Concurrent Cross-Sectional and Longitudinal Analyses of Multivariate White Matter Profiles and Clinical Functioning in Pre-Diagnosis Huntington Disease

Abstract

Background:

Gray matter (GM) atrophy in the striatum and across the brain is a consistently reported feature of the Huntington Disease (HD) prodrome. More recently, widespread prodromal white matter (WM) degradation has also been detected. However, longitudinal WM studies are limited and conflicting, and most analyses comparing WM and clinical functioning have also been cross-sectional.

Objective:

We simultaneously assessed changes in WM and cognitive and motor functioning at various prodromal HD stages.

Methods:

Data from 1,336 (1,047 prodromal, 289 control) PREDICT-HD participants were analyzed (3,700 sessions). MRI images were used to create GM, WM, and cerebrospinal fluid probability maps. Using source-based morphometry, independent component analysis was applied to WM probability maps to extract covarying spatial patterns and their subject profiles. WM profiles were analyzed in two sets of linear mixed model (LMM) analyses: one to compare WM profiles across groups cross-sectionally and longitudinally, and one to concurrently compare WM profiles and clinical variables cross-sectionally and longitudinally within each group.

Results:

Findings illustrate widespread prodromal changes in GM-adjacent-WM, with premotor, supplementary motor, middle frontal and striatal changes early in the prodrome that subsequently extend sub-gyrally with progression. Motor functioning agreed most with WM until the near-onset prodromal stage, when Stroop interference was the best WM indicator. Across groups, Trail-Making Test part A outperformed other cognitive variables in its similarity to WM, particularly cross-sectionally.

Conclusions:

Results suggest that distinct regions coincide with cognitive compared to motor functioning. Furthermore, at different prodromal stages, distinct regions appear to align best with clinical functioning. Thus, the informativeness of clinical measures may vary according to the type of data available (cross-sectional or longitudinal) as well as age and CAG-number.

Keywords

Cognition magnetic resonance imaging movement multivariate analysis prodromal symptoms white matter

INTRODUCTION

Huntington disease (HD) is a progressive, delayed-onset disorder primarily associated with involuntary motion (chorea) and also characterized by cognitive alterations spanning executive functioning (including task-switching, response inhibition, and working memory) [1 –3], odorant recognition [4], and emotional processing [5, 6]. HD-associated disturbances also extend beyond the central nervous system, affecting cardiac, respiratory, and sleep systems.

HD is a rare condition, affecting around 1 in 7,300 individuals in the United States (global prominence is variable) [7]. However, as both a proteinopathy and the most common PolyQ disease (Q is the symbol for glutamine, the amino acid encoded by the CAG codon), it shares key features with disorders that are more difficult to disentangle due to increased genetic complexity. Proteinopathies, including HD, Alzheimer’s disease, and Parkinson’s disease, are distinguished by regionally-selective neuronal death (brain structural alterations are widespread in HD but concentrated in medium spiny striatal neurons) and protein misfolding (manifesting as expanded huntingtin in HD, Lewy bodies in Parkinson’s disease, and β-amyloid plaques in Alzheimer’s disease). These conditions typically demonstrate delayed onset and disproportionately affect the elderly [8].

There is no known cure for HD, though its cause has been known since the nascence of human genome sequencing. In 1993, an abnormally expanded cytosine-adenine-guanine (CAG) repeat at an exon 1 locus of the chromosome 4 HTT gene was identified as causal for HD [9]. The normal range of CAG-expansion at this locus is 17–20 [10]; repeat-numbers of 36 and above confer highly elevated risk of HD, with higher CAG-numbers being associated with earlier onset and more rapid symptom progression [11]. In individuals with abnormally-expanded CAG, brain structural and cognitive differences from non-expanded individuals are measurable a decade or more before HD diagnosis, during a period known as the prodrome. The relationship between CAG and onset makes it possible to index the prodrome using CAG-number and age [12]; many researchers have leveraged this fact to identify the earliest prodromal changes as potential targets for HD interventions. Because HD typically manifests in mid-to-late life, treatments that further delay symptoms could prevent disease manifestation in many individuals.

PREDICT-HD is among the most extensive research groups to undertake characterization of the prodrome; this multi-site study has collected genomic, brain imaging, and clinical (motor and cognitive) data from 33 sites in the United States, Canada, Australia, and Europe [13]. PREDICT-HD has identified many early prodromal indicators, including widespread gray matter (GM) concentration reductions [14], longitudinal motor, cognitive, and volumetric changes (in the putamen, caudate, and nucleus accumbens) [15] as well as clinically-associated subcortical brain volume differences [2 , 17], and resting state functional connectivity alterations [18].

Brain structure in prodromal HD

GM atrophy, particularly in the basal ganglia, is a long-established symptom of the HD prodrome. More recently, prodromal white matter (WM) deterioration around the striatum [19 –25] and across the brain [16 , 26–29] has also been reported. Cerebral WM can offer independent power for predicting diagnosis proximity in prodromal HD [27]. Prodromal WM changes may also precede those observed in GM, although reports have differed.

To date, most prodromal HD WM studies have examined diffusion magnetic resonance imaging (dMRI) measures such as fractional anisotropy (FA), mean diffusivity (MD), radial diffusivity (RD), and axial diffusivity (AD) to compare WM integrity between prodromal and control participants as well as among prodromal groups with varying estimated proximities to diagnosis. This work has generally demonstrated prodromal anisotropy reductions and diffusivity increases in WM adjacent to the striatum [22 , 30–32] as well as in the thalamus [22 , 31–34], corpus callosum [22 , 35], and parietal and occipital areas [22 , 35]. Frontal regions are also common sites of reported prodromal diffusivity differences [22 , 34–36]. In one of the only studies probing GM-adjacent-WM, the effect sizes for atrophy increases were more significant in GM-adjacent-WM than whole WM [22].

Longitudinal findings have been more limited and conflicting. The largest longitudinal diffusion-weighted imaging (DWI) study to date (N = 191 pre-HD and 70 control participants) reported cross-sectional and longitudinal differences that were consistent with degeneration in WM tracts connecting the striatum to premotor, primary motor, and sensory cortices, with the most prominent longitudinal effects observed in putamen-connected tracts [25]. Another two-year longitudinal WM diffusivity study of 64 prodromal participants reported differences in longitudinal diffusivity that were restricted to the superior fronto-occipital fasciculus [33]. An additional two-year study that included 211 prodromal individuals found increased rates of atrophy in cerebral WM that were most pronounced in the frontal lobe [37]. Notably, no changes were observed in lobe-specific or cortical GM volumes, and disease-related whole-brain WM atrophy exceeded striatal atrophy when age-related atrophy was included as a statistical nuisance variable. Other studies, however, did not report any changes over a 12- or 30-month period [26 , 39].

Prodromal WM and clinical functioning

Most prodromal studies linking WM to motor and cognitive functioning have also been cross-sectional, associating volumetric and FA reductions as well as MD, RD, and AD increases with clinical impairment [16 , 41] (Table 1). In the HD prodrome, correlations between WM and clinical measures have indicated widespread regional involvement in cognitive and motor functioning. Few studies have compared WM and clinical change trajectories concurrently. In one such study, motor symptom increases were associated with diffusivity changes in the superior fronto-occipital fasciculus as well as volumetric changes in the corpus callosum, cerebrospinal fluid, and lateral ventricle [33]. In another longitudinal study, the Unified Huntington Disease Rating Scale Total Motor Score (UHDRS-TMS) was the only independent predictor of longitudinal FA decline in both prodromal and symptomatic HD [38].

Table 1

Reported Prodromal Cross-sectional WM-Clinical Associations

		SDMT	SC	SI	SW	TMTA	TMTB	UHDRS-TMS	Source(s)
volumes	Frontal							√	[16]
	Parietal	√							[16]
	Occipital	√						√	[16]
	Temporal	√						√	[16]
tracts	Putamen-prefrontal	RD	RD	RD	RD				[24]
	Corpus callosum-prefrontal	FA, MD, RD				FA, MD, RD	FA, MD, RD	FA, MD, RD	[38, 40]
	Anterior thalamic radiations	MD, RD	MD, RD				MD, RD		[40]
	Inferior fronto-occipital fasciculi	MD, RD	MD, RD	MD, RD	MD, RD	MD, RD	MD, RD		[40]
	Uncinate fasciculi					FA	RD		[40]
	R medial orbitofrontal						FA		[36]
	R pars opercularis						FA, RD		[36]
	R pars triangularis						FA, RD		[36]

Previously reported associations between prodromal WM volumes/tracts and cognitive and motor functioning variables. R, right; SDMT, Symbol Digit Modalities Test; SC, Stroop Color; SW, Stroop Word; SI, Stroop Interference; TMTA, Trail-Making Test A; TMTB, Trail-Making Test B; UHDRS-TMS, Unified Huntington Disease Rating Scale Total Motor Score; RD, Radial Diffusivity; FA, Fractional Anisotropy; MD, Mean Diffusivity. All reported FA differences are reductions, and all other diffusivity differences reflect increases.

The present study

This work complements and extends the current literature regarding brain-structural and clinical changes in the HD prodrome by: investigating co-occurring regional patterns in the brain; examining cross-sectional and longitudinal WM changes; concurrently comparing change trajectories in WM and clinical functioning variables; and analyzing one of the largest prodromal HD brain imaging data sets ever collected.

To accomplish these goals, we extracted brain imaging patterns primarily consisting of GM-adjacent-WM from our population using source-based morphometry (SBM), an independent component analysis (ICA)-based method optimized for brain imaging data [42]. The ICA-derived structural patterns allowed us to examine co-occurring changes across multiple brain regions and reduce the number of required statistical tests (compared to voxel-by-voxel testing). We applied Linear Mixed Model (LMM) analyses to concurrently compare cross-sectional differences and longitudinal trajectories of these WM profiles with those of clinical (cognitive and motor) functioning measures.

Cross-sectionally, we predicted that the prodromal group closest to estimated diagnosis would most drastically differ from controls, whereas the group furthest from onset would present with the most modest differences. We also expected the most robust longitudinal changes in the near-onset group, particularly in motor functioning and in areas underlying diagnosis-associated motor symptoms, such as the striatum. Across groups, we anticipated that frontal changes would be the most similar to cognitive changes, whereas changes in the striatum, thalamus, and cortical areas important for movement (such as motor cortices) would align most with changes in motor functioning. These predictions reflect the broader underlying hypothesis that different brain structural changes accompany early prodromal cognitive changes compared to the delayed motor dysfunction that typically instigates diagnosis.

MATERIALS AND METHODS

Participants

The study data included 1,336 PREDICT-HD participants (1,047 prodromal and 289 control) who took part in at least two separate MRI sessions. A total of 3,700 sessions were analyzed. Participant demographics are presented in Table 2. Prodromal participants were stratified into Low, Medium, or High disease progression levels based on estimated time to diagnosis as indexed by CAG-age-product (CAP) score, a widely used metric for establishing clinical diagnosis likelihood that was developed based on an accelerated failure time model analysis with the entire PREDICT-HD dataset [12]. CAP was calculated using the formula CAP = (age at first MRI scan session)×(CAG– 33.6). Based on this grouping, the participant pool (at baseline) included 288 Low (CAP scores <287.16), 365 Medium (CAP: 287.16–367.12), and 394 High (CAP > 367.12) prodromal participants, with High group designation indicating the highest likelihood of a looming HD diagnosis. Because CAP-score encompasses age and CAG-repeat-number, age differences between CAP-groups were anticipated. The High group was older than the Medium and Low groups at their earliest scans, and the Medium group was older than the Low group. The Low and Medium groups were also younger than the Control group. The High group had fewer years of education than the Control and Medium groups, and the Low group had fewer years of education than the Control group. Significant CAP-group differences in gender were also observed (there were more female participants overall) but were driven by the High vs. Low contrast. These statistics are presented in Supplementary Material 1.

Table 2

Participant Demographics

Group	Total Participants	CAP Score		Total Sessions	Participant Sessions		Baseline Age (years)		Gender		Education (years)
Group	Total Participants	Mean	SD	Total Sessions	Mean	SD	Mean	SD	Male	Female	Mean	SD
Control	289	–	–	875	3.22	1.82	44.44	11.93	100	189	15	2.44
Low CAP	288	235.75	40.17	768	2.81	2.05	33.83	8.72	92	196	14.49	2.46
Medium CAP	365	330.14	22.76	1,009	3.06	1.87	40.68	9.62	128	237	14.71	2.39
High CAP	394	429.03	54.12	1,048	2.86	1.54	44.89	10.40	166	228	14.41	2.61
All	1,336	341.39	87.94	3,700	2.95	1.82	41.26	11.05	486	850	14.64	2.51

Participant demographics, including baseline CAG-age-product (CAP), age, gender, years of education, and the number of sessions.

All PREDICT-HD participants provided written, informed consent and were treated in accordance with protocols approved by each participating institution’s internal review board. Participants underwent genotyping before study enrollment; those with less than 36 HTT CAG-repeats were designated as controls and those with 36 or more as prodromal. Participants with any other central nervous system condition or an unstable medical or psychiatric disorder were excluded from the study.

Imaging data acquisition

High-resolution anatomical MR images were collected at 33 sites using General Electric, Philips, and Siemens scanners with 1.5 T (Tesla) or 3 T field strengths and various acquisition parameters. Each site underwent a calibration process to minimize inherent differences between site and scanner manufacturer capabilities. The intent was to collect images with similar properties at each site under the constraints of each site’s scanner capabilities. T1 images were obtained using three-dimensional (3D) T1-weighted inversion recovery turboflash (MP-RAGE) sequences. The Siemens protocol was designed to be similar to the General Electric scan parameters: GRAPPA factor, 900 ms TI (inversion time), 2,530 ms TR (relaxation time), 3.09 ms TE (excitation time), 256 mm×256 mm field of view (FoV), 10° flip angle, 240 coronal slices (1 mm thickness), 256×128 matrix with 1/4 phase FoV, and 220 Hz/pixel receiver bandwidth. 3 T scanner protocol commonly included a sagittal localizing series followed by an axial 3D volumetric spoiled gradient recalled acquisition in steady state (GRASS) sequence acquisition and the following scan parameters: ∼1 mm×1 mm×1.5 mm voxel size, 0 mm gap, 18 ms TR, 3 ms TE, 24 cm FoV, 20° flip angle, 124 slices (1.5 mm thickness), 256× 192 matrix with 3/4 phase FoV, number of excitations (NEX) = 2.

Imaging data preprocessing

T1-weighted images were processed using a Nipype pipeline optimized for efficient large-scale multicenter data processing [43] using the BRAINSTools suite (http://github.com/BRAINSia/BRAINSTools.git) and the advanced normalization tools (ANTs) package [44]. Images were spatially normalized to an anterior commissure-posterior commissure (AC– PC) and interhemispheric fissure reference orientation, then segmented into tissue types (GM, WM, and cerebrospinal fluid) using an expectation maximization (EM) and a fuzzy k-Nearest Neighbor (KNN) classification incorporating bias-field correction and image registration [45 –47]. The extracted WM probability maps were used in further analyses.

Clinical variables

Seven clinical variables were selected for analyses based on their established clinical reliability and sensitivity to prodromal HD progression [1], as well as data availability (e.g., although speeded tapping and emotion recognition exhibit strong effect sizes, data were available for <1000 participants): Symbol Digit Modalities Test (SDMT), three conditions from the Stroop Color Word Test, Trail-Making Test parts A (TMTA) and B (TMTB), and the Unified Huntington Disease Rating Scale Total Motor Score (UHDRS-TMS). Table 3 includes examples of each clinical measure. The SDMT, an adaptation of the Wechsler Digit Symbol subtest, is an indicator of working memory, complex scanning, and processing speed capabilities [48, 49]; participants are tasked with matching numbers to corresponding symbols using a provided key. The Stroop Color and Word Test includes the color (SC), word (SW), and interference (SI) conditions. The color and word conditions measure basic attention, whereas the interference condition also entails inhibition of a dominant (automatic processing) response [50]. For the color condition, participants identify colors presented on stimulus cards. For the word condition, the task is to read color names that are presented in black ink. For the interference condition, participants identify the ink color in which a color-name is presented (e.g., for the word “blue” printed in green ink, the correct response is “green”). TMTA primarily measures visual attention, whereas its more complex counterpart, TMTB, also measures task-switching [51, 52]. TMTA entails sequentially connecting a series of numbered circles (e.g., 1-2-3-4). For TMTB, consecutive numbers and letters are alternately connected in ascending/alphabetical order (e.g., 1-A-2-B-3-C). The UHDRS-TMS encompasses several indicators of motor functioning, including oculomotor function, bradykinesia, chorea, dystonia, gait, and postural stability [1, 53].

Table 3

Clinical Variables

Examples of each clinical variable analyzed in the present study. For SC, SW, SI, TMTA, and TMTB, correct responses are shown. For SDMT, an example key and uncompleted response sheet are displayed, and two sample questions from the UHDRS-TMS are presented.

Source-based morphometry

ICA, a blind-source separation technique, was applied to processed WM probability maps using SBM via the GIFT Matlab toolbox (http://mialab.mrn.org/software/gift) to extract maximally independent spatial maps and their subject loading parameters [42 , 55]. This method was chosen because it enables the examination of co-occurring brain structural patterns across several brain regions, in addition to requiring far fewer statistical tests than voxel-by-voxel comparisons. In order to extract a spatial map for every participant at every time point, enabling longitudinal examination of structural changes in the ensuing analyses, all available imaging data (3,700 sessions) were included in the SBM analysis. The number of components representing the 3,700 images was estimated to be 45 using a minimum description length (MDL) criterion modified to account for correlated voxels [56]. In a general ICA model, X = AS: X is an observation (i.e., participant-by-variable matrix); S is a statistically independent component matrix (component-by-variable); and A is the loading coefficient or mixing matrix, the representation of each component in the participant or sample (participant-by-component). SBM is an extension of ICA optimized for brain imaging data; each image is converted into a one-dimensional vector (a participant-by-voxel data matrix) [42]. This matrix is then decomposed into a mixing matrix (participant-by-component, where components are the 45 white matter profiles) and a source matrix (component-by-voxel) using the infomax algorithm for spatial ICA. The goal of ICA is to determine the mixing matrix (or rather, its inverse, the unmixing matrix), which represents how much each participant’s brain structure resembles each maximally independent white matter pattern extracted from the population. Rows of the mixing matrix are weights (or loading coefficients) representing how much each of the 45 components/profiles contributes to one participant’s data, and columns represent how strongly a component is represented in each participant. For the source matrix, by contrast, rows indicate the degree to which each voxel is represented in a component and columns denote how much a single voxel contributes to each of the 45 components. For each participant session, this decomposition produces a loading coefficient for each component, and each component is a spatial map.

As quality control, component stability was confirmed using Icasso software [57] and the most representative decomposition was moved forward for analysis [58]. All 45 components were stable, with 41 components having minimum stabilities above 0.95 and only two components with minimum stabilities below 0.90 (components C and A had stabilities of 0.85 and 0.72, respectively).

Linear mixed model analyses

To determine mean baseline (intercept) differences and rates of longitudinal change (slopes) in the SBM WM profiles among CAP and Control groups, we employed LMMs for longitudinal data [59]. LMMs are particularly appropriate for large, complex datasets because they can furnish unbiased parameter estimates even when some data is missing, as long as the data is missing at random (MAR). MAR describes variables with missing datapoints, for which the missing information can be accounted for by complete information at other datapoints. Maximum likelihood estimation (MLE) can be used to provide unbiased parameter estimates of the mean and variance of the MAR variable under assumptions of normality. We employed the MLE method to estimate random effects and random error [59].

For each WM profile/component (outcome variable), two models were fit and compared via a likelihood ratio test to assess the omnibus null hypothesis of no intercept or slope group differences: (1) a full model, inclusive of CAP-group-specific intercepts and slopes; (2) a reduced model that lacked CAP-specific effects, in which each group shared the same slope and intercept. Any component for which the two models were not significantly different was excluded from further analyses and results.

Both LMMs included a time variable representing years of study participation (with 0 denoting study entry), as well as covariates for gender, education (years), and CAP-group. Because the study was longitudinal, random intercepts for each participant were included in the model to account for correlations due to repeated measurements. Because data from several collection sites and scanners were included in the study, random site intercepts were also added to account for correlations due to common scanner or site-specific features. General linear model hypothesis testing was used to compare component intercepts and slopes among groups.

Stacked linear mixed model analyses

To compare longitudinal trajectories of the SBM WM profiles to those of clinical (cognitive and motor) functioning measures, we used what we hereafter refer to as a “stacked” LMM regression analysis, which permitted simultaneous analysis of the brain structural and clinical data. A detailed description of the statistical approach is provided in the supplemental materials of Shaffer et al., 2017 [25]. Concurrent testing of these variables enabled us to address hypotheses involving direct comparisons of clinical functioning and brain structural integrity, both at baseline and over time across groups. For example, we tested the hypothesis that motor functioning in the prodromal group furthest from estimated HD-onset deteriorates at a rate (and in a direction) similar to that of WM in motor-implicated frontal regions (e.g., supplementary and premotor cortices), whereas motor functioning changes in the group closest to onset is more comparable to striatal and subcortical changes.

To allow meaningful comparisons across variables that are scaled differently, outcome vectors (i.e., cognitive and motor variables) were standardized by subtracting the vector mean (across participants and times) and dividing by the standard deviation. Following standardization of the outcome vectors, the repeated measures outcome variables and their respective covariate data were combined into a stacked format to enable concurrent analysis [60]. For example, to test the hypothesis that a supplementary motor WM profile and motor performance are changing at similar rates, vectors containing the repeated measure values for each participant’s supplementary motor profile were combined with vectors containing each participant’s repeated UHDRS-TMS scores. The covariate matrices were likewise concatenated. Like the previously described LMMs, the stacked analysis included fixed effect covariates for time (years of study participation), gender, years of education and CAP-group, and random intercepts by participant and site. The stacked analysis also included random intercepts by ‘domain’, with ‘domain’ referring to the two stacked outcome variables.

Dummy variables were coded for each outcome variable and its covariates, enabling estimation of baseline and longitudinal covariance for outcome variables. The use of these dummy variables facilitates direct comparison of two outcome variables by generating a confidence interval representing the slope difference between two outcome variables. Random effects (site, participant, and domain) were assumed to be joint-normally distributed and have a general covariance matrix. To permit accurate comparisons of outcome variable trajectories, component (WM profile) values were multiplied by – 1 for comparisons with clinical variables for which higher scores indicate poorer performance (TMTB, TMTA, and UHDRS-TMS).

RESULTS

In the combined sample, 22 out of 45 components (∼49%) yielded Chi-square values with associated p-values of <0.001, representing the most robust CAP group effects observed in our cohort. Nine components exhibited no differences between the null and CAP models; these components were excluded from further analyses.

Source-based morphometry

A summary of the 15 SBM components (WM profiles) that exhibited the most prominent effects in our results (based on the number and strength of significant contrasts) is presented in Table 4. An expanded table with all components is available in Supplementary Material 2, and accompanying images are provided in Supplementary Material 3. A glossary of regions within components is provided in Supplementary Material 4.

Table 4

SBM Component Regions

Comp.	Regions	Max. Stat.	Coordinates (x, y, z)
A	Angular Gyrus/BA39, Middle Temporal Gyrus, Superior Temporal Gyrus, Middle Occipital	– 9.0	– 41, – 62, 27
B	Angular Gyrus/BA39, Inferior Parietal Lobule, Precuneus, Middle Occipital, Superior Parietal Lobule/BA7, Superior Occipital, BA19, Inferior Parietal, Supramarginal/BA40	– 18.2	32, – 60, 39
C	Angular Gyrus/BA39, Supramarginal/BA40, Superior Temporal Gyrus, Inferior Parietal Lobule, Middle Temporal Gyrus	16.5	39, – 55, 31
D	Superior Occipital^L, Cuneus, BA19, Superior Parietal Lobule/BA7, BA18, Precuneus, BA31	– 4.6	– 18, – 85, 29
E	Corpus Callosum, Limbic Lobe, Lateral Ventricle, Cingulate Gyrus/BA23, Posterior Cingulate	13.1	– 1, – 19, 24
F	Supplementary Motor Area/BA6, Medial Frontal Gyrus, Superior Frontal Gyrus, BA8	– 11.7	10, – 7, 64
G	Cerebellum Anterior Lobe, Culmen, Vermis 4,5, Cerebellum 4,5, Cerebellum Posterior Lobe, Declive, Vermis 3, Cerebellar Lingual, Midbrain, Brainstem^R, Cerebellum 3, Vermis 6	– 9.5	– 1, – 56, – 14
H	Cuneus, Superior Occipital, Middle Occipital Gyrus, BA18, BA19, Temporal Lobe	– 10.3	– 24, – 86, 21
I	Precuneus, Superior Parietal Lobule/BA7, Superior Parietal, Superior Occipital, Cuneus	– 15.8	17, – 67, 46
J	Postcentral Gyrus, Inferior Parietal, Inferior Parietal Lobule, Supramarginal/BA40, Superior Parietal, Paracentral Lobule, Superior Parietal Lobule/BA7, BA5, Precuneus	23.6	30, – 39, 41
K	Precentral^L, Middle Frontal Gyrus, Inferior Frontal Gyrus, DLPFC/BA9, Premotor/Supplementary Motor/BA6, Frontal Inferior Operculum	– 25.3	– 36, 7, 38
L	Postcentral Gyrus/BA2/BA3, Inferior Parietal, Inferior Parietal Lobule, Supramarginal/BA40	17.7	– 30, – 36, 45
M	Middle Temporal/BA21, Inferior Temporal	9.5	– 55, – 18, – 12
N	Superior Temporal Gyrus/BA22, Middle Temporal^L, Angular Gyrus/BA39, Middle Occipital	24.1	– 42, – 58, 19
O	Precuneus, Superior Parietal Lobule/BA7, Angular Gyrus, Superior Parietal, Inferior Parietal, Inferior Parietal Lobule, Superior Occipital	23.9	26, – 56, 45

Regions within SBM components displaying the most robust significant effects, including maximum z-scores (Max. Stat. column) and coordinates of the most strongly represented regions. Components were ordered alphabetically based on the number and strength of significant contrasts across all analyses (i.e., component A displayed the maximum observed number of significant results). Primary contributing regions corresponding to maximum z-scores appear in bold, and regions in each component are listed in order of largest-to-smallest voxel cluster size in the most prominent cluster in the component. In cases where regional representation is hemisphere-specific, this is denoted with a superscript L (left) or R (right).

Baseline (intercept) differences in WM profiles

The t- and p-values for components with three or more significant pairwise CAP-group comparisons are presented in Table 5, and an expanded table including all 36 analyzed components is provided in Supplementary Material 5.

Table 5

Baseline (intercept) pairwise group comparisons for components with 3 + significant contrasts

Component	Low vs. Control	Medium vs. Control	High vs. Control	Medium vs. Low	High vs. Low	High vs. Medium
A	3.68/2×10^–4	0.56/0.57	– 1.25/0.21	– 3.36/8×10^–4	– 5.24/<1×10^–4	– 1.96/0.05
B	– 0.63/0.53	1.81/0.07	2.08/0.04	2.52/0.01	2.80/0.01	0.28/0.78
C	– 3.36/1×10^–3	– 2.96/3×10^–3	– 3.73/2×10^–4	0.57/0.57	– 0.18/0.86	– 0.80/0.42
D	3.04/2×10^–3	– 0.05/0.96	– 1.40/0.16	– 3.29/1×10^–3	– 4.69/<1×10^–4	– 1.46/0.14
E	2.59/0.01	– 1.15/0.25	– 1.91/0.06	– 3.93/1×10^–4	– 4.73/<1×10^–4	– 0.82/0.41
I	– 3.19/1×10^–3	– 0.04/0.97	1.72/0.09	3.36/8×10^–4	5.17/<1×10^–4	1.90/0.06
K	– 0.06/0.95	– 0.72/0.47	– 2.74/0.01	– 0.66/0.51	– 2.72/0.01	– 2.18/0.03
N	3.96/1×10^–4	3.69/2×10^–4	7.56/<1×10^–4	– 0.47/0.64	3.43/6×10^–4	4.16/<1×10^–4
O	3.94/1×10^–4	2.59/0.01	5.42/<1×10^–4	– 1.58/0.11	1.27/0.20	3.05/2×10^–3
Q	– 2.34/0.02	1.57/0.12	9.20/<1×10^–4	4.09/<1×10^–4	11.89/<1×10^–4	8.23/<1×10^–4
T	– 0.96/0.34	2.55/0.01	8.76/<1×10^–4	3.62/3×10^–4	9.95/<1×10^–4	6.70/<1×10^–4
X	– 3.37/1×10^–3	– 2.35/0.02	– 1.96/0.05	1.21/0.23	1.65/0.10	0.45/0.65
AA	1.14/0.25	3.87/1×10^–4	5.21/<1×10^–4	2.72/7×10^–3	4.07/<1×10^–4	1.42/0.16
AB	2.03/0.04	3.01/3×10^–3	4.80/<1×10^–4	0.90/0.37	2.71/7×10^–3	1.91/0.06
AD	– 2.10/0.04	– 0.02/0.10	1.20/0.23	2.22/0.03	3.50/1×10^–3	1.32/0.19
AE	– 0.27/0.79	– 0.52/0.60	– 2.71/0.01	– 0.25/0.81	– 2.47/0.01	– 2.36/0.02
AG	1.27/0.20	2.14/0.03	7.26/<1×10^–4	0.82/0.41	6.01/<1×10^–4	5.52/<1×10^–4

Baseline (intercept) pairwise group comparisons. Results (t-value/p-value) for components yielding three or more significant pairwise contrasts are displayed. Bold denotes significant contrasts, and the most robust effects (p < 1×10^–4) are also underlined.

Cross-sectionally, components most sensitive to differences between Controls and all three CAP groups (i.e., those significant in all three Control group contrasts: X, N, O, AB, and C) were most commonly parietal. Areas of overlap among these components included angular gyrus (in three of the five components), superior temporal gyrus, superior and middle occipital, and inferior parietal lobule. Component X also included premotor, primary motor, and supplementary motor cortices.

Three components (T, Q, and N) exhibited the most robust intercept effects across prodromal and control group comparisons, with significant differences in five out of six contrasts (no component was significant in every intercept contrast). These components had few overlapping regions; component T most strongly represented cerebellar posterior lobe (along with cerebellum crus 1 and uvula), component Q sub-gyral inferior parietal WM, and component N superior and middle temporal WM (with weaker parietal and occipital contributions). These three components are superimposed in Fig. 1. Although the non-significant contrast was different among the three components (Control vs. Low for component T, Control vs. Medium for component Q, Low vs. Medium for component N), it reassuringly never included the High group (i.e., the High group was always different from the other groups for these components).

Fig.1

Components with the most robust cross-sectional group differences: superimposition of the three components with the strongest cross-sectional (intercept) effects across groups (components T, Q, and N, thresholded at p = 0.05). These components were significant in all but one of the possible comparisons, highlighting differences across prodromal groups. Primary regions represented in each component are listed along with corresponding Montreal Neurological Institute (MNI) coordinates (x, y, z).

The most common significant baseline differences were observed between the High and Low groups (16 significant components, four at p < 1×10^–4), the High and Control groups (16 significant components, seven at p < 1×10^–4), and the Low and Control groups (15 significant components, minimum p = 1×10^–4). Significant differences from the Medium group were slightly less common, observed in 11 components for the Medium vs. Control (minimum p = 1×10^–4) and Medium vs. Low contrasts (one component significant at p < 1×10^–4) and 12 components in the Medium vs. High contrast (four components significant at p < 1×10^–4).

Longitudinal (slope) differences in WM profiles

Component t- and p-values for pairwise CAP-group slope comparisons are presented in Table 6, and an expanded table is provided in Supplementary Material 6. The rate of change in parietal component Q was the most distinct among groups, exhibiting significant differences in every group contrast Table 6. Component Q was also strongly significant, yielding p < 1×10^–4 in every group contrast except Control >Low. Other components with the most robust rate of change differences among groups were frontal and sub-cortical, including F (premotor, supplementary motor, medial frontal), H (cuneus, middle occipital), T (cerebellum posterior lobe, cerebellum crus 1, uvula), AA (left middle frontal), and AF (thalamus, ventral anterior nucleus), which were each significant in all but one group contrast.

Table 6

Longitudinal (slope) pairwise group comparisons for components with 3 + significant contrasts

Component	Low vs. Control	Medium vs. Control	High vs. Control	Medium vs. Low	High vs. Low	High vs. Medium
A	– 0.13/0.90	2.92/4×10^–3	3.73/2×10^–4	3.10/2×10^–3	3.89/1×10^–4	0.93/0.35
B	0.65/0.52	0.35/0.73	2.96/3×10^–3	– 0.36/0.72	2.29/0.02	2.87/4×10^–3
C	2.96/3×10^–3	1.69/0.09	– 0.36/0.72	– 1.51/0.13	– 3.52/4×10^–4	– 2.23/0.03
D	– 0.69/0.49	2.36/0.02	1.50/0.13	3.16/2×10^–3	2.26/0.02	– 0.92/0.36
F	– 1.82/0.07	2.04/ 0.04	3.99/1×10^–4	4.02/1×10^–4	5.93/<1×10^–4	2.17/0.03
G	2.83/0.01	0.25/0.80	2.94/3×10^–3	– 2.82/0.01	– 0.07/0.95	2.95/3×10^–3
H	0.40/0.69	– 4.71/<1×10^–4	– 1.92/0.05	– 5.19/<1×10^–4	– 2.35/0.02	2.99/3×10^–3
I	2.97/3×10^–3	0.68/0.50	– 0.67/0.50	– 2.55/0.01	– 3.87/1×10^–4	– 1.48/0.14
J	0.98/0.33	1.52/0.13	3.93/1×10^–4	0.46/0.64	2.89/4×10^–3	2.66/0.01
K	– 2.69/0.01	– 2.39/0.02	– 2.36/0.02	0.50/0.62	0.49/0.63	– 0.003/1.0
L	– 1.17/0.24	– 0.04/0.97	2.10/0.04	1.23/0.22	3.36/1×10^–3	2.35/0.02
M	– 0.68/0.50	2.20/0.03	1.44/0.15	2.95/3×10^–3	2.17/0.03	– 0.80/0.42
O	– 0.36/0.72	1.97/0.05	– 0.17/0.86	2.38/0.02	0.21/0.83	– 2.32/0.02
P	– 2.51/0.01	1.81/0.07	0.62/0.54	4.55/<1×10^–4	3.30/1×10^–3	– 1.28/0.20
Q	3.16/2×10^–3	8.86/<1×10^–4	13.43/<1×10^–4	5.50/<1×10^–4	10.11/<1×10^–4	5.14/<1×10^–4
S	– 3.12/2×10^–3	– 2.47/0.01	– 2.78/0.01	0.89/0.37	0.53/0.60	– 0.38/0.71
T	2.77/0.01	3.83/1×10^–4	6.70/<1×10^–4	0.87/0.38	3.81/1×10^–4	3.21/1×10^–3
W	0.18/0.86	– 0.13/0.90	2.78/5×10^–3	– 0.32/0.75	2.61/9×10^–3	3.19/1×10^–3
X	0.58/0.57	0.76/0.45	– 2.4/0.02	0.14/0.89	– 3.03/2×10^–3	– 3.46/1×10^–3
Y	1.57/0.12	3.80/1×10^–4	2.61/0.01	2.17/0.03	0.96/0.34	– 1.27/0.21
AA	– 2.45/0.01	– 5.53/<1×10^–4	– 2.69/7×10^–3	– 2.96/3×10^–3	– 0.10/0.92	3.04/2×10^–3
AC	3.91/1×10^–4	5.05/<1×10^–4	3.48/5×10^–4	0.85/0.40	– 0.67/0.50	– 1.65/0.10
AF	1.35/0.18	3.86/1×10^–4	7.49/<1×10^–4	2.44/0.01	6.12/<1×10^–4	4.05/1×10^–4
AG	0.97/0.33	5.38/<1×10^–4	3.97/1×10^–4	4.42/<1×10^–4	2.98/3×10^–3	– 1.48/0.14
AI	– 1.14/0.25	– 1.07/0.29	3.04/2×10^–3	0.16/0.87	4.30/<1×10^–4	4.50/<1×10^–4
AJ	– 2.08/0.04	– 1.75/0.08	1.15/0.25	0.49/0.63	3.42/1×10^–3	3.17/2×10^–3

Longitudinal (slope) pairwise group comparisons for components with three or more significant contrasts (t-value/p-value) in the unstacked LMM analyses. Bold denotes significant contrasts, and the most robust effects (p < 1×10^–4) are also underlined.

Like the baseline comparisons, the most common significant slope differences were seen between the High and Low groups (22 significant components, four at p < 1×10^–4) and the High and Control groups (21 significant components, three at p < 1×10^–4). The next most frequently observed rate of change differences were between the Low and Medium groups (17 significant components, four at p < 1×10^–4), the Medium and Control groups (16 significant components, five at p < 1×10^–4), and the Low and Control groups (11 significant components, minimum p = 1×10^–4).

Baseline (intercept) differences between WM profiles and clinical variables

For the LMM analyses comparing WM profiles at baseline and over time among groups, we were most interested in significant group differences. The stacked analyses, by contrast, compared baseline values (intercepts) and change trajectories (slopes) between clinical functioning variables and WM profiles; consequently, here we were most interested in non-significant differences, as they indicate the most similarity between clinical variables and WM profiles. Ergo, in Table 7, we summarize the results by listing the components yielding the highest p-values in comparisons with clinical variables. Full results from the clinical test comparisons, including p-values for every contrast and group, are provided in Supplementary Materials 7–13. Across the clinical variables, WM aligned least with cognitive and motor functioning in the High and Medium groups (in that order) and most in the Low and Control groups. The UHDRS-TMS and TMTA scores were most similar to baseline WM, with the UHDRS-TMS being the only clinical variable that did not differ from at least one WM component at baseline (Table 7).

Table 7

WM components most similar to clinical variables at baseline

Clinical variable	Max. # of NS contrasts	Control	Low	Medium	High	Top regions
UHDRS-TMS	4	H, I	H, I	H, I	H, I	precuneus, cuneus, middle occipital
SDMT	1	–	A	–	–	gyrus, middle temporal
SC	1	C	A	–	–	gyrus, supramarginal gyrus, middle temporal
SI	2	C	A, C, D, E, H	–	–	gyrus, supramarginal gyrus, superior occipital, cuneus, middle occipital, middle temporal, corpus callosum
SW	1	–	A	–	–	gyrus, middle temporal
TMTA	3	A, C, F, H, I	A, B, C, D, E, F, G, H	A, C, H	–	motor, premotor, medial frontal, precuneus, angular gyrus, supramarginal gyrus, cuneus, middle occipital, superior occipital, middle temporal, corpus callosum, cerebellum anterior lobe, culmen, vermis 4,5
TMTB	1	C	A	–	–	gyrus, supramarginal gyrus, middle temporal

The maximum number of non-significant (NS) contrasts (out of four) observed in at least one component is provided for each test along with corresponding component identities and the participant groups in which these associations were observed (e.g., components H and I were NS different from UHDRS-TMS at baseline in all four groups). Regions most contributing to these effects are listed in the far-right column and color-coded according to location (frontal = green, temporal = blue, parietal = purple, occipital = red, sub-gyral/other = black).

Longitudinal (slope) differences between WM profiles and clinical variables

Generally, the similarities between WM profiles and clinical functioning variables were much more apparent longitudinally than cross-sectionally (i.e., similarities between rates of change in WM profiles and clinical functioning variables were more readily detectable than similarities at baseline); the sole exception to this pattern of findings was the UHDRS-TMS. In Table 8 we list regions within components that were most similar to clinical variables. In Table 9, we display significant contrasts (indicating dissimilarity) as well as groups aligning most and least with each clinical variable in the study. These results are visualized in Fig. 2 and Table 10, and p-values for every contrast and group are provided in Supplementary Materials 7–13. Supplementary Material 14 contains superimposed clinical variable and component plots.∥

Fig.2

WM Changes and Clinical Functioning across the Prodrome: simplified schematic depicting areas in which rates of change were most similar to changing motor (UHDRS-TMS) and cognitive functioning in each CAP group (yellow = Low, orange = Medium, red = High). UHDRS-TMS, Unified Huntington Disease Rating Scale – Total Motor Score; TMTA, Trail-Making Test part A; TMTB, Trail-Making Test part B; SDMT, Symbol Digit Modalities Test; SW, Stroop word condition; SC, Stroop color condition; SI, Stroop interference condition.

Table 8

WM profiles and clinical variables with similar change trajectories

Clinical variable	Max. # NS contrasts	Control	Low	Medium	High	Top regions
UHDRS-TMS	2	G	D, G, N, AH	D, N, AH	-	occipital, cuneus, superior temporal, middle temporal, cerebellum anterior lobe, culmen, vermis 4,5
SDMT	2	J	A, D, E, F, I, J, P	A, D, E, F, I, L, P	L	motor, premotor, medial frontal, precentral, precuneus, postcentral, angular gyrus, superior occipital, cuneus, middle temporal, corpus callosum
SC	2	G	A, D, E, F, M, P, R	A, D, E, F, L, M, P, R, W	L, W	supplementary motor, premotor, medial frontal, precentral, angular gyrus, postcentral, superior occipital, cuneus, BA18, middle occipital, middle temporal, cerebellum anterior lobe, culmen, vermis 4,5, corpus callosum
SI	2	J	J	B, V, AE	B, V, AE	angular gyrus, inferior parietal, superior parietal lobule, calcarine, cuneus, BA18
SW	3	A	A, D, E, K, L, R	A, D, E, K, L, R	–	middle frontal, DLPFC, angular gyrus, postcentral, superior occipital, cuneus, BA18, middle occipital, middle temporal, corpus callosum
TMTA	2	J	A, B, E, F, J, K, M, P, U	A, B, E, F, K, M, P, U	–	supplementary motor, premotor, medial frontal, precentral, middle frontal, DLPFC, postcentral, angular gyrus, middle temporal, corpus callosum, putamen
TMTB	3	B, G	B, C, G, H, U, V	B, C, H, U, V	–	gyrus, inferior parietal, superior parietal lobule, cuneus, middle occipital, cerebellum anterior lobe, culmen, vermis 4,5, putamen

The maximum number of non-significant contrasts (out of four) observed in at least one component is listed for each test, and components with two or more NS contrasts for a given clinical measure are listed for each group. Regions most contributing to these effects are displayed in the far-right column and color-coded according to location (frontal = green, temporal = blue, parietal = purple, occipital = red, sub-gyral/other = black).

Table 9

Significant component - clinical variable trajectory differences

Significant associations	Control	Low	Medium	High	Most alignment	Least alignment
UHDRS-TMS	34/22	18/9	22/14	36/34	Low	High
SDMT	35/33	26/17	25/14	29/15	Medium	Control
SC	35/30	26/16	23/15	30/17	Medium	Control
SI	35/31	35/14	28/15	19/10	High	Control
SW	32/21	24/13	26/13	32/22	Low	High
TMTA	35/25	24/15	25/12	27/13	Low	Control
TMTB	35/33	21/12	30/16	30/17	Low	Control

Significant differences between component and clinical variable trajectories, presented as: (# of significant differences)/(# significant at p < 1×10^–4). Bold denotes the most similarity with a given group (e.g., in the High group, WM changes were most similar to changes in Stroop Interference performance), as significant differences indicate less similarity. The two rightmost columns list groups showing the most and least alignment with each clinical variable in the study (e.g., change in UHDRS-TMS was most similar to WM change in the High group).

Summary: Prodromal WM and clinical functioning

There were overlaps and differences among WM regions that most readily changed with prodromal progression (unstacked analyses) and those most implicated in cognitive and motor functioning across the prodrome (stacked analyses). Two components (A and B) stood out in both sets of analyses (Fig. 3). One of these (A) is a mixture of parietal, temporal (especially middle) and to a lesser extent, occipital areas; the other (B) is mostly parietal (superior and inferior) and occipital (superior and middle). Both of these components have strong representation from the angular gyrus and displayed prominent differences and changes across the prodrome as well as similarity to clinical changes.

Fig.3

Components with Robust Effects Across Analyses: two WM profiles containing angular gyrus exhibited robust baseline and longitudinal group differences in addition to reliably aligning well with clinical functioning measures.

DISCUSSION

This study used T1-weighted images and tissue-specific probability maps to investigate WM morphology. It should be noted that the findings thus do not reflect WM integrity to the same extent that DWI studies do. Many of our findings were in GM-adjacent-WM, and changes to GM and CSF may influence the morphology of adjacent WM. Nonetheless, our findings aligned quite well with those of previous DWI studies as well as work examining GM-adjacent-WM in this population. Like previous studies, our analyses detected widespread WM differences inclusive of regions in every lobe of the brain. The anticipated Low-to-High prodromal disease gradient of WM degradation was observed both cross-sectionally and longitudinally; cross-sectional and longitudinal differences were most commonly detected in comparisons with the High group.

Hemispheric asymmetry

Although WM changes were mostly bilateral, the left hemisphere seemed overall more affected, mainly by the late prodrome. Few prior studies have directly addressed brain asymmetry in HD outside of the striatum, where left-hemispheric asymmetry is often reported [61, 62]. One volumetric meta-analysis reported converging evidence for left-biased atrophy [61] and an additional study reported bilateral GM atrophy that became more left-lateralized with prodromal progression [62]. In our previous cross-sectional GM analysis, hemispheric effects were more mixed [14], indicating that WM atrophy may proceed along a more uniformly left-hemispheric trajectory than GM atrophy; this may in part reflect connectivity increases that have been observed in right hemispheric regions (such as the dorsolateral prefrontal cortex, or DLPFC), particularly in the late prodrome, which have been proposed as possible compensatory mechanisms [63].

Striatal changes

Regarding the striatum, we observed the most prominent striatal WM changes in the putamen, consistent with reportedly greater WM connectivity disruption within the putamen compared to the caudate [25]. In this study, striatal WM signal came mostly from two putamen-driven components; one contained the frontal lobe, cingulate cortex, and caudate, and the other co-localized with temporal regions (including the amygdala and parahippocampal gyrus) and lacked caudate contributions. The former, fronto-striatal WM component exhibited group differences as early as the Control vs. Low contrast, but the latter, temporo-limbic striatal WM component (which lacked caudate) did not exhibit significant differences until the High prodromal phase. These results suggest that early WM changes in the putamen may be more robust than those in the caudate and occur in conjunction with changes in other functionally-related areas. Although the caudate is generally thought to be affected before the putamen in HD [64], greater atrophy in the putamen relative to the caudate (50.1% and 27.7% reductions compared to controls, respectively) has also been reported [65], and putamen volumes have been identified as more reliable indicators of disease progression than caudate volumes [1, 66].

Motor cortices

The prominence of supplementary motor and premotor cortices in our results is also in keeping with previous findings from our group and others. In two previous PREDICT-HD studies, a similar motor cortical profile in GM exhibited robust group differences [14], alignment with clinical functioning, and an association with a genetic profile related to brain-derived neurotropic factor [67] (Fig. 4). Furthermore, in the most extensive longitudinal study of prodromal WM tract changes to date, these regions also exhibited the most robust effects [25].

Fig.4

Supplementary Motor Components in WM and GM: components that maximally represented supplementary motor cortex in three separate PREDICT-HD studies. A) GM source-based morphometry (SBM) component, from [14]. B) WM SBM component, from the present study. C) GM parallel independent component analysis (pICA) component, from [67]. Images thresholded at p < 0.05, with cross-hairs placed at the global maxima. Coordinates and Z-scores corresponding to the global maxima are shown to the right of each component.

Table 10

WM changes most coinciding with clinical functioning changes

Rows are clinical variables and columns are components thresholded to show effects (z-scores) of p = 0.05 or less. These components correspond to the plots in Fig. 2 (solid lines).

These converging findings provide compelling evidence that motor cortical regions change across the prodrome and likely play a vital role in both early and late prodromal symptoms. In the present study, this supplementary motor profile was also unique in its relevance to both cognitive and motor functioning, which otherwise was evident only in parietal regions. The prominence of inferior frontal regions in our results is also in agreement with their known roles in planning and coordinating complex movements (supplementary motor) and selecting and timing motor sequences (premotor cortex), and aligns with previous studies implicating these regions in HD pathology. For example, inferior frontal (but not striatal) cortical thinning distinguished HD patients with more prominent bradykinesia, rigidity, and dystonia from those with more prominent chorea, indicating that cortical thinning may underlie clinical heterogeneity in HD that has previously been attributed to the striatum [68].

Default mode network

The robust group differences, changes, and alignment with clinical functioning variables that we observed in angular gyrus and precuneus are also in keeping with previous findings, and may reflect an early disruption of the default mode network DMN). The striatum is structurally and functionally connected to DMN hubs, including the precuneus, posterior cingulate cortex, anterior cingulate, inferior parietal and DLPFC. DMN disturbances have been reported in prodromal and diagnosed HD during rest as well as task performance, including increased DMN connectivity that was evident before volumetric loss [69], intrinsic functional connectivity increases [70], and greater M1 connectivity with a posterior cingulate DMN hub (which was associated with cognitive and motor dysfunction) [71]. Although the angular gyrus is not typically reported as a primary region of change in prodromal HD, one study examining post-mortem brain tissue from HD patients found a 55% reduction in angular gyrus pyramidal neurons, and the authors suggested that HD may disproportionately affect posterior cortical regions [72].

WM throughout the HD prodrome

Regarding comparisons with the Control group, frontal components that were most significant longitudinally were least likely to show noticeable cross-sectional effects (e.g., components AC and K), and conversely other components like striatal component S and parietal component Q demonstrated significant intercept and slope effects. Thus, the comparison with the Control group suggests that prodromal differences from Controls in striatal and parietal WM may be more detectible in cross-sectional comparisons than some frontal differences, even though changes in all these areas occur throughout the prodrome.

The suggested patterns of most robust change among the intercept (cross-sectional) and slope (longitudinal) results are similar for the Medium and High groups but distinct for the early prodrome. In other words, cross-sectional and longitudinal findings highlighted many of the same regions in the Medium and High groups, whereas areas most implicated in the Low group differed in the intercept compared to slope results. All results highlighted inferior parietal component Q as strongly representative of changes in the middle and late prodrome. However, the most prominent early (Control vs. Low) cross-sectional differences manifested in temporal and superior parietal regions, whereas the longitudinal results highlighted many early frontal effects, with the most substantial early changes observed in premotor, supplementary motor and middle frontal component AC (in addition to commonly reported striatal and sub-gyral areas in component S). Across groups, cross-sectional effects were weaker overall, but it is comforting for the sake of study feasibility everywhere that similar interpretations become apparent across cross-sectional and longitudinal analyses, especially at later prodromal phases, when intercept and slope results shared more regional overlap.

Taken together, these results suggest robust early prodromal changes in premotor, supplementary motor and middle frontal gyral GM-adjacent-WM, in agreement with previous reports of robust changes in these regions. These findings illustrate a widespread prodromal pattern of WM change that becomes increasingly sub-gyral with prodromal progression (with steeper changes in the cingulum, thalamus, and other sub-gyral and striatal regions). Cerebellum component T, inferior parietal component Q, and frontal component AA (inclusive of premotor and supplementary motor cortices and frontal eye fields) were highly and commonly significant in both intercept and slope contrasts, hinting that these regions may show the most robust group differences and changes over time. If longitudinal data collection is not possible, measures from these regions (in addition to the striatum) may offer the most insight into disease progression. The parietal and frontal regions within these components are in complete agreement with striatal connectivity findings [25, 73]. Parietal projections to striatal convergence zones begin in the inferior parietal lobule (the primary contributor to component Q) and angular gyrus (another prominent region highlighted in our results), and two prefrontal connectivity clusters that displayed prodromal group differences in an independent study included middle frontal gyrus and superior frontal gyrus, respectively [25]. Previous studies have also implicated the cerebellum in HD pathology, including cerebellar volume loss in both GM and WM [74]. Most recently, Wolf et al. reported reduced volumes in right cerebellar lobule VII in early manifest HD patients, in addition to abnormal coupling with this region; right cerebellar lobule VII was functionally coupled with paracentral, lingual, and inferior frontal regions in early HD and bilateral cerebellar, right prefrontal, and cingulate areas in control participants [75]. Furthermore, paracentral connectivity was related to motor symptoms and disease burden in HD patients. An additional study reported abnormal diffusion in cerebellar GM and WM that was associated with poorer motor functioning (e.g., higher total motor score and saccade, finger tapping, and tandem walking impairments) [76], and a further post-mortem study of eight HD patients found multiple neurodegenerative cerebellar features that included macroscopic atrophy and Purkinje cell loss, in addition to nerve cell loss in all deep cerebellar nuclei [77]. These features were present even in patients with milder striatal atrophy.

Clinical functioning and WM throughout the HD prodrome

WM components that aligned most with clinical functioning typically also exhibited many significant group differences cross-sectionally and over time, indicating that areas essential for prodromal clinical functioning also differ across prodromal stages as well as between prodromal and control individuals. The overall patterns of alignment among component and clinical variable trajectories support our prediction that distinct regional changes accompany changes in different clinical functioning domains (Fig. 2). However, we observed some regional overlap in brain areas that were most implicated in clinical functioning, particularly in superior occipital (SC, SI, SW, TMTA, TMTB), precuneus (SC, SI, TMTA, TMTB), superior parietal lobule (SC, SI, TMTA, TMTB), angular gyrus (SI, TMTB, UHDRS-TMS), middle occipital (SDMT, SI, TMTB), associative visual cortex (SW, TMTA, TMTB), supramarginal gyrus (SI, TMTB), cuneus (SW, TMTA), and inferior parietal lobule (SI, TMTB).

Additionally, the results overall suggest that clinical functioning is best reflected by different brain regions depending on prodromal stage, although there was also some overlap among regions implicated at different prodromal stages. Parietal regions (including precuneus, superior parietal lobule, and angular gyrus) appear to play an influential role in clinical functioning throughout the prodrome, with the angular gyrus standing out as one of the only regions actively involved in both cognitive and motor functioning.

The involvement of parietal and frontal regions in cognitive functioning during early stages of (and throughout) the prodrome is consistent with proposed compensatory mechanisms involving increased cortical recruitment of these regions during movement (e.g., the Simon button press task) in HD patients [78]. Intrinsic neural activity alterations have been observed in the precuneus and angular gyrus of HD patients, and significant correlations between cognitive performance and precuneus activation and thinning have also been reported [79, 80]. Furthermore, parietal regions BA 5 and 7 contain neurons that selectively respond to specific hand and mouth movements [81], and these regions were both present in components with noticeable clinical effects in our studies. Frontal regions implicated in movement (premotor, supplementary motor, and middle frontal) showed some of the strongest similarities with clinical functioning in the early prodrome. In a positron emission tomography (PET) study comparing brain activation during hand movements in HD patients and controls, HD patients exhibited abnormal movement-related activation in the supplementary motor cortex, premotor cortex, and anterior cingulate, as well as the striatum, in conjunction with enhanced parietal activity that may reflect compensatory mechanisms [82]. Similarly, an additional PET study reported frontal lobe and putamen activity alterations in HD patients during a movement task, with inadequate activation in supplementary and premotor cortices (in addition to the cingulum) correlating with performance accuracy [81]; here, enhanced left parietal cortex activation was associated with improved performance in HD patients.

By the middle prodrome, the results suggest a more pivotal role of the angular gyrus, limbic lobe, and middle occipital gyrus in clinical functioning, with comparatively diminished frontal involvement. Previous work has shown that middle occipital diffusivity changes can distinguish prodromal stages, and poorer Stroop Word performance has been associated with increased diffusivity in the superior and middle occipital gyri [83]. An additional study found that the only diffusivity changes distinguishing prodromal groups were in the superior fronto-occipital fasciculus, with the most considerable changes observed in the group closest to diagnosis. Diffusivity increases in the superior fronto-occipital fasciculus were also associated with worsening motor functioning over time (i.e., motor symptom increases) [33].

By the late prodrome, parietal areas implicated in early prodromal clinical functioning in our results (including the precuneus and superior parietal lobule) again aligned well with clinical functioning, and change in the angular gyrus continued to strongly resemble changes in both cognitive and motor functioning. Other regions that appeared to play an increasing role in late prodromal clinical functioning included the superior occipital gyrus and inferior parietal lobule. Our results suggest shifting occipital involvement in prodromal clinical functioning, with early prodromal clinical changes aligning most with associative visual cortex changes, and middle and superior occipital being most implicated in middle and late prodromal clinical changes, respectively.

Conclusions

Overall, these results support the hypothesis that distinct regions most underlie cognitive compared to motor impairments in the HD prodrome, and further suggest that regions underlying these processes differ at different prodromal stages. These findings portray a regionally widespread prodromal pattern of WM changes marked by intense, early motor cortical and middle frontal gyral (in additional to striatal) changes that increasingly extend sub-gyrally with increased prodromal progression. The results are also in keeping with previous reports highlighting UHDRS-TMS as the best indicator of prodromal WM changes. However, at the High prodromal stage, cognitive variables (particularly Stroop Interference) outperformed UHDRS-TMS, and TMTA exhibited the most similarity to WM profiles overall across groups (especially cross-sectionally).

Many observations were in keeping with predictions; WM in the High and Low groups respectively differed most and least drastically from that of the Control group, and the most robust longitudinal clinical functioning changes were also observed in the High group. The prediction that motor performance changes would most resemble striatal and motor-implicated frontal change was partially supported, but results indicated a stronger frontal involvement in the early prodrome than anticipated; supplementary motor and striatal changes most agreed with changing UHDRS-TMS performance in the Low and High groups, respectively. Prodromal frontal WM changes also did not always align most with cognitive changes. Instead, many structural components that best reflected changing clinical functioning also highlighted non-frontal regions with prominent striatal connections and essential roles in cognition, such as the angular gyrus.

Limitations

The present study used T1-weighted images and tissue-specific probability maps to examine WM morphology. These findings can thus not be interpreted as analogous to DTI and DWI studies examining WM tract integrity. As the structural patterns in this study are adjacent to GM and CSF, it is likely that changes in these adjacent areas influence the morphology of, and are not completely separate from, the WM morphology examined in this study. Nonetheless, the findings largely overlapped with previous WM studies, including a DWI LMM study involving a cohort of the same participants included in this study, and the issue of prodromal changes in GM-adjacent-WM is in itself an interesting question that has not been examined in many previous studies. It will be necessary and interesting to compare these findings to those in the GM morphology maps from the same data set to investigate this issue further.

Because HD is a rare condition, there is an unavoidable tradeoff between maximizing power (afforded by a large sample) and data homogeneity (enabled by uniform data collection). Like most studies involving many participants with an uncommon condition, PREDICT-HD collected data from multiple sites using different MRI scanners. Several steps were taken to minimize the complications of analyzing multi-site data. Before data collection, PREDICT-HD established uniform protocols across sites. Following data collection, sites were examined for outliers and unbalanced participant demographics. For analyses, random site intercepts were included to account for correlations due to inherent scanner or site-specific features. Notably, the SBM method itself has also been shown to eradicate site effects [84], and the reliability of DWI data collected across these sites has also been previously demonstrated [85].

Future directions

Reports regarding the chronology of brain structural changes in prodromal HD are conflicting. It is still unclear whether WM and gray matter (GM) changes occur mostly simultaneously or if robust changes in WM precede those in GM (or vice versa); thus, a promising future direction for this work is to directly compare WM changes with those in GM, which is achievable by applying SBM to the GM probability maps extracted in the first phase of this study and performing similar stacked LMM analyses to compare WM and GM trajectories from the same participants and scans. This analysis could also address some of the uncertainty regarding how much GM morphology changes influence the morphology of adjacent WM (examined in this study). It is also possible that prodromal GM change more aptly reflects cognitive or motor changes, which could similarly be addressed in a similar LMM study assessing GM in conjunction with clinical variables.

CONFLICT OF INTEREST

The authors have no conflict of interest to report.

Footnotes

ACKNOWLEDGMENTS

We thank Drs. Jeffrey Long and Spencer Lourens for their assistance developing the statistical approach.

This project was supported by 1U01NS082074 (V.C. and J.T., co-principal investigators) from the National Institutes of Health, National Institute of Neurological Disorders and Stroke. The PREDICT-HD study was supported by NIH/NINDS grant 5R01NS040068 awarded to J.P.; CHDI Foundation, Inc., A3917 and 6266 awarded to J.P.; Cognitive and Functional Brain Changes in Preclinical Huntington Disease (HD) 5R01NS054893 awarded to J.P.; 4D Shape Analysis for Modeling Spatiotemporal Change Trajectories in Huntington 1U01NS082086; Functional Connectivity in Premanifest Huntington Disease 1U01NS082083; and Basal Ganglia Shape Analysis and Circuitry in Huntington Disease 1U01NS082085 awarded to Christopher A. Ross. This work was also supported by grants from the National Institutes of Health (5R01NS040068, 1U01NS082083, 5R01 NS054893, NS050568, P20GM103472, and R01REB020407) and the CHDI Foundation (A-5008), National Alliance for Medical Image Computing (NAMIC; EB005149 / Brigham and Women’s Hospital), and Enterprise Storage in a Collaborative Neuroimaging Environment (S10 RR023392/NCCR Shared Instrumentation Grant).

We thank the PREDICT-HD Investigators and coordinators of the Huntington Study Group; PREDICT-HD sites: University of Iowa, Cleveland Clinic, University College London, Indiana University, Washington University in St. Louis, University California San Francisco; the study participants; the National Research Roster for Huntington Disease Patients and Families; the Huntington Disease Society of America.

The supplementary material is available in the electronic version of this article: .

References

Paulsen

JS ,

Long

JD ,

Johnson

HJ ,

Aylward

EH ,

Ross

CA ,

Williams

JK , et al.Clinical and biomarker changes in premanifest Huntington disease show trial feasibility: A decade of the PREDICT-HD Study. Front Aging Neurosci. 2014;6:78.

Harrington

DL ,

Liu

D ,

Smith

MM ,

Mills

JA ,

Long

JD ,

Aylward

EH , et al.Neuroanatomical correlates of cognitive functioning in prodromal Huntington disease. Brain Behav. 2014;4(1):29–40.

Paulsen

JS . Cognitive impairment in Huntington disease: Diagnosis and treatment. Curr Neurol Neurosci Rep. 2011;11(5):474–83.

Lazic

SE ,

Goodman

AO ,

Grote

HE ,

Blakemore

C ,

Morton

AJ ,

Hannan

AJ , et al. Olfactory abnormalities in Huntington’s disease: Decreased plasticity in the primary olfactory cortex of R6/1 transgenic mice and reduced olfactory discrimination in patients. Brain Res. 2007;1151:219–26.

Dogan

I ,

SaB

C ,

Mirzazade

S ,

Kleiman

A ,

Werner

CJ ,

Pohl

A , et al. Neural correlates of impaired emotion processing in manifest Huntington’s disease. Soc Cogn Affect Neurosci.2014;9(5):671–80.

Kalkhoven

C ,

Sennef

C ,

Peeters

A , van den Bos

. Risk-taking and pathological gambling behavior in Huntington’s disease. Front Behav Neurosci. 2014;8:103.

Rawlins

MD ,

Wexler

NS ,

Wexler

AR ,

Tabrizi

SJ ,

Douglas

I ,

Evans

SJ , et al. The prevalence of Huntington’s disease. Neuroepidemiology. 2016;46(2):144–53.

Golde

TE ,

Borchelt

DR ,

Giasson

BI ,

Lewis

J . Thinking laterally about neurodegenerative proteinopathies. J Clin Invest. 2013;123(5):1847–55.

A novel gene containing a trinucleotide repeat that is expanded and unstable on Huntington’s disease chromosomes. The Huntington’s Disease Collaborative Research Group. Cell. 1993;72(6):971–83.

10.

Myers

RH . Huntington’s disease genetics. NeuroRx. 2004;1(2):255–62.

11.

Long

JD ,

Paulsen

JS , Group P-HIaCotHS. Multivariate prediction of motor diagnosis in Huntington’s disease: 12 years of PREDICT-HD. Mov Disord. 2015;30(12):1664–72.

12.

Zhang

Y ,

Long

JD ,

Mills

JA ,

Warner

JH ,

W ,

Paulsen

JS , et al. Indexing disease progression at study entry with individuals at-risk for Huntington disease. Am J Med Genet B Neuropsychiatr Genet. 2011;156B(7):751–63.

13.

Biglan

KM ,

Zhang

Y ,

Long

JD ,

Geschwind

M ,

Kang

GA ,

Killoran

A , et al. Refining the diagnosis of Huntington disease: The PREDICT-HD study. Front Aging Neurosci. 2013;5:12.

14.

Ciarochi

JA ,

Calhoun

VD ,

Lourens

S ,

Long

JD ,

Johnson

HJ ,

Bockholt

HJ , et al. Patterns of co-occurring gray matter concentration loss across the Huntington disease prodrome. Front Neurol. 2016;7:147.

15.

Paulsen

JS ,

Long

JD ,

Ross

CA ,

Harrington

DL ,

Erwin

CJ ,

Williams

JK , et al. Prediction of manifest Huntington’s disease with clinical and imaging measures: A prospective observational study. Lancet Neurol. 2014;13(12):1193–201.

16.

Aylward

EH ,

Harrington

DL ,

Mills

JA ,

Nopoulos

PC ,

Ross

CA ,

Long

JD , et al. Regional atrophy associated with cognitive and motor function in prodromal Huntington disease. J Huntingtons Dis. 2013;2(4):477–89.

17.

Misiura

MB ,

Lourens

S ,

Calhoun

VD ,

Long

J ,

Bockholt

J ,

Johnson

H , et al. Cognitive control, learning, and clinical motor ratings are most highly associated with basal ganglia brain volumes in the premanifest Huntington’s disease phenotype. J Int Neuropsychol Soc. 2017;23(2):159–70.

18.

Espinoza

FA ,

Turner

JA ,

Vergara

VM ,

Miller

RL ,

Mennigen

E ,

Liu

J , et al. Whole-brain connectivity in a large study of Huntington’s disease gene mutation carriers and healthy controls. Brain Connect. 2018;8(3):166–78.

19.

Tabrizi

SJ ,

Langbehn

DR ,

Leavitt

BR ,

Roos

RA ,

Durr

A ,

Craufurd

D , et al. Biological and clinical manifestations of Huntington’s disease in the longitudinal TRACK-HD study: Cross-sectional analysis of baseline data. Lancet Neurol. 2009;8(9):791–801.

20.

Tabrizi

SJ ,

Scahill

RI ,

Durr

A ,

Roos

RA ,

Leavitt

BR ,

Jones

R , et al. Biological and clinical changes in premanifest and early stage Huntington’s disease in the TRACK-HD study: The 12-month longitudinal analysis. Lancet Neurol. 2011;10(1):31–42.

21.

Tabrizi

SJ ,

Reilmann

R ,

Roos

RA ,

Durr

A ,

Leavitt

B ,

Owen

G , et al. Potential endpoints for clinical trials in premanifest and early Huntington’s disease in the TRACK-HD study: Analysis of 24 month observational data. Lancet Neurol. 2012;11(1):42–53.

22.

Faria

AV ,

Ratnanather

JT ,

Tward

DJ ,

Lee

DS , van den Noort

D , et al. Linking white matter and deep gray matter alterations in premanifest Huntington disease. Neuroimage Clin. 2016;11:450–60.

23.

D ,

Faria

AV ,

Younes

L ,

Mori

S ,

Brown

T ,

Johnson

H , et al. Mapping the order and pattern of brain structural MRI changes using change-point analysis in premanifest Huntington’s disease. Hum Brain Mapp. 2017;38(10):5035–50.

24.

Poudel

GR ,

Stout

JC , Domínguez

D JF

Salmon

L ,

Churchyard

A ,

Chua

P , et al. White matter connectivity reflects clinical and cognitive status in Huntington’s disease. Neurobiol Dis. 2014;65:180–7.

25.

Shaffer

JJ ,

Ghayoor

A ,

Long

JD ,

Kim

RE ,

Lourens

S ,

O’Donnell

LJ , et al. Longitudinal diffusion changes in prodromal and early HD: Evidence of white-matter tract deterioration. Hum Brain Mapp. 2017;38(3):1460–77.

26.

Orth

M ,

Gregory

S ,

Scahill

RI ,

Mayer

IS ,

Minkova

L ,

Klöppel

S , et al. Natural variation in sensory-motor white matter organization influences manifestations of Huntington’s disease. Hum Brain Mapp. 2016;37(12):4615–28.

27.

Paulsen

JS ,

Nopoulos

PC ,

Aylward

E ,

Ross

CA ,

Johnson

H ,

Magnotta

VA , et al. Striatal and white matter predictors of estimated diagnosis for Huntington disease. Brain Res Bull. 2010;82(3-4):201–7.

28.

Paulsen

JS ,

Magnotta

VA ,

Mikos

AE ,

Paulson

HL ,

Penziner

E ,

Andreasen

NC , et al. Brain structure in preclinical Huntington’s disease. Biol Psychiatry. 2006;59(1):57–63.

29.

Crawford

HE ,

Hobbs

NZ ,

Keogh

R ,

Langbehn

DR ,

Frost

C ,

Johnson

H , et al. Corpus callosal atrophy in premanifest and early Huntington’s disease. J Huntingtons Dis. 2013;2(4):517–26.

30.

Magnotta

VA ,

Kim

J ,

Koscik

T ,

Beglinger

LJ ,

Espinso

D ,

Langbehn

D , et al. Diffusion tensor imaging in preclinical Huntington’s disease. Brain Imaging Behav. 2009;3(1):77–84.

31.

Rosas

HD ,

Tuch

DS ,

Hevelone

ND ,

Zaleta

AK ,

Vangel

M ,

Hersch

SM , et al. Diffusion tensor imaging in presymptomatic and early Huntington’s disease: Selective white matter pathology and its relationship to clinical measures. Mov Disord. 2006;21(9):1317–25.

32.

Stoffers

D ,

Sheldon

S ,

Kuperman

JM ,

Goldstein

J ,

Corey-Bloom

J ,

Aron

AR . Contrasting gray and white matter changes in preclinical Huntington disease: An MRI study. Neurology. 2010;74(15):1208–16.

33.

Harrington

DL ,

Long

JD ,

Durgerian

S ,

Mourany

L ,

Koenig

K ,

Bonner-Jackson

A , et al. Cross-sectional and longitudinal multimodal structural imaging in prodromal Huntington’s disease. Mov Disord. 2016;31(11):1664–75.

34.

Phillips

O ,

Squitieri

F ,

Sanchez-Castaneda

C ,

Elifani

F ,

Caltagirone

C ,

Sabatini

U , et al. Deep white matter in Huntington’s disease. PLoS One. 2014;9(10):e109676.

35.

Phillips

O ,

Sanchez-Castaneda

C ,

Elifani

F ,

Maglione

V , Di Pardo

Caltagirone

C , et al. Tractography of the corpus callosum in Huntington’s disease. PLoS One. 2013;8(9):e73280.

36.

Matsui

JT ,

Vaidya

JG ,

Johnson

HJ ,

Magnotta

VA ,

Long

JD ,

Mills

JA , et al. Diffusion weighted imaging of prefrontal cortex in prodromal Huntington’s disease. Hum Brain Mapp. 2014;35(4):1562–73.

37.

Aylward

EH ,

Nopoulos

PC ,

Ross

CA ,

Langbehn

DR ,

Pierson

RK ,

Mills

JA , et al. Longitudinal change in regional brain volumes in prodromal Huntington disease. J Neurol Neurosurg Psychiatry. 2011;82(4):405–10.

38.

Poudel

GR ,

Stout

JC ,

Domínguez D

JF ,

Churchyard

A ,

Chua

P ,

Egan

GF , et al. Longitudinal change in white matter microstructure in Huntington’s disease: The IMAGE-HD study. Neurobiol Dis. 2015;74:406–12.

39.

Odish

OF ,

Leemans

A ,

Reijntjes

RH ,

van den Bogaard

SJ ,

Dumas

EM ,

Wolterbeek

R , et al. Microstructural brain abnormalities in Huntington’s disease: A two-year follow-up. Hum Brain Mapp. 2015;36(6):2061–74.

40.

Matsui

JT ,

Vaidya

JG ,

Wassermann

D ,

Kim

RE ,

Magnotta

VA ,

Johnson

HJ , et al. Prefrontal cortex white matter tracts in prodromal Huntington disease. Hum Brain Mapp. 2015;36(10):3717–32.

41.

Papp

KV ,

Snyder

PJ ,

Mills

JA ,

Duff

K ,

Westervelt

HJ ,

Long

JD , et al. Measuring executive dysfunction longitudinally and in relation to genetic burden, brain volumetrics, and depression in prodromal Huntington disease. Arch Clin Neuropsychol. 2013;28(2):156–68.

42.

L ,

Groth

KM ,

Pearlson

G ,

Schretlen

DJ ,

Calhoun

VD . Source-based morphometry: The use of independent component analysis to identify gray matter differences with application to schizophrenia. Hum Brain Mapp. 2009;30(3):711–24.

43.

Gorgolewski

K ,

Burns

CD ,

Madison

C ,

Clark

D ,

Halchenko

YO ,

Waskom

ML , et al. Nipype: A flexible, lightweight and extensible neuroimaging data processing framework in python. Front Neuroinform. 2011;5:13.

44.

Avants

BB ,

Tustison

NJ ,

Song

G ,

Cook

PA ,

Klein

A ,

Gee

JC . A reproducible evaluation of ANTs similarity metric performance in brain image registration. Neuroimage. 2011;54(3):2033–44.

45.

Young Kim

E ,

Johnson

HJ . Robust multi-site MR data processing: Iterative optimization of bias correction, tissue classification, and registration. Front Neuroinform. 2013;7:29.

46.

Ghayoor

A ,

Paulsen

JS ,

Kim, EY

R ,

Johnson

HJ . Tissue classification of large-scale multi-site MR data using fuzzy k-nearest neighbor method. Proc SPIE Med Imaging. 2016;9784:97841V.

47.

Ghayoor

A ,

Vaidya

JG ,

Johnson

HJ . Robust automated constellation-based landmark detection in human brain imaging. Neuroimage. 2018;170:471–81.

48.

Wechsler

D . Manual for the Wechsler Adult Intelligence Scale. Oxford, England: Psychological Corporation; 1955.

49.

Smith

A . Symbol Digit Modalities Test (SDMT) Manual (Revised). Los Angeles, CA: Western Psychological Services.; 1982.

50.

Stroop

JR . Studies of interference in serial verbal reactions.J Exp Psychol. 1935;18(6):643–62.

51.

O’Rourke

JJ ,

Beglinger

LJ ,

Smith

MM ,

Mills

J ,

Moser

DJ ,

Rowe

KC , et al. The Trail Making Test in prodromal Huntington disease: Contributions of disease progression to test performance. J Clin Exp Neuropsychol. 2011;33(5):567–79.

52.

Reitan

. Validity of the trail making test as an indicator of organic brain damage. Percept Mot Skills. 1958;8:271–6.

53.

Genetic Modifiers of Huntington’s Disease (GeM-HD) Consortium. Identification of genetic factors that modify clinical onset of Huntington’s disease. Cell. 2015;162(3):516–26.

54.

Gupta

CN ,

Calhoun

VD ,

Rachakonda

S ,

Chen

J ,

Patel

V ,

Liu

J , et al. Patterns of gray matter abnormalities in schizophrenia based on an international mega-analysis. Schizophr Bull. 2015;41(5):1133–42.

55.

Segall

JM ,

Allen

EA ,

Jung

RE ,

Erhardt

EB ,

Arja

SK ,

Kiehl

K , et al. Correspondence between structure and function in the human brain at rest. Front Neuroinform. 2012;6:10.

56.

YO ,

Adali

T ,

Calhoun

VD . Estimating the number of independent components for functional magnetic resonance imaging data. Hum Brain Mapp. 2007;28(11):1251–66.

57.

Himberg

J ,

Hyvärinen

A ,

Esposito

F . Validating the independent components of neuroimaging time series via clustering and visualization. Neuroimage. 2004;22(3):1214–22.

58.

S ,

Correa

NM ,

XL ,

Eichele

T ,

Calhoun

VD , AdaliT . Automatic identification of functional clusters in FMRI data using spatial dependence. IEEE Trans Biomed Eng. 2011;58(12):3406–17.

59.

Verbeke

G ,

Molenberghs

G . Linear Mixed Models for Longitudinal Data. New York: Springer; 2000.

60.

Long

JD . Longitudinal Data Analysis for the Behavioral Sciences Using R. Thousand Oaks, California: SAGE Publications, Inc.; 2011.

61.

Lambrecq

V ,

Langbour

N ,

Guehl

D ,

Bioulac

B ,

Burbaud

P ,

Rotge

JY . Evolution of brain gray matter loss in Huntington’s disease: A meta-analysis. Eur J Neurol. 2013;20(2):315–21.

62.

Mühlau

M ,

Gaser

C ,

Wohlschläger

AM ,

Weindl

A ,

Städtler

M ,

Valet

M , et al. Striatal gray matter loss in Huntington’s disease is leftward biased. Mov Disord. 2007;22(8):1169–73.

63.

Klöppel

S ,

Gregory

S ,

Scheller

E ,

Minkova

L ,

Razi

A ,

Durr

A , et al. Compensation in preclinical Huntington’s disease: Evidence from the Track-On HD Study. EBioMedicine. 2015;2(10):1420–9.

64.

Aylward

EH ,

Codori

AM ,

Barta

PE ,

Pearlson

GD ,

Harris

GJ ,

Brandt

J . Basal ganglia volume and proximity to onset in presymptomatic Huntington disease. Arch Neurol. 1996;53(12):1293–6.

65.

Harris

GJ ,

Pearlson

GD ,

Peyser

CE ,

Aylward

EH ,

Roberts

J ,

Barta

PE , et al. Putamen volume reduction on magnetic resonance imaging exceeds caudate changes in mild Huntington’s disease. Ann Neurol. 1992;31(1):69–75.

66.

Tabrizi

SJ ,

Scahill

RI ,

Owen

G ,

Durr

A ,

Leavitt

BR ,

Roos

RA , et al. Predictors of phenotypic progression and disease onset in premanifest and early-stage Huntington’s disease in the TRACK-HD study: Analysis of 36-month observational data. Lancet Neurol. 2013;12(7):637–49.

67.

Ciarochi

JA ,

Liu

J ,

Calhoun

V ,

Johnson

H ,

Misiura

M ,

Bockholt

HJ , et al. High and low levels of an NTRK2-driven genetic profile affect motor- and cognition-associated frontal gray matter in prodromal Huntington’s disease. Brain Sci. 2018;8(7):E116.

68.

Rosas

HD ,

Salat

DH ,

Lee

SY ,

Zaleta

AK ,

Pappu

V ,

Fischl

B , et al. Cerebral cortex and the clinical expression of Huntington’s disease: Complexity and heterogeneity. Brain. 2008;131(Pt 4):1057–68.

69.

Sánchez-Castañeda

C ,

de Pasquale

F ,

Caravasso

CF ,

Marano

M ,

Maffi

S ,

Migliore

S , et al. Resting-state connectivity and modulated somatomotor and default-mode networks in Huntington disease. CNS Neurosci Ther. 2017;23(6):488–97.

70.

Werner

CJ ,

Dogan

I ,

SaB

C ,

Mirzazade

S ,

Schiefer

J ,

Shah

NJ , et al. Altered resting-state connectivity in Huntington’s disease. Hum Brain Mapp. 2014;35(6):2582–93.

71.

Koenig

KA ,

Lowe

MJ ,

Harrington

DL ,

Lin

J ,

Durgerian

S ,

Mourany

L , et al. Functional connectivity of primary motor cortex is dependent on genetic burden in prodromal Huntington disease. Brain Connect. 2014;4(7):535–46.

72.

Macdonald

V ,

Halliday

GM ,

Trent

RJ ,

McCusker

EA . Significant loss of pyramidal neurons in the angular gyrus of patients with Huntington’s disease. Neuropathol Appl Neurobiol. 1997;23(6):492–5.

73.

Jarbo

K ,

Verstynen

TD . Converging structural and functional connectivity of orbitofrontal, dorsolateral prefrontal, and posterior parietal cortex in the human striatum. J Neurosci. 2015;35(9):3865–78.

74.

Fennema-Notestine

C ,

Archibald

SL ,

Jacobson

MW ,

Corey-Bloom

J ,

Paulsen

JS ,

Peavy

GM , et al. In vivo evidence of cerebellar atrophy and cerebral white matter loss in Huntington disease. Neurology. 2004;63(6):989–95.

75.

Wolf

RC ,

Thomann

PA ,

Sambataro

F ,

Wolf

ND ,

Vasic

N ,

Landwehrmeyer

GB , et al. Abnormal cerebellar volume and corticocerebellar dysfunction in early manifest Huntington’s disease. J Neurol. 2015;262(4):859–69.

76.

Rees

EM ,

Farmer

R ,

Cole

JH ,

Haider

S ,

Durr

A ,

Landwehrmeyer

B , et al. Cerebellar abnormalities in Huntington’s disease: A role in motor and psychiatric impairment?Mov Disord. 2014;29(13):1648–54.

77.

Rüb

U ,

Hoche

F ,

Brunt

ER ,

Heinsen

H ,

Seidel

K ,

Del Turco

D , et al. Degeneration of the cerebellum in Huntington’s disease (HD): Possible relevance for the clinical picture and potential gateway to pathological mechanisms of the disease process. Brain Pathol. 2013;23(2):165–77.

78.

Georgiou-Karistianis

N ,

Sritharan

A ,

Farrow

M ,

Cunnington

R ,

Stout

J ,

Bradshaw

J , et al. Increased cortical recruitment in Huntington’s disease using a Simon task. Neuropsychologia. 2007;45(8):1791–800.

79.

Liu

W ,

Yang

J ,

Chen

K ,

Luo

C ,

Burgunder

J ,

Gong

Q , et al. Resting-state fMRI reveals potential neural correlates of impaired cognition in Huntington’s disease. Parkinsonism Relat Disord. 2016;27:41–6.

80.

Rosas

HD ,

Hevelone

ND ,

Zaleta

AK ,

Greve

DN ,

Salat

DH ,

Fischl

B . Regional cortical thinning in preclinical Huntington disease and its relationship to cognition. Neurology. 2005;65(5):745–7.

81.

Bartenstein

P ,

Weindl

A ,

Spiegel

S ,

Boecker

H ,

Wenzel

R ,

Ceballos-Baumann

AO , et al. Central motor processing in Huntington’s disease. A PET study. Brain. 1997;120 (Pt 9):1553–67.

82.

Weeks

RA ,

Ceballos-Baumann

A ,

Piccini

P ,

Boecker

H ,

Harding

AE ,

Brooks

DJ . Cortical control of movement in Huntington’s disease. A PET activation study. Brain. 1997;120 (Pt 9):1569–78.

83.

Odish

OFF ,

Reijntjes

RHAM ,

van den Bogaard

SJA ,

Roos

RAC ,

Leemans

A . Progressive microstructural changes of the occipital cortex in Huntington’s disease. Brain Imaging Behav. 2018;12(6):1786–94.

84.

Chen

J ,

Liu

J ,

Calhoun

VD ,

Arias-Vasquez

A ,

Zwiers

MP ,

Gupta

CN , et al. Exploration of scanning effects in multi-site structural MRI studies. J Neurosci Methods. 2014;230:37–50.

85.

Magnotta

VA ,

Matsui

JT ,

Liu

D ,

Johnson

HJ ,

Long

JD ,

Bolster

BD , et al. Multicenter reliability of diffusion tensor imaging. Brain Connect. 2012;2(6):345–55.