCA2, analogously to SCA1 and SCA3 diseases, (2) cortical thinning is detectable only several years after disease onset, and (3) the progression of cerebellar cortex thinning appears much slower than that observed for cerebellar volumes. Overall these data suggest that cerebellar volume rather than thickness would be a better biomarker to track the progression of SCA2 disease, including the presymptomatic phase. It has to be noted that in our study all preSCA2 subjects were very far from disease onset (median 16 years, range 8-30 years). This aspect was not predetermined by our inclusion-exclusion criteria, but was unintentional due to unbiased recruitment of the at-risk relatives belonging to our SCA2 families. The advantage of this cohort was to have the opportunity of studying a very large time-window of the pre-ataxic stage, the disadvantage was the large heterogeneity.
We also have to consider the limitations of this work, mainly represented by the small sample size and the high rate of dropout in preclinical SCA2 participants at longitudinal evaluation. Both factors considerably reduce the study power, and may affect statistical analyses particularly in the preSCA population. Though these limitations highlight the need of multicenter studies, our results support the importance of specific clinical and MRI longitudinal observations in SCA gene carriers. Particularly, in preSCA2 cerebellar and pons volume loss is detectable ∼10-years from expected disease manifestation. Our extensive analysis across the whole brain, allowed us to demonstrate a significant trend of cortical thinning in bilateral rostralmiddle frontal and in right precuneus in preSCA2 subjects, suggesting that also these cortical areas could also be affected very early in the course of the disease. The rate of progression of cerebellar volume reduction is very different in the period before and after clinical onset, but it can be appreciated at 1-year interval. Structural MRI of these specific areas may allow early quantification of neurodegeneration, and may represent possible outcome measures to implement trial readiness for upcoming disease-modifying clinical trials.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by Comitato Etico Regione Lombardia Sezione Fondazione IRCCS Istituto Neurologico "Carlo Besta". The patients/participants provided their written informed consent to participate in this study.