Article Text
Abstract
Background: The mechanisms guiding the progression of neuronal damage in patients with Huntington disease (HD) are not completely understood. It is unclear whether the genotype—that is, the length of the expanded CAG repeat—guides the location and speed of grey matter decline once HD is clinically manifested. Moreover, the relationship between cortical and subcortical grey matter atrophy and the severity of motor symptoms of HD is controversial.
Objectives: In this article, we longitudinally studied, over the period of 1 year, a cohort of 49 patients with HD. We investigated: first, the clinical relevance of regional progressive grey matter atrophy; and second, the relationship between the ratio of atrophy progression and genotype.
Methods: The length of the expanded CAG repeat was quantified for all patients and the United Huntington’s Disease Rating Scale (UHDRS) was used to rate the severity of clinical symptoms. Grey matter atrophy was determined using voxel-based morphometry (VBM) of brain MRI. Progression of atrophy was quantified in 37 patients who were submitted to two different MRI scans, the second scan 1 year later than the first.
Results: Overall, patients exhibited progressive atrophy involving the caudate, pallidum, putamen, insula, cingulate cortex, cerebellum, orbitofrontal cortex, medial temporal lobes and middle frontal gyri. Patients with a larger UHDRS score exhibited selective atrophy of the caudate, thalamus, midbrain, insula and frontal lobes. Patients with longer, expanded CAG repeat sequences showed faster rates and more widespread atrophy, particularly those patients with more than 55 expanded CAG repeats.
Conclusions: These results confirm that brain atrophy progresses after the clinical onset of HD and that regional atrophy is related to symptom severity. Moreover, our results also indicate that intensity and rate of progression of brain atrophy are more pronounced in patients with larger, expanded CAG repeat sequences.
- huntington disease
- magnetic resonance image
- voxel-based morphometry
- brain atrophy
- expanded CAG repeats
Statistics from Altmetric.com
Huntington disease (HD) is an inherited neurodegenerative disorder that follows an autosomal-dominant pattern of transmission, usually manifesting itself during adulthood. The classical symptoms of HD are defined by a triad that is composed of: prominent motor dysfunction, cognitive deterioration and psychiatric symptoms. The physiopathology of HD can be attributed to the abnormal protein Huntingtin, which has an enlarged polyglutamine stretch that is believed to promote neuronal apoptosis in a variety of brain regions, particularly within the basal nuclei.1 It is well known that the most affected brain structure in HD patients is the striatum.1 However, the mechanisms underlying the regional preferences for grey matter loss are still controversial.
It is not completely clear how neuronal loss is distributed outside the basal nuclei. MRI studies using manual morphometry of brain structures have confirmed that the caudate, putamen and globus pallidum are significantly atrophied in patients with HD.2 Pre-symptomatic at-risk subjects also show volume reduction of these structures, which is more intense closer to the onset of symptoms.3 4 However, the results from morphometric analyses of the whole brain are less consistent. For instance, studies using a voxel-wise comparison of grey matter maps have shown different patterns of atrophy. Reduction in grey matter volume was observed in the left striatum, bilateral insula, dorsal midbrain and bilateral intra-parietal,5 but also in the hypothalamus, opercular cortex and right paracentral lobule,6 and in thalamic dorsomedial subnucleus.7
In this study, we evaluated the progression of neuronal damage in patients with HD. We performed a longitudinal analysis of grey matter loss. We aimed to investigate whether the distribution and speed of neuronal loss in HD is related to the genotype—that is, the length of the expanded CAG repeat. We also explored the clinical relevance of grey matter atrophy by investigating the association between severity of symptoms, defined by the Unified Huntington’s Disease Rating Scale,8 and the regional distribution of atrophy.
MATERIALS AND METHODS
Subjects
We studied 49 consecutive patients with clinical and molecular diagnosis of HD and who were followed up at the Neurogenetics Clinic of State University of Campinas, São Paulo, Brazil. We excluded patients with significant co-morbidities, such as alcoholism, head injury, arterial hypertension, diabetes mellitus, stroke, drug abuse and malignancy. Neurological exams were performed in all patients by the same investigator (HHR). All subjects signed a written informed consent to participate in this study, which was approved by the Ethics Committee of our University (UNICAMP).
Clinical and molecular investigation
We used a structured clinical questionnaire for patients and family members. Age of onset of HD was determined as first occurrence of symptoms associated with HD, which did not happen as an isolated incident, but rather heralded a progressive clinical decline. Specifically, patients were asked about the onset of symptoms, such as chorea, rigidity, irritability, sleep disturbance, frequent falls, sexual dysfunction, loss of energy, altered social behaviour and failing memory. Each subject underwent formal motor examination according to the motor component of the Unified Huntington’s Disease Rating Scale (UHDRS).8 UHDRS was performed by an experienced neurologist (HHR), who is an attending physician at the HD clinic at State University of Campinas. Molecular testing was performed following international standard protocols for laboratory testing of HD9 at the Molecular Genetics Laboratory at the State University of Campinas, São Paulo, Brazil.
Statistical analyses
We used analyses of variance (ANOVA) to compare continuous variables between groups, and p values less than 0.05 were accepted as significant. We used linear regression to assess the relationship between age at onset of disease and CAG expansion. For the clinical score analyses, we used Spearman correlation test. Statistical differences between subgroups with respect to the particular criterion were calculated and tested using the F test.
IMAGING
Voxel-based morphometry (VBM)
We applied a refined VBM protocol, aiming to investigate longitudinal grey matter volume changes within patients with HD. VBM was performed on T1-weighted magnetic resonance volumetric images with 1 mm isotropic voxels. All images were acquired on the same Elscint Prestige 2 Tesla scanner using a spoiled gradient-echo sequence (TR = 22 ms, TE = 9 ms, flip angle = 35°, matrix = 256 × 220). We used Analyze® format images that were generated from raw Dicom images using MRIcro.10 The VBM analysis was performed using modified routines present in the SPM2 software package (Wellcome Department of Cognitive Neurology, www.fil.ion.ucl.ac.uk).11 For normalisation and segmentation purposes, we used a customised template with tissue priors constructed from images from the 96 normal healthy volunteers recruited from our local community, ranging from 9 to 67 years (mean = 30 (SD 11), median = 27). Prior images and the template were convolved with an Isotropic Gaussian Kernel (IGK) of 8 mm and were used for optimising the non-linear normalisation of the raw skull-stripped images that were analysed by VBM. We opted for not using a HD-specific template, as there is considerable variability in the extent and location of atrophy exhibited by patients with HD. Hence, we decided to use a template constructed from a large number of healthy individuals as a form to control for the demographics of our local population and for scanner-specific magnetic field homogeneities. Moreover, a HD-specific template would probably be largely influenced by extreme cases of atrophy and possibly require a large number of subjects to preserve the authenticity of disease-specific variations. Therefore, a template based on a healthy population with less variability permits a more robust normalisation and segmentation of HD images. As explained below, all HD images were submitted to pre-processing using the same template, and we then applied a modulation parameter to correct for the deformations encountered during registration. In this way, we intended to maximise the accuracy of spatial normalisation and segmentation, while at the same time preserving the genuineness of the HD.
Forty-nine patients with HD (mean age = 43 years (SD 13.3), ranging from 4 to 73 years) were scanned once, and the UHDRS was computed within a month of the scanning session. Out of these 49 patients, 37 were scanned 1 year later than the first scan (mean age = 44 years (SD 12), ranging from 4 to 73 years). All patients were submitted to the same scanning protocol both in the first and second scan.
All images were submitted to the same pre-processing steps, composed by spatial normalisation, segmentation, tissue modulation and smoothing. Spatial normalisation was performed using 16 non-linear iterations, medium regularisation and a 25 mm cut-off. Spatially normalised images were re-sliced to an isotropic 1 mm. The images underwent segmentation of grey matter using the built-in routines of SPM2, which estimate the probability that each voxel is grey matter. The segmented images were modulated, in order to preserve the quantity of tissue (eg grey matter), while ensuring a good spatial alignment between patients and controls. Finally, the images were convolved with an Isotropic Gaussian Kernel (IGK) of 10 mm in order to minimise gyral inter-individual variability. This smoothing converts maps of tissue classification probabilities (ie segments) into images of local grey matter concentration. After smoothing, the data represent the proportion of the volume under the smoothing kernel that has been classified probabilistically as grey matter. In addition, this smoothing renders grey matter concentration normally distributed, enabling the application of standard statistical parametric mapping techniques. All scans were visually inspected for errors in normalisation and segmentation.
The normalised, segmented, modulated and smoothed data were analysed using SPM2. The following statistical analyses were performed:
1- The differences between the first and the second scan were assessed using a paired t-test. Thirty-seven patients were included in this analysis.
2- The rate of progression of the disease was quantified as the percentage change according to the formula: rate of progression (r) = ((i1−i2)/i1)*100, where i1 represents the first scan and i2 the second scan. Those patients who had both scans (excluding the patient with childhood onset)—that is, 36 patients—were grouped according to the length of their CAG repeat sequence. The ‘r’ images, symbolising the rate of progression for every subject, were then compared between subjects who had 40–45, 46–50, 51–55 and more than 55 CAG repeats. We performed comparisons between groups of individuals with repeat lengths above and below a certain threshold—for example, subjects with more than 45 repeats versus subjects with less than 45 repeats, and so forth.
3- The regional grey matter volume for the 49 patients with HD who underwent the first scan was correlated with their UHDRS score. This analysis was performed by a whole-brain voxel-based simple regression. In order to improve power, we also performed a region of interest (ROI) analysis of the relationship between the grey matter densities of the basal nuclei and clinical performance. ROIs were defined according to the freeware library Anatomical Automatic Labeling (http://www.cyceron.fr/freeware/). We investigated the grey matter density in ROIs corresponding to the thalamus, putamen, caudate and pallidum. The software MarsBar was used to extract the mean grey matter probability in each ROI,12 and data was exported to SPSS (v12); where possible, significant linear and logarithmic regressions between the grey matter within the basal nuclei and the UHDRS were investigated.
Statistical analyses 1 (paired t-test) and 3 (correlation) were performed using the statistical module of SPM2, and analysis 2 (simple t-test) was performed with the in-house developed software NPM (http://www.sph.sc.edu/comd/rorden/npm/), using the non-parametric Brunner and Munzel test.13 Importantly, regarding analysis 2, when analysing groups with large CAG repeats, we were limited to small sample sizes. This could have led to reduced statistical power, thereby increasing the likelihood of false-negatives. Therefore, we opted to use a non-parametric test to avoid overestimation of group effects of outliers, at the expense of reduced statistical power.13
Results were corrected for multiple comparisons using a false discovery rate (FDR) corrected statistical threshold of p<0.05.
RESULTS
Clinical and molecular investigation
We studied 49 patients with clinical and molecular diagnosis of HD (21 women and 28 men) (demographics are summarised in table 1). The mean age was 43.2 years (ranging from 4 to 73 years) and age at onset varied between 2 to 66 years (mean 35.3 years). The expanded CAG repeat at the HD gene ranged from 40 to 69 (mean 47.2 CAGs) and normal alleles varied from 18 to 26 CAG units (mean 22 CAGs). There was no significant difference in the size of the expanded alleles between affected men (mean 45.7) and affected women (mean 49.1) (p = 0.08).
There was a significant negative correlation between age at onset of HD and length of expanded CAG repeat (p<0.001, r = −0.57) (fig 1), but not between duration of disease and length of expanded CAG repeat (p = 0.54, r = −0.08).
The UHDRS scores ranged from 15 to 128 (mean 65.5). There was a significant correlation between the UHDRS score and age at onset (p = 0.011, r = −0.33) and expanded CAG repeat (p = 0.034, r = 0.26) (fig 1).
Imaging
Progressive atrophy
A significant mean difference in grey matter was observed in the second scan compared with the first scan (a histogram showing the overall reduction in brain size is shown in fig 2), and confirmed by a paired t-test. On average, the second scan exhibited a widespread reduction of grey matter in regions such as the caudate, pallidum, putamen, insula, cingulate cortex, cerebellum, orbitofrontal cortex, medial temporal lobes and middle frontal gyri (fig 3).
Rate of atrophy
There were 16 patients with 40–45 expanded CAG repeats (with mean age = 50 (SD 11.2) and mean duration of HD = 8.7 (SD 4)), 12 patients with 46–50 expanded CAG (mean age = 43 (SD 8) and mean duration of HD = 7.5 (SD 3.2)), 6 with 51–55 expanded CAG (mean age = 33 (SD 14) and mean duration of HD = 7.2 (SD 4.7)) and 2 patients who had more than 55 expanded CAG repeats (both were 21 years old and the mean duration of HD was 8.5 years (SD 4.7)). We excluded from this analysis the patient who had an onset of HD during childhood, who had expanded alleles of 53 CAG repeats.
Although there was a significant difference between the age of onset among groups with different lengths of expanded CAG repeats (F(32,3) = 7.5, p<0.01), there was no difference in the duration of symptoms (F(32,3) = 0.3, p = 0.82).
The overall reduction in brain volume across different expanded CAG repeat length groups can be seen in figure 4. There was a significant difference in the rate of atrophy among different patient groups. Although all patients exhibited progressive atrophy, particularly within the basal ganglia and neocortex (fig 5), there was a striking increase in the rate of atrophy that was directly related to the length of the expanded CAG repeat. For example, patients with more than 45 repeats exhibited a higher degree of atrophy in frontal and cerebellar areas than patients with fewer than 45 repeats. In turn, patients with more than 50 repeats showed a more intense rate of atrophy in frontal, occipital, parietal and cerebellar regions. Furthermore, patients who had more than 55 CAG repeats had a widespread increase in the rate of atrophy, affecting virtually the whole brain, frontal, temporal, occipital and parietal areas (fig 6).
Relationship between regional atrophy and motor symptomatology
There was a negative correlation between regional atrophy and the score in the UHDRS. Whole-brain voxel-wise analysis showed that patients with a larger UHDRS exhibited more intense selective atrophy of the thalamus, midbrain, insula and frontal areas (fig 7).
ROI analysis confirmed the significant association between the thalamus and the score in the UHDRS, and also revealed a significant association between the caudate and the UHDRS (fig 8; table 2). There was not a significant association between grey matter concentration in the pallidum or putamen and the UHDRS score.
DISCUSSION
In this study, we aimed to investigate genotypic influences (length of expanded CAG repeat) in brain atrophy in patients with HD. A longitudinal analysis of grey matter decline in patients with HD revealed that progressive neuronal loss associated with HD affects the basal ganglia and widespread cortical areas. Importantly, the length of the expanded CAG repeat was crucial for determining the rate of atrophy. Patients with longer repeats exhibited a steep increase in the rate of damage. Moreover, we observed that atrophy to specific brain areas, such as the basal nuclei and frontal lobes, is related to the severity of disease and motor impairment. Post-mortem studies1 suggest that the first symptoms (both motor and cognitive) appear in the absence of overt neuronal cell loss, suggesting that impaired cognition is likely to be caused by a synaptic dysfunction rather than a consequence of neuronal cell death and it is possible that this dysfunction is regulated by CAG repeat expansion.
A recent VBM study of subjects with preclinical positive genetic testing for HD5 demonstrated grey matter atrophy in the striatum, insula, midbrain and parietal lobe in subjects with positive testing compared with subjects with negative testing. Even though the VBM results from that study were not controlled for multiple comparisons, possibly because of the mild atrophy of pre-clinical subjects, they indicated that the neuronal damage effects of the HD gene involves not only the basal ganglia but also widespread cortical areas. Moreover, these results indicate that grey matter atrophy in preclinical patients can be monitored using automatic brain analysis. These concepts are corroborated by Testa et al.14 who have suggested that VBM is marginally more accurate than ROI volumetry at detecting hippocampal atrophy in early Alzheimer’s disease relative to controls. In our study, we investigated patients with HD, who had already developed symptoms and detectable brain atrophy. Therefore, VBM was highly sensitive to the location, extent and progression of brain atrophy, and its clinical relevance. This is reflected by our reporting of only results that survived a stringent statistical threshold.
We demonstrated that atrophy is more intense and more widespread at 1-year follow-up in those patients whose expanded CAG repeats are longer, suggesting that atrophy is highly dependent on the size of the expanded CAG repeat.
Rosenblatt et al., studying a large survey of patients with HD,15 observed that subjects with a smaller number of expanded CAG repeats tended to follow a more benign course. They also concluded that expanded CAG repeat lengths have a small effect on the rate of progression that may be clinically important over time. However, in a different study, the same group observed a partial correlation between a post-mortem neuropathological assessment of neuronal loss, the Vonsattel score and the length of the expanded CAG repeat.16 In addition, there was a strong relationship between motor impairment scores and neuropathological changes, leading the authors to conclude that motor impairment is a good clinical measure of neuronal cell loss.
Ward et al., examining the effect of expanded CAG repeat length on motor and neuropsychological performance,17 observed a closer association between motor symptoms and the genetic profile, than between the length of the expanded repeat and neuropsychological performance. They suggested that the expanded CAG repeat length is more closely linked to changes in basal ganglia that predominate in early- to mid-stage HD than with cortical degeneration seen later in disease progression. Nonetheless, our results suggest that disease progression affects basal ganglia and isocortical areas, which in turn might indicate that the use of neuropsychological measurements might not be a sensitive way to evaluate cortical tissue loss. The cortical damage may disrupt crucial cognitive networks early on during the disease course and, therefore, after a threshold of crucial loss, the performance on neuropsychological tests may have a floor effect and not be a good reflection of tissue loss.18–20 Cognitive tests are of value as a marker of symptom onset, but might not be sensitive for later stages of disease progression. On the other hand, assessment of eye movement abnormalities, which is part of the routine clinical assessment of symptomatic patients with HD, may be detected in presymptomatic individuals.21 22
We have demonstrated that the severity of motor symptoms is directly related to tissue loss affecting specific brain areas. Possibly, a combination of thalamic atrophy and loss of fronto-striatal connections is crucial for generating the motor symptoms of HD.
Acknowledgments
Study supported by Fundaçao de Amparo è Pesquisa do Estado de Sao Paulo (FAPESP) (grant 01/05439-3). We gratefully thank patients and their relatives for their invaluable collaboration with the research.
REFERENCES
Footnotes
Funding: Drs Ruocco, Cendes and Lopes-Cendes received grants from CNPq (Conselho Nacional de Desenvolvimento Científico e Tecnológico), Brazil.
Competing interests: None.