Introduction
A substantial body of evidence indicates that in Huntington's disease (HD), a neurodegenerative cytosine–adenine–guanine (CAG) repeat movement disorder with dementia, abnormal brain structure can be detected as far as two decades prior to motor onset (Aylward et al. Reference Aylward, Nopoulos, Ross, Langbehn, Pierson, Mills, Johnson, Magnotta, Juhl and Paulsen2011; Tabrizi et al. Reference Tabrizi, Scahill, Durr, Roos, Leavitt, Jones, Landwehrmeyer, Fox, Johnson, Hicks, Kennard, Craufurd, Frost, Langbehn, Reilmann and Stout2011, Reference Tabrizi, Reilmann, Roos, Durr, Leavitt, Owen, Jones, Johnson, Craufurd, Hicks, Kennard, Landwehrmeyer, Stout, Borowsky, Scahill, Frost and Langbehn2012; Wolf et al. Reference Wolf, Thomann, Thomann, Vasic, Wolf, Landwehrmeyer and Orth2013). Substantial effort at present is dedicated to identifying therapeutic targets and developing treatments to delay the onset of the disease, or to slow down the progression of the disease once it has become clinically manifest (Ross & Tabrizi, Reference Ross and Tabrizi2011). Similarly important is the development of objective measures that reliably reflect disease progression since these biomarkers may serve as surrogate endpoints in interventional trials (Weir et al. Reference Weir, Sturrock and Leavitt2011). Multiple neuroimaging modalities have been explored in this regard in clinically manifest and preclinical carriers of the HD gene mutation (Hennenlotter et al. Reference Hennenlotter, Schroeder, Erhard, Haslinger, Stahl, Weindl, von Einsiedel, Lange and Ceballos-Baumann2004; Reading et al. Reference Reading, Dziorny, Peroutka, Schreiber, Gourley, Yallapragada, Rosenblatt, Margolis, Pekar, Pearlson, Aylward, Brandt, Bassett and Ross2004; Wolf et al. Reference Wolf, Vasic, Schonfeldt-Lecuona, Landwehrmeyer and Ecker2007, Reference Wolf, Vasic, Schonfeldt-Lecuona, Ecker and Landwehrmeyer2009b ; Zimbelman et al. Reference Zimbelman, Paulsen, Mikos, Reynolds, Hoffman and Rao2007; Klöppel et al. Reference Klöppel, Draganski, Siebner, Tabrizi, Weiller and Frackowiak2009a , Reference Klöppel, Henley, Hobbs, Wolf, Kassubek, Tabrizi and Frackowiak b ). Large-scale longitudinal studies have shown that imaging brain structure in manifest HD can track disease-related changes over time (Aylward et al. Reference Aylward, Nopoulos, Ross, Langbehn, Pierson, Mills, Johnson, Magnotta, Juhl and Paulsen2011; Tabrizi et al. Reference Tabrizi, Scahill, Durr, Roos, Leavitt, Jones, Landwehrmeyer, Fox, Johnson, Hicks, Kennard, Craufurd, Frost, Langbehn, Reilmann and Stout2011, Reference Tabrizi, Reilmann, Roos, Durr, Leavitt, Owen, Jones, Johnson, Craufurd, Hicks, Kennard, Landwehrmeyer, Stout, Borowsky, Scahill, Frost and Langbehn2012). Structural magnetic resonance imaging (MRI) could thus be a powerful objective measure that allows the reliable assessment of disease progression. Functional MRI (fMRI) can add the dimension of brain function to structural analyses and may be sensitive to very early neural changes in HD gene mutation carriers (Paulsen et al. Reference Paulsen, Zimbelman, Hinton, Langbehn, Leveroni, Benjamin, Reynolds and Rao2004; Wolf et al. Reference Wolf, Vasic, Schonfeldt-Lecuona, Landwehrmeyer and Ecker2007; Zimbelman et al. Reference Zimbelman, Paulsen, Mikos, Reynolds, Hoffman and Rao2007). fMRI may also provide new insights into the relationship between the dynamics of neural activity and brain and clinical measures, since these parameters may change over time in a disease-specific way (Klöppel et al. Reference Klöppel, Henley, Hobbs, Wolf, Kassubek, Tabrizi and Frackowiak2009b ; Wolf & Klöppel, Reference Wolf and Klöppel2013). At present, brain function in manifest HD has been investigated using task-based fMRI (Georgiou-Karistianis et al. Reference Georgiou-Karistianis, Sritharan, Farrow, Cunnington, Stout, Bradshaw, Churchyard, Brawn, Chua, Chiu, Thiruvady and Egan2007; Wolf et al. Reference Wolf, Vasic, Schonfeldt-Lecuona, Ecker and Landwehrmeyer2009b ; Gray et al. Reference Gray, Egan, Ando, Churchyard, Chua, Stout and Georgiou-Karistianis2013), which has to be carefully interpreted within its experimental framework. Task type, task performance and individual cognitive capacity have to be taken into account as much as task re-test effects (Raemaekers et al. Reference Raemaekers, Vink, Zandbelt, van Wezel, Kahn and Ramsey2007; Zandbelt et al. Reference Zandbelt, Gladwin, Raemaekers, van Buuren, Neggers, Kahn, Ramsey and Vink2008; Maitra, Reference Maitra2009; Wolf et al. Reference Wolf, Vasic, Sambataro, Höse, Frasch, Schmidt and Walter2009a ; Gray et al. Reference Gray, Egan, Ando, Churchyard, Chua, Stout and Georgiou-Karistianis2013).
In this study, we assessed brain activity in early manifest HD using measures of resting-state functional connectivity (Fox & Greicius, Reference Fox and Greicius2010; van den Heuvel & Hulshoff Pol, Reference van den Heuvel and Hulshoff Pol2010). This approach aims to identify temporally synchronous neural systems characterized by ongoing spontaneous modulations of the blood oxygen level-dependent (BOLD) signal in the absence of explicit stimulation (Biswal et al. Reference Biswal, Yetkin, Haughton and Hyde1995; Fox et al. Reference Fox, Snyder, Vincent, Corbetta, Van Essen and Raichle2005, Reference Fox, Corbetta, Snyder, Vincent and Raichle2006). Independent component analysis (ICA), a technique which can, without bias, decompose resting-state (rs)-fMRI data into multiple spatiotemporally distinct components (Calhoun et al. Reference Calhoun, Adali and Pekar2004, Reference Calhoun, Kiehl and Pearlson2008), has been successfully applied to investigate so-called resting state networks (RSNs). These neural systems are thought to reflect ‘intrinsic’ neural activity patterns (Kelly et al. Reference Kelly, Biswal, Craddock, Castellanos and Milham2012). Multiple RSNs have been robustly identified across multiple independent datasets; these include spatially stable patterns of sensorimotor, auditory, visual and several discrete lateral and medial–prefrontal systems (Damoiseaux et al. Reference Damoiseaux, Rombouts, Barkhof, Scheltens, Stam, Smith and Beckmann2006; Smith et al. Reference Smith, Fox, Miller, Glahn, Fox, Mackay, Filippini, Watkins, Toro, Laird and Beckmann2009; Fox & Greicius, Reference Fox and Greicius2010). Importantly, RSNs correspond closely to activity patterns underlying a wide range of distinct higher cognitive functions linked to frontal cortical control (Smith et al. Reference Smith, Fox, Miller, Glahn, Fox, Mackay, Filippini, Watkins, Toro, Laird and Beckmann2009; Laird et al. Reference Laird, Fox, Eickhoff, Turner, Ray, McKay, Glahn, Beckmann, Smith and Fox2012). Two recent studies provided evidence for abnormal cortico-striatal activity at rest in preclinical and clinically manifest carriers of the HD gene mutation (Seibert et al. Reference Seibert, Majid, Aron, Corey-Bloom and Brewer2012; Unschuld et al. Reference Unschuld, Joel, Liu, Shanahan, Margolis, Biglan, Bassett, Schretlen, Redgrave, van Zijl, Pekar and Ross2012). Impaired resting-state connectivity of cortical midline regions, often referred to as critical nodes of the ‘default-mode network’ (Broyd et al. Reference Broyd, Demanuele, Debener, Helps, James and Sonuga-Barke2008; Buckner et al. Reference Buckner, Andrews-Hanna and Schacter2008), was further reported for early manifest patients (Quarantelli et al. Reference Quarantelli, Salvatore, Giorgio, Filla, Cervo, Russo, Cocozza, Massarelli, Brunetti and De Michele2013). It is, however, noteworthy that these studies relied on seed correlation approaches that are influenced by the choice of the brain regions of interest, thus limiting the number of models and comparisons to be tested. Using multivariate statistics a very recent study reported increased activity in a mixed group of preclinical and early manifest HD patients in several cortical and subcortical RSNs (Werner et al. Reference Werner, Dogan, Saß, Mirzazade, Schiefer, Shah, Schulz and Reetz2013). These studies suggest abnormal RSN activity in HD. However, it remains unclear to what extent regional brain atrophy may affect brain activity in manifest HD. This is an important issue since neural activity is bound to brain structure so that brain atrophy can have effects on neural activity in HD (Reading et al. Reference Reading, Dziorny, Peroutka, Schreiber, Gourley, Yallapragada, Rosenblatt, Margolis, Pekar, Pearlson, Aylward, Brandt, Bassett and Ross2004; Wolf et al. Reference Wolf, Vasic, Schonfeldt-Lecuona, Ecker and Landwehrmeyer2009b , Reference Wolf, Gron, Sambataro, Vasic, Wolf, Thomann, Saft, Landwehrmeyer and Orth2011; Unschuld et al. Reference Unschuld, Joel, Liu, Shanahan, Margolis, Biglan, Bassett, Schretlen, Redgrave, van Zijl, Pekar and Ross2012).
The aims of this study were twofold. First, using ICA and a ‘functional network connectivity’ approach (Jafri et al. Reference Jafri, Pearlson, Stevens and Calhoun2008) we investigated changes within RSNs and differences in between-network coupling in patients with early manifest HD compared with healthy individuals. To account for effects of brain atrophy, we integrated high-resolution structural data with the functional data analyses, thus correcting for regional brain atrophy across the whole brain. Second, in HD patients we explored the relationship between RSN connectivity and clinical measures, such as scores of disease burden, motor function or cognition. Given that motor dysfunction and cognitive impairment are inherent to HD, and since prefrontal cortex activity is highly relevant to cognition in HD gene mutation carriers (Wolf & Klöppel, Reference Wolf and Klöppel2013), we focused on neural systems associated with motor processing (i.e. cortical motor and striatal RSNs) and higher-order cognitive function (i.e. lateral and medial prefrontal RSNs). We predicted that regionally abnormal activity within these networks would be associated with motor or cognitive dysfunction in patients. Moreover, given our recent findings of between-network connectivity changes in preclinical HD gene mutation carriers (Wolf et al. Reference Wolf, Sambataro, Vasic, Wolf, Thomann, Saft, Landwehrmeyer and Orth2012), we expected that in symptomatic HD patients between-network coupling would also differ from that of controls.
Method
Participants
We recruited 21 right-handed HD patients with a genetically confirmed CAG repeat expansion (⩾39) in the HTT (huntingtin) gene. All HD patients were motor manifest, i.e. they had unequivocal motor signs of HD and a diagnostic confidence score of 4 on the motor part of the Unified Huntington's Disease Rating Scale (UHDRS; Huntington Study Group, 1996). Disease staging was based on the score on the Total Functional Capacity (TFC) scale of the UHDRS (Huntington Study Group, 1996). The patients' TFC score ranged between 8 and 13 points, corresponding to Shoulson and Fahn disease stages I and II.
Patients were excluded from participation if they met Diagnostic and Statistical Manual of Mental Disorders, fourth edition, text revision (DSM IV-TR; APA, 2000) criteria for a current psychiatric disorder or if they presented with a history of another neurological disease or a history of head trauma. From the initial sample, one patient was excluded because of head motion of >2 mm during fMRI. Of the included 20 patients, seven were unmedicated, five were on anti-dyskinetics, three were on anti-depressants, three were on a combination of anti-dyskinetics and anti-depressants, and two were on memantine. The burden of HD pathology was estimated using a formula based on age and CAG repeat length [(CAGn-35.5) × age] (Penney et al. Reference Penney, Vonsattel, MacDonald, Gusella and Myers1997; Tabrizi et al. Reference Tabrizi, Langbehn, Leavitt, Roos, Durr, Craufurd, Kennard, Hicks, Fox, Scahill, Borowsky, Tobin, Rosas, Johnson, Reilmann, Landwehrmeyer and Stout2009).
A total of 20 healthy controls were matched for age, education, intelligence quotient (IQ), gender and handedness. Participants with neurological or psychiatric diseases were excluded. All participants underwent a neuropsychiatric examination including a semi-structured psychiatric interview to exclude Axis I and Axis II disorders according to DSM-IV-TR, the Hospital Anxiety and Depression Scale (HADS; Zigmond & Snaith, Reference Zigmond and Snaith1983) and the complete UHDRS. The ‘Multiple Choice Vocabulary Test’ (‘Mehrfach Wortschatz Intelligenztest’), version B (MWT-B; Lehrl et al. 1995) was used to estimate individual IQs. All participants were recruited exclusively for rs-fMRI, i.e. no other experimental tasks were involved. None of these patients was included in our previous task-based fMRI study (Wolf et al. Reference Wolf, Vasic, Schonfeldt-Lecuona, Ecker and Landwehrmeyer2009b ). The project was approved by the local ethics committee (Ulm University, Germany). Written informed consent was obtained from all participants following a complete description of the study.
MRI data acquisition
Structural MRI data acquisition
Three-dimensional magnetization-prepared rapid gradient-echo (3D-MPRAGE) data were acquired using a 3 T Magnetom ALLEGRA (Siemens, Germany) head MRI system [echo time (TE) = 3.67 ms, repetition time (TR) = 2200 ms, inversion time = 1200 ms, field of view (FOV) = 282 mm, slice plane = axial, slice thickness = 1.1 mm, number of slices = 208].
rs-fMRI data acquisition
T2*-weighted images were obtained using echo-planar imaging in an axial orientation (TR = 2000 ms, TE = 30 ms, FOV = 192 mm, flip angle 80°, voxel size 3 × 3×3 mm, 33 slices, slice thickness 3 mm, gap 1 mm). Within a session 180 volumes were acquired. MRI scanning was carried out in darkness. Participants were explicitly instructed to relax without falling asleep, to keep their eyes closed, not to think about anything special and move as little as possible. Adherence to these instructions was verified by verbally contacting participants immediately after the resting-state scan, prior to structural data acquisition. None of the patients reported that they had fallen asleep during the rs-fMRI scan.
MRI data analysis
This study employed two distinct neuroimaging modalities to investigate spatial and temporal characteristics of neural activity (see also the data analysis flowchart in online Supplementary Fig. S1). First, we analysed structural data using voxel-based morphometry (VBM) (Ashburner & Friston, Reference Ashburner and Friston2005). Next, we computed a spatial ICA to identify cortical motor, striatal and prefrontal RSNs of interest, which were then compared between patients and controls. For fMRI, spatial differences between the groups were assessed with or without correcting for brain atrophy. Eventually, for fMRI data, between-network analyses were computed taking into account temporal relationships between RSNs.
Structural data analysis
A VBM analysis was computed using a VBM toolbox running within the Statistical Parametric Mapping software package version 8 (SPM8; http://dbm.neuro.uni-jena.de/vbm8/). During data segmentation, each participant's original T1 image was spatially normalized and segmented into grey and white matter and cerebrospinal fluid. This procedure was followed by partial volume estimation (Tohka et al. Reference Tohka, Zijdenbos and Evans2004), data denoising (Manjon et al. Reference Manjon, Coupe, Marti-Bonmati, Collins and Robles2010) and application of Markov random fields (Rajapakse et al. Reference Rajapakse, Giedd and Rapoport1997). The VBM toolbox also included normalization using ‘Diffeomorphic Anatomic Registration through Exponentiated Lie (DARTEL) Algebra’ (Ashburner, Reference Ashburner2007). Modulated normalized grey matter segments were smoothed using an 8 mm full-width at half-maximum (FWHM) Gaussian kernel. To test for grey matter volume (GMV) differences between groups, a t test was computed where age, gender and total intracranial volume (TIV) were included as nuisance variables. An absolute threshold of 0.2 was used to prevent effects located at tissue borders. Differences were assessed using a threshold of p < 0.05, family-wise error corrected. For exploratory purposes, in a second comparison step we lowered the significance threshold to p < 0.001, uncorrected at the voxel level, p < 0.05 corrected for spatial extent.
fMRI data analysis
Data pre-processing was performed with SPM8 (http://www.fil.ion.ucl.ac.uk/spm) and MATLAB 7.3 (MathWorks, USA). Prior to data processing, the first eight volumes of the time-series were discarded. The remaining functional images were corrected for motion artifacts, co-registered to the individual T1 images, spatially normalized using DARTEL and smoothed with a 9 mm FWHM isotropic Gaussian kernel. After calculating the mean relative Euclidean distance between movement parameters derived from the individual realignment files (controls: mean = 0.15, s.d. = 0.09, range = 0.05–0.45; patients: mean = 0.19, s.d. = 0.10, range = 0.06–0.53) a two-sample t test did not reveal significant head movement differences between the groups (p = 0.19). A spatial ICA was then computed on the entire dataset using the Group ICA for fMRI Toolbox (GIFT; http://mialab.mrn.org/software/gift) (Correa et al. Reference Correa, Adali, Yi-Ou and Calhoun2005). To increase the stability of the components, we used the Icasso algorithm (Himberg et al. Reference Himberg, Hyvarinen and Esposito2004) after repeating the ICA estimation 50 times with bootstrapping and permutation. The dimensionality of the functional data was reduced using principal component analysis alternated with data concatenation across subjects, resulting in one aggregate mixing matrix. An ICA decomposition using the Infomax algorithm was used to extract 21 independent components (ICs). The ‘minimum description length’ criteria (Li et al. Reference Li, Adali and Calhoun2007) were used to estimate the order selection. The estimated ICs were used for a back reconstruction into individual ICs using the aggregate mixing matrix created during the dimensionality data reduction steps. The individual ICs consisting of individual spatial independent maps and time courses were spatially sorted using a priori masks, as defined by the Automatic Anatomical Labeling (AAL) atlas (Tzourio-Mazoyer et al. Reference Tzourio-Mazoyer, Landeau, Papathanassiou, Crivello, Etard, Delcroix, Mazoyer and Joliot2002). To identify and select ICs/RSNs of interest for further between-group analyses, three a priori masks were used for spatial sorting. First, we computed a ‘motor mask’ comprising the precentral gyrus, the cingulate cortex and bilateral medial and superior frontal regions. Second, a ‘striatal’ mask was computed comprising the bilateral caudate nucleus and the putamen. Third, we computed a ‘prefrontal mask’, comprising the bilateral superior, inferior, middle and medial frontal gyri and the anterior cingulate cortex. RSNs that were significantly (p < 0.001) spatially correlated with these masks were chosen for second-level analyses. Note that the three masks described above were only used to identify networks of interest. They were not used as ‘seeds’ or to constrain second-level within- and between-group analyses on a certain set of brain regions.
Voxel-wise one-sample t tests against the null hypothesis of zero magnitude were used to calculate within-group maps for each RSN. Between-group comparisons were performed using two-sample t tests covaried for age and gender (see Allen et al. Reference Allen, Erhardt, Damaraju, Gruner, Segall, Silva, Havlicek, Rachakonda, Fries, Kalyanam, Michael, Caprihan, Turner, Eichele, Adelsheim, Bryan, Bustillo, Clark, Feldstein Ewing, Filbey, Ford, Hutchison, Jung, Kiehl, Kodituwakku, Komesu, Mayer, Pearlson, Phillips, Sadek, Stevens, Teuscher, Thoma and Calhoun2011; Unschuld et al. Reference Unschuld, Joel, Liu, Shanahan, Margolis, Biglan, Bassett, Schretlen, Redgrave, van Zijl, Pekar and Ross2012). To fully include those brain regions that were recruited by both groups we masked these analyses with a combined mask, which was created as follows. First we computed one-sample t tests per RSN and each group. Second, binarized t-map masks of the combined effect of each diagnostic group were thresholded at p < 0.05 (uncorrected) and used to explicitly mask the between-group comparisons computed for each RSN. We performed two analysis modalities. First, we investigated RSN connectivity differences between the groups without including measures of brain atrophy (‘standard SPM analyses’). Second, to investigate RSN connectivity differences controlled for regional atrophy, we used the Biological Parametric Mapping (BPM) toolbox (http://www.ansir.wfubmc.edu) (Casanova et al. Reference Casanova, Srikanth, Baer, Laurienti, Burdette, Hayasaka, Flowers, Wood and Maldjian2007). Based on a voxel-wise use of the general linear model, this toolbox allows the incorporation of information obtained by other imaging modalities as regressors in a voxel-wise analysis. Using BPM we computed an analysis of covariance model, where the individual GMV maps (as obtained by the VBM analysis) were included as covariates, together with age, gender and TIV. Prior to BPM, to achieve similar resolution the individual normalized and modulated GMV maps were interpolated to the voxel size of the functional data and smoothed with a 9 mm Gaussian kernel.
A threshold of p < 0.001, uncorrected at the voxel level, p < 0.05 corrected for spatial extent (Worsley et al. Reference Worsley, Cao, Paus, Petrides and Evans1998), was chosen for all second-level between-group comparisons. Stereotaxic coordinates are reported as coordinates of cluster maxima in Montreal Neurological Institute (MNI) space. Anatomical regions emerging from the between-group comparisons were labelled according to Talairach Daemon (TD) labels and the AAL atlas, implemented in the Wake Forest University (WFU) PickAtlas toolbox (http://fmri.wfubmc.edu/software/PickAtlas).
Functional network connectivity
We assessed temporal correlations between RSNs (‘functional network connectivity’) by means of a constrained maximal lag correlation approach (Functional Network Connectivity Toolbox, FNCTB version 2.3, http://mialab.mrn.org/software/fnc/) (Jafri et al. Reference Jafri, Pearlson, Stevens and Calhoun2008). Analyses were performed using the ‘motor’ and the ‘striatal’ RSNs and the three prefrontal RSNs (as described below in more detail). For these analyses, atrophy correction was not performed, since regional atrophy correction using BPM is valid for spatial comparisons within RSNs but not for analyses of temporal features. To assess functional network connectivity, the individual time courses of the RSNs were first interpolated to 50 ms bins to allow detection of sub-TR haemodynamic differences. Subsequently, the temporal correlation for the RSNs was estimated by shifting the lag between the time courses from –4 to 4 s and calculating the maximal correlation value and the corresponding lag value for each participant. After Z-transformation between-group differences of significant maximal correlations and lag values were compared using two-sample t tests (p < 0.05, uncorrected). A significant correlation between the RSNs indicates their temporal dependency whereas the lag values represent the delay between two correlated RSNs averaged across each group (Jafri et al. Reference Jafri, Pearlson, Stevens and Calhoun2008).
Correlations between functional connectivity and clinical measures
Correlation analyses within the HD group were calculated between indices of RSN connectivity and clinical measures (p < 0.05, uncorrected for multiple comparisons). CAG repeat length, UHDRS motor and cognitive scores, duration of illness and disease burden were correlated with mean voxel weights from significant clusters emerging from atrophy-corrected between-group comparisons. Extraction of β parameters was performed using the MarsBar toolbox (Brett et al. Reference Brett, Anton, Valabregue and Poline2002) and then processed off-line using the Statistica software package (version 6.0; StatSoft Inc., USA).
Results
Participants
There were no significant differences between patients and controls in terms of age, gender, education or IQ (Table 1). UHDRS scores were all significantly worse in patients than in controls. No patient showed depressive symptoms (as indicated by mean HADS scores) at the time of study participation. The HADS scores were similar in both groups.
Table 1. Demographics and clinical scores for controls and HD patients
Data are given as mean (standard deviation).
HD, Huntington's disease; IQ, intelligence quotient; n.a., value was not available; CAG, cytosine–adenine–guanine; UHDRS, Unified Huntington's Disease Rating Scale; HADS, Hospital Anxiety and Depression Scale.
MRI results
Structural MRI results
Patients had less GMV than controls mainly in the striatum (Fig. 1). The largest cluster (peak voxel, given in MNI coordinates, at x = 5, y = 6, z = 3, Z = 7.44, k = 10488) comprised the bilateral caudate nucleus, the bilateral putamen, the left insula, and the left inferior frontal gyrus. Other clusters comprised the middle and left superior temporal gyri (x = − 54, y = − 19, z = − 8, Z = 5.56, k = 244), and the left insula (x = − 35, y = − 24, z = 10, Z = 5.44, k = 365). There were no brain regions with increased GMV in HD patients compared with controls.
Fig. 1. Brain regions with decreased grey matter volume in manifest Huntington's disease (HD) patients compared with healthy controls. The results are of the second-level analysis of variance, p < 0.05, family-wise error corrected. The colour bar indicates t values. The figure displays the Statistical Parametric Mapping (SPM) ‘glass brain’ output (bottom, left) together with maps rendered onto the anatomical template implemented in SPM8. Hemispheric laterality is according to neurological convention (left = left).
When the threshold was lowered to p < 0.001, uncorrected for height, widespread cortical and subcortical GMV differences were detected between HD patients and controls: apart from the striatal regions described above, reduced GMV was found in regions of the frontal and cingulate cortex, temporal and occipital areas and in the thalamus (online Supplementary Fig. S2).
fMRI results
Component estimation and RSN selection
From a total of 21 estimated ICs, five networks (RSNs) of interest were identified using a spatial correlation with motor, striatal and prefrontal a priori masks (see fMRI data analysis section). These masks were only used for RSN selection; a priori masks were not used for subsequent within- or between-group analyses.
One RSN showed a significant correlation with the striatal mask (r = 0.369, p < 0.001). This RSN contained the bilateral caudate nucleus, regions of the putamen, the thalamus and regions of the anterior and lateral prefrontal cortices (‘striatal’ RSN, Fig. 2 a and online Supplementary Fig. S3). A second RSN showed a significant correlation with the motor mask (r = 0.314, p < 0.001). This ‘motor’ RSN included cortical primary motor areas, supplementary motor areas (SMAs), superior frontal and cerebellar regions (Fig. 2 a and online Supplementary Fig. S4). Three RSNs showed a significant correlation with the prefrontal mask. These ‘prefrontal’ RSNs included a network comprising anterior and medial prefrontal areas, cingulate and parietal regions, the precuneus and cerebellar areas (‘anterior prefrontal’ RSN, r = 0.443, p < 0.001), a left lateralized fronto-temporo-parietal system (‘left lateral prefrontal’ RSN, r = 0.248, p < 0.001) and a right lateralized fronto-temporo-parietal system (‘right lateral prefrontal’ RSN, r = 0.201, p < 0.001) (Fig. 2 b and online Supplementary Figs S5–S7).
Fig. 2. (a) Patterns of functional connectivity within the motor and striatal resting state networks (RSNs). Common to controls and Huntington's disease (HD) patients, the figure displays independent components (ICs) and their corresponding time courses, as identified by the group independent component analysis (ICA). Blue, Striatal RSN; red, motor RSN; L, left; R, right. (b) Patterns of functional connectivity within the prefrontal RSN. Common to controls and HD patients, the figure displays ICs and their corresponding time courses, as identified by the group ICA. Blue, Left lateral prefrontal RSN; green, anterior prefrontal RSN; red, right lateral prefrontal RSN. The composite IC maps were obtained using the graphical output function provided by the Group ICA for fMRI Toolbox (GIFT) (http://mialab.mrn.org/software/gift/index.html). The colour bars indicate Z values. For illustrative purposes, ICs are thresholded at Z = 1.
Between-group comparisons
Table 2 displays stereotaxic coordinates and Z scores for both outputs in order to facilitate comparisons between analysis modalities. Figs 3 and 4 display between-group differences in RSN connectivity, as revealed by BPM analyses. For comparison purposes, between-group activity maps from both the ‘standard’ SPM and the BPM analyses are displayed in online Supplementary Figs S3–S7.
Fig. 3. Striatal and motor resting state network (RSN) functional connectivity differences between Huntington's disease (HD) patients and controls. Results are derived from Biological Parametric Mapping (BPM) between-group analyses, p < 0.001 (uncorrected at the voxel level), p < 0.05 corrected for spatial extent. ‘Standard’ Statistical Parametric Mapping (SPM) analyses were covaried for age and gender; BPM analyses were covaried for age, gender, grey matter volume and total intracranial volume. (a) Independent component (IC) maps obtained using the graphical output function provided by the Group ICA for fMRI Toolbox (GIFT) (http://mialab.mrn.org/software/gift/index.html). The colour bars indicate Z values. For illustrative purposes, ICs were thresholded at Z = 1. L, Left; R, right. (b) Between-group differences, rendered onto the anatomical template implemented in SPM8. n.s., Non-significant findings.
Fig. 4. Anterior and lateral prefrontal resting state network (RSN) functional connectivity differences between Huntington's disease (HD) patients and controls. Results are derived from Biological Parametric Mapping (BPM) between-group analyses, p < 0.001 (uncorrected at the voxel level), p < 0.05 corrected for spatial extent. ‘Standard’ Statistical Parametric Mapping (SPM) analyses were covaried for age and gender; BPM analyses were covaried for age, gender, grey matter volume and total intracranial volume. (a) Independent component (IC) maps obtained using the graphical output function provided by the Group ICA for fMRI Toolbox (GIFT) (http://mialab.mrn.org/software/gift/index.html). The colour bars indicate Z values. For illustrative purposes, ICs were thresholded at Z = 1. L, Left; R, right. (b) Between-group differences, rendered onto the anatomical template implemented in SPM8. n.s., Non-significant findings.
Table 2. Brain areas showing functional connectivity differences between controls and HD patients a
HD, Huntington's disease; SPM, Statistical Parametric Mapping; BPM, Biological Parametric Mapping; RSN, resting state network; GMV, grey matter volume; TIV, total intracranial volume; UHDRS, Unified Huntington's Disease Rating Scale; CAG, cytosine–adenine–guanine.
a Results of second-level t tests [standard SPM analysis (left), BPM analysis (right)], p < 0.001 uncorrected at the voxel level, p < 0.05 corrected for spatial extent. ‘Standard’ SPM analyses were covaried for age and gender; BPM analyses were covaried for age, gender, GMV and TIV.
b Parentheses indicate that these regions were not confirmed after additionally covarying for medication status and movement.
* Correlation with UHDRS motor score (r = 0.44, p = 0.025).
** Correlation with CAG repeat length (r = –0.67, p = 0.001); correlation with UHDRS cognitive score (r = 0.50, p = 0.013).
*** Correlation with UHDRS motor score (r = 0.43, p = 0.026).
‘Standard’ SPM analysis (i.e. data not covaried for brain atrophy) revealed, in HD patients compared with controls, increased connectivity in the caudate bilaterally within the ‘striatal’ RSN. Within the ‘motor’ RSN, patients had increased connectivity of the SMA. Increased connectivity was found in the left inferior and middle frontal cortices within the ‘anterior prefrontal’ RSN. In patients, decreased connectivity of the precuneus and the right inferior parietal lobule was evident in the ‘anterior prefrontal’ RSN. Within the ‘left lateral prefrontal’ RSN, patients exhibited decreased connectivity of the middle cingulate gyrus, and within the ‘right lateral prefrontal’ RSN, there was decreased connectivity of the left middle temporal gyrus in patients.
Next, we employed BPM analyses to control for brain atrophy. In the ‘striatal’ RSN, connectivity differences in the caudate nucleus were not evident. However, BPM revealed increased connectivity of the left inferior frontal gyrus and decreased connectivity of the bilateral thalamus and the right superior temporal gyrus in patients compared with controls (Fig. 3 and Table 2). Within the ‘motor’ RSN, as with SPM increased connectivity with the SMA was also seen by BPM (see Fig. 3 and Table 2). In the ‘anterior prefrontal’ RSN, BPM analyses revealed increased connectivity of the left middle frontal gyrus and decreased connectivity of the precuneus and the right inferior parietal lobule in patients (Fig. 4 and Table 2). In the ‘left lateral prefrontal’ RSN, connectivity of the anterior cingulate (ACC) and the medial frontal cortex was increased in patients (Fig. 4 and Table 2). In the ‘right lateral prefrontal’ RSN, BPM analyses revealed increased connectivity of the right inferior frontal gyrus and the right paracentral lobule along with decreases of right precentral cortex connectivity in patients (Fig. 4 and Table 2).
To test for potential effects of psychotropic medication, we repeated ‘standard’ SPM and BPM analyses using medication status (defined as the presence or absence of psychotropic medication) and movement indices (mean relative Euclidean distance between the six movement parameters) as additional covariates of no interest. Stereotaxic coordinates and Z scores for both analysis modalities are shown in online Supplementary Table S1. The majority of the between-group differences detected by our previous analyses were found to be significant after additional covariation, albeit with a smaller cluster extent (see Table 2). For ‘standard’ SPM analyses the right inferior parietal cortex and the precuneus within the anterior prefrontal RSN were not confirmed to be significant after further covariation. For BPM analyses the right inferior parietal cortex within the anterior prefrontal RSN and the right precentral gyrus within the right lateral prefrontal RSN were not confirmed to be significant after further covariation (see also Table 2).
Functional network connectivity
Functional network connectivity analyses revealed significant differences in between-RSN coupling between the ‘motor’, the ‘striatal’ and the ‘prefrontal’ RSNs in controls and patients (Fig. 5). The functional coupling, i.e. the significant maximal correlations and the lag values, between the right and left ‘lateral prefrontal’ RSNs and the ‘anterior’ prefrontal and the coupling between the ‘right lateral prefrontal’ and the ‘motor’ RSNs was significantly greater in patients than in controls (see Fig. 5 and online Supplementary Fig. S8).
Fig. 5. Functional network connectivity within and between Huntington's disease (HD) and control groups. The spatial maps and their corresponding arrows indicate significance and directionality of maximal correlations using within- and between-group t tests (p < 0.05, uncorrected). The corresponding correlation graphs and p values are shown in online Supplementary Fig. 7. The numbers associated with their corresponding image refer to the respective independent component numbers, as obtained by independent component analysis. Arrow colours reflect the lag time, as indicated by corresponding bars.
Correlations between functional connectivity and clinical measures
In the patient group, significant correlations between neural activity and clinical measures were found for the SMA within the ‘motor’ RSN, for the left middle frontal gyrus within the ‘anterior prefrontal’ RSN, and for the ACC within the ‘left lateral prefrontal’ RSN. The higher the connectivity of the left SMA (r = 0.44, p = 0.025) or the ACC (r = 0.43, p = 0.026), the higher the UHDRS motor score. Connectivity of the left middle frontal gyrus was negatively correlated with CAG repeat length (r = –0.67, p = 0.001). A positive relationship between connectivity of the left middle frontal gyrus and the UHDRS cognitive score was found (r = 0.50, p = 0.013); correlation plots are shown in online Supplementary Fig. S9.
Discussion
In this study we investigated, in patients with early HD, brain structure and neural activity within and between anatomically and functionally distinct RSNs. Four main findings emerged. (1) In patients, GMV was reduced in the caudate, putamen, left insular cortex, left inferior frontal cortex and left superior temporal gyrus. (2) Functional connectivity was abnormal within distinct RSNs. For some of the RSN nodes abnormal connectivity corresponded to GMV loss while in other regions connectivity differed in HD in the absence of GMV change. (3) In HD patients, functional coupling between the RSNs was increased. (4) Connectivity changes were related to motor, or cognitive, performance: an increase of SMA connectivity within the motor RSN, or ACC connectivity increases in the left lateral prefrontal RSN, were associated with higher motor scores while higher left middle frontal gyrus connectivity within the anterior prefrontal RSN was associated with better cognitive performance.
Striatal atrophy is a neuropathological hallmark of HD even during the pre-manifest stage (Tabrizi et al. Reference Tabrizi, Langbehn, Leavitt, Roos, Durr, Craufurd, Kennard, Hicks, Fox, Scahill, Borowsky, Tobin, Rosas, Johnson, Reilmann, Landwehrmeyer and Stout2009, Reference Tabrizi, Scahill, Durr, Roos, Leavitt, Jones, Landwehrmeyer, Fox, Johnson, Hicks, Kennard, Craufurd, Frost, Langbehn, Reilmann and Stout2011; Wolf et al. Reference Wolf, Thomann, Thomann, Vasic, Wolf, Landwehrmeyer and Orth2013). The pattern of GMV decrease in our study is in good accordance with findings from large-scale multicentre studies and with recent meta-analyses (Dogan et al. Reference Dogan, Eickhoff, Schulz, Shah, Laird, Fox and Reetz2013; Lambrecq et al. Reference Lambrecq, Langbour, Guehl, Bioulac, Burbaud and Rotgea2013). We add to this in that we explored what these structural changes mean for neural activity at rest. This is important because the relationship between behaviour and neural activity may be closer than between brain structure and behaviour. We first performed functional between-group analysis without correcting for structural change. This revealed that, within the striatal RSN, patients had increased functional connectivity in the caudate bilaterally, increased connectivity of the SMA within the motor RSN, and increased connectivity in the left inferior and middle frontal cortices within the anterior prefrontal RSN. Decreased connectivity was limited to the prefrontal RSN. Within the anterior prefrontal RSN the connectivity of the precuneus and the right inferior parietal lobule was reduced in patients, within the left lateral prefrontal RSN connectivity was reduced in the middle cingulate gyrus, and within the right lateral prefrontal RSN connectivity was decreased in the middle temporal gyrus, respectively. Differences in precuneus and right inferior parietal lobule activity, however, were no longer significant after covarying for medication and movement, indicating less robust findings of functional change in these regions.
Without brain structure, neural activity would be inconceivable, so that any relationship between neural activity and behaviour needs to take brain structure into account. We next employed a technique that integrates information from fMRI and structural MRI. BPM confirmed increased connectivity of the SMA to the motor RSN, and the left middle frontal cortices within the anterior prefrontal RSN, indicating that these brain activity differences in early HD are not solely the consequence of a loss of brain structure. Some differences in functional connectivity between early HD patients and controls were not evident when accounting for brain volume changes. After atrophy correction, we could not confirm the initially highly significant finding of bilaterally increased caudate, or middle temporal, middle cingulate, left inferior and middle frontal cortical regions connectivity in patients. This indicates that striatal and extra-striatal functional connectivity changes can occur simply because of regional brain volume loss and thus may not reflect activity changes independent from atrophy. This notion is in line with a recent study using BPM in early HD (Quarantelli et al. Reference Quarantelli, Salvatore, Giorgio, Filla, Cervo, Russo, Cocozza, Massarelli, Brunetti and De Michele2013), which demonstrated that connectivity changes in the caudate nucleus and other cortical regions can be explained by regional atrophy. On the other hand, in patients BPM analyses also revealed areas of aberrant activity that were not apparent without atrophy correction. BPM revealed increased connectivity of the left inferior frontal gyrus and decreased connectivity of the bilateral thalamus and the right superior temporal gyrus in HD patients compared with controls. Taken together these functional connectivity and brain structural assessments revealed that neural activity can change in the same direction as brain volume: with atrophy correction functional connectivity differences between controls and early HD patients no longer exist. However, functional connectivity differences can also ‘survive’ atrophy correction, suggesting that neural activity and brain volume change in the same direction; however, relative to brain volume changes, activity changes are more pronounced. Functional change in HD can also be in the opposite direction, and in this case neural activity changes only emerge when correcting for atrophy. In addition, further correction for medication and movement indicates significant effects of these parameters and brain activity. Although the majority of our initial findings remained stable, in two cortical regions – the right inferior parietal cortex within the anterior prefrontal RSN and the right precentral gyrus within the right lateral prefrontal RSN – findings were no longer significant. As with the ‘standard’ SPM analysis the effects of psychotropic medication status and within-scanner movements on functional connectivity at rest may be relatively small. However, covariation beyond age and gender is recommendable for future rs-fMRI studies in HD gene mutation carriers.
Since we were interested in the clinical relevance of any of these findings we next examined the association between functional connectivity changes when correcting for atrophy with clinical measures. Increased functional connectivity of the SMA within the motor RSN, or the ACC within the left lateral prefrontal RSN, was associated with higher UHDRS motor scores. This suggests a neural signature of abnormal motor function in patients, which is task-independent and may reflect the neural activity underlying the movement disorder of early HD. Previous motor task-based neuroimaging studies revealed, depending on task requirements and imaging modalities, activity changes in the SMA and ACC in manifest HD and in presymptomatic individuals (preHD) (Bartenstein et al. Reference Bartenstein, Weindl, Spiegel, Boecker, Wenzel, Ceballos-Baumann, Minoshima and Conrad1997; Paulsen et al. Reference Paulsen, Zimbelman, Hinton, Langbehn, Leveroni, Benjamin, Reynolds and Rao2004; Gavazzi et al. Reference Gavazzi, Nave, Petralli, Rocca, Guerrini, Tessa, Diciotti, Filippi, Piacentini and Mascalchi2007; Klöppel et al. Reference Klöppel, Draganski, Siebner, Tabrizi, Weiller and Frackowiak2009a ). These changes may reflect a cortical effort to compensate for cortico-subcortical circuit dysfunction and related behavioural deficits (Paulsen et al. Reference Paulsen, Zimbelman, Hinton, Langbehn, Leveroni, Benjamin, Reynolds and Rao2004; Klöppel et al. Reference Klöppel, Draganski, Siebner, Tabrizi, Weiller and Frackowiak2009a ). Projections from the striatum to frontal motor regions are well known (Takada et al. Reference Takada, Tokuno, Hamada, Inase, Ito, Imanishi, Hasegawa, Akazawa, Hatanaka and Nambu2001). Increased activity of the SMA and ACC may thus reflect lower striatal activity (Paulsen et al. Reference Paulsen, Zimbelman, Hinton, Langbehn, Leveroni, Benjamin, Reynolds and Rao2004). In addition, a study of resting-state function in preHD reported, without correcting for atrophy, lower coupling between the caudate nucleus and the SMA (Unschuld et al. Reference Unschuld, Joel, Liu, Shanahan, Margolis, Biglan, Bassett, Schretlen, Redgrave, van Zijl, Pekar and Ross2012). Since a significant relationship was reported between striatal atrophy and functional connectivity measures it is not clear whether functional connectivity differences go beyond those explained by atrophy (Unschuld et al. Reference Unschuld, Joel, Liu, Shanahan, Margolis, Biglan, Bassett, Schretlen, Redgrave, van Zijl, Pekar and Ross2012). In contrast, and similar to our findings, a recent rs-fMRI study in manifest HD emphasized a positive relationship between motor scores and SMA connectivity (Werner et al. Reference Werner, Dogan, Saß, Mirzazade, Schiefer, Shah, Schulz and Reetz2013). Of note, with the present data we found abnormal SMA and ACC connectivity within systems that were spatially and temporally distinct from the striatal RSN. This spatial segregation may reflect pathological processes following independent temporal trajectories, as implied by neuroimaging data (Wolf et al. Reference Wolf, Sambataro, Vasic, Schonfeldt-Lecuona, Ecker and Landwehrmeyer2008, Reference Wolf, Sambataro, Vasic, Wolf, Thomann, Saft, Landwehrmeyer and Orth2012) and recent neuropathological studies (Rub et al. Reference Rub, Hoche, Brunt, Heinsen, Seidel, Del Turco, Paulson, Bohl, von Gall, Vonsattel, Korf and den Dunnen2012). Consistent with this notion, between-network connectivity analyses showed functional coupling between motor and striatal RSNs in both controls and patients, but no between-group differences. This suggests that abnormal cortical network connectivity at rest in manifest HD is not fully explained by abnormal striatal function.
We also identified that higher left middle frontal gyrus connectivity within the anterior prefrontal RSN was associated with better cognitive performance. This is in line with growing evidence for abnormalities within the frontal cortex in HD (Gomez-Anson et al. Reference Gomez-Anson, Alegret, Munoz, Monte, Alayrach, Sanchez, Boada and Tolosa2009; Gray et al. Reference Gray, Egan, Ando, Churchyard, Chua, Stout and Georgiou-Karistianis2013; Wolf & Klöppel, Reference Wolf and Klöppel2013). In a series of tasked-based fMRI studies we have shown that lateral prefrontal cortex function is a critical neural node in individuals who carry the HD gene mutation (Wolf et al. Reference Wolf, Sambataro, Vasic, Schonfeldt-Lecuona, Ecker and Landwehrmeyer2008, Reference Wolf, Vasic, Schonfeldt-Lecuona, Ecker and Landwehrmeyer2009b ). In an independent preHD sample, we also demonstrated that left lateral prefrontal blood flow at rest is abnormal in preHD (Wolf et al. Reference Wolf, Gron, Sambataro, Vasic, Wolf, Thomann, Saft, Landwehrmeyer and Orth2011). We here extend these findings by showing that lateral prefrontal cortex connectivity in manifest HD is also impaired during resting-state conditions. Given the association with CAG repeat length and measures of cognition this is clinically relevant (Hasselbalch et al. Reference Hasselbalch, Oberg, Sorensen, Andersen, Waldemar, Schmidt, Fenger and Paulson1992; Sax et al. Reference Sax, Powsner, Kim, Tilak, Bhatia, Cupples and Myers1996; Weeks et al. Reference Weeks, Ceballos-Baumann, Piccini, Boecker, Harding and Brooks1997; Gomez-Anson et al. Reference Gomez-Anson, Alegret, Munoz, Monte, Alayrach, Sanchez, Boada and Tolosa2009; Wolf et al. Reference Wolf, Vasic, Schonfeldt-Lecuona, Ecker and Landwehrmeyer2009b , Reference Wolf, Gron, Sambataro, Vasic, Wolf, Thomann, Saft, Landwehrmeyer and Orth2011; Gray et al. Reference Gray, Egan, Ando, Churchyard, Chua, Stout and Georgiou-Karistianis2013). Activity increases at rest have also been demonstrated by a recent rs-fMRI study in a mixed preHD and manifest HD sample (Werner et al. Reference Werner, Dogan, Saß, Mirzazade, Schiefer, Shah, Schulz and Reetz2013). However, in the absence of longitudinal data or complementary task-based data the significance of these findings is unclear. Higher cortical activity levels may reflect increased neural effort in patients to achieve similar levels of cognitive competence as controls (Gray et al. Reference Gray, Egan, Ando, Churchyard, Chua, Stout and Georgiou-Karistianis2013). The increase in functional coupling between the prefrontal networks, as revealed by our functional network connectivity analyses, could reflect an additional neural mechanism of maintaining function (Wolf et al. Reference Wolf, Sambataro, Vasic, Wolf, Thomann, Saft, Landwehrmeyer and Orth2012). However, in contrast to our previous findings, we demonstrated increased, not reduced, left lateral prefrontal activity in patients. Longitudinal data from the TRACK-HD study (Tabrizi et al. Reference Tabrizi, Scahill, Durr, Roos, Leavitt, Jones, Landwehrmeyer, Fox, Johnson, Hicks, Kennard, Craufurd, Frost, Langbehn, Reilmann and Stout2011) indicate that despite progressive loss of brain volume clinical measures do not necessarily deteriorate in HD gene mutation carriers. This supports the notion that neural activity suffices to maintain normal behaviour despite the loss of brain volume. Increased neural activity in HD gene mutation carriers could thus reflect ‘neural compensation’ of ongoing brain structural loss even in spatially remote brain regions (Wolf & Klöppel, Reference Wolf and Klöppel2013). Our study was not designed to test the concept of neural compensation, though. For example, UHDRS cognitive scores differed between controls and patients, indicating significant behavioural differences even in the presence of increased prefrontal connectivity. However, our data suggest that rs-fMRI can detect differences between patients and controls that have relevance for behavioural measures. In an appropriately designed study rs-fMRI could therefore complement structural MRI and task-based fMRI.
Limitations of this study include the cross-sectional design and the inclusion of patients receiving psychotropic drugs, which may affect neural activity (McCabe & Mishor, Reference McCabe and Mishor2011; Posner et al. Reference Posner, Hellerstein, Gat, Mechling, Klahr, Wang, McGrath, Stewart and Peterson2013). As in our patient sample, psychotropic medication in manifest HD is frequent, with heterogeneous drug regimens (Priller et al. Reference Priller, Ecker, Landwehrmeyer and Craufurd2008; Orth et al. Reference Orth, Handley, Schwenke, Dunnett, Craufurd, Ho, Wild, Tabrizi and Landwehrmeyer2010). In agreement with our results the effects of psychotropic drugs on resting-state function in manifest HD has been shown to be very limited in a recent study (Quarantelli et al. Reference Quarantelli, Salvatore, Giorgio, Filla, Cervo, Russo, Cocozza, Massarelli, Brunetti and De Michele2013). However, medication and head movements had an effect on some regions such as the parietal and precentral cortices. Thus, we recommend controlling for both variables in future studies. Due to the study design it remains unknown at this stage if, over time, RSN activity and structural loss follow similar trajectories of decline and how progressive volume loss may affect neural activity changes. This information is crucial to fully appreciate the biomarker potential of resting-state activity changes in patients. In our analyses, we focused on RSNs associated with motor processing (cortical motor and striatal RSNs) and higher-order cognition (prefrontal RSN), since the behavioural relevance of these systems is well documented in HD (see Gomez-Anson et al. Reference Gomez-Anson, Alegret, Munoz, Monte, Alayrach, Sanchez, Boada and Tolosa2009; Gray et al. Reference Gray, Egan, Ando, Churchyard, Chua, Stout and Georgiou-Karistianis2013; Wolf & Klöppel, Reference Wolf and Klöppel2013). We are aware that this approach may constrain the amount of information that can be obtained from rs-fMRI data, since abnormal activity has recently been reported in several other RSNs in HD (Werner et al. Reference Werner, Dogan, Saß, Mirzazade, Schiefer, Shah, Schulz and Reetz2013). rs-fMRI is conducted within a poorly controlled experimental environment, and brain activity at rest can also be modulated by unspecific factors, such as vigilance changes (Olbrich et al. Reference Olbrich, Mulert, Karch, Trenner, Leicht, Pogarell and Hegerl2009). While rs-fMRI may offer some advantages over task-based methods, it is more difficult to relate differences in brain function at rest to specific behavioural processes. Thus, when investigating many different RSNs it is desirable to include, for each RSN, behavioural or task-based data relevant for the respective RSN's function. In addition, our between-network (functional network connectivity) analyses were not corrected for multiple comparisons. Thus, our results should be regarded as preliminary and need to be replicated. Finally, we employed one specific method to evaluate brain structure and subsequently to control for atrophy within the functional data, i.e. VBM. VBM has been widely used in HD research (Dogan et al. Reference Dogan, Eickhoff, Schulz, Shah, Laird, Fox and Reetz2013; Lambrecq et al. Reference Lambrecq, Langbour, Guehl, Bioulac, Burbaud and Rotgea2013), but several other validated methods for structural data analysis do exist (see Tabrizi et al. Reference Tabrizi, Scahill, Durr, Roos, Leavitt, Jones, Landwehrmeyer, Fox, Johnson, Hicks, Kennard, Craufurd, Frost, Langbehn, Reilmann and Stout2011, Reference Tabrizi, Reilmann, Roos, Durr, Leavitt, Owen, Jones, Johnson, Craufurd, Hicks, Kennard, Landwehrmeyer, Stout, Borowsky, Scahill, Frost and Langbehn2012; Wolf et al. Reference Wolf, Thomann, Thomann, Vasic, Wolf, Landwehrmeyer and Orth2013). At present, the number of approaches for multimodal neuroimaging data integration is limited. Here, we opted for a voxel-wise, whole-brain method for atrophy correction using VBM-derived measures. It is possible that our atrophy-corrected results may also be constrained by the sensitivity of this approach.
Conclusions
In conclusion, we have shown that early manifest HD is associated with abnormal connectivity within several distinct resting-state networks subserving motor function and cognition. Our data also suggest that measures of brain structure should be taken into account whenever brain activity is the primary measure in manifest HD. The relationship between functional connectivity increases despite atrophy and clinical measures indicates contributions of cortical regions to the expression of motor signs of HD. In addition, the data suggest that lateral prefrontal cortical connectivity increases in patients are associated with better cognitive performance. Therefore, in early HD differences in functional connectivity may not be limited to a specific task but may extend to the resting brain. Information on brain activity provided by rs-fMRI data acquisition could complement task-based protocols in studying HD.
Supplementary material
For supplementary material accompanying this paper visit http://dx.doi.org/10.1017/S0033291714000579.
Acknowledgements
The authors thank all participants and their families for their time and interest in this study, and Kathrin Brändle for excellent technical support. This research received no specific grant from any funding agency, commercial or not-for-profit sectors.The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008.
Declaration of Interest
None.
