Investigation of the large-scale white-matter functional networks in spinocerebellar ataxia type 3

Introduction

Spinocerebellar ataxia type 3 (SCA3) is one of the most common subtypes of autosomal dominant ataxia, caused by a cytosine-adenine-guanine (CAG) repeat expansion in the ATXN3 gene (1,2). Substantial evidence suggests the impairment of white matter (WM) fibers in SCA3. Even before clinical onset, patients with SCA3 already exhibit structural damage in the WM in brainstem and cerebellum (3-5). Longitudinal studies have further revealed that, whether over 5 years or just 6 to 12 months of observation, and whether in advanced symptomatic phases or in the early clinical stage, there is marked, ongoing deterioration in WM volume (WMV) and microstructural integrity in patients with SCA3 (6-8). As such, impairments within WM play a key role in its pathological processes.

Recently, an increasing number of functional magnetic resonance imaging (fMRI) studies have begun to focus on exploring the neural activation and functional organization of WM by using blood-oxygenation-level-dependent (BOLD) signals. It is reported that, during task conditions, neural activity in WM occurs in synchrony with gray matter (GM) and distributes along the anatomical bundles (9,10). Moreover, the BOLD signal in WM is found to be driven by glial-vascular coupling and the energy demands of axonal conduction (11-13). These findings consistently indicate that neural activity in WM fibers is not merely a passive reflection of GM signals; rather, WM actively participates in neural processes, including information transmission and coordination across different GM regions. In the last years, studies have verified the intrinsic functional organization of WM, which manifests as interacting networks of functional modules that can be reproducibly detected (14-16). Studies integrating the functional properties of WM functional networks (WMFNs) into brain networks have yielded a more complete understanding of brain functions and their disturbances during aging and degenerative brain disorders (17-19). Given the previously reported spatial similarity between abnormalities of functional activity and WM integrity in SCA3, it seems relevant to integrate the structural and functional properties of WM pathways into analyses (8). However, alterations of neural activity and interactions within large-scale WMFNs remain uninvestigated in SCA3. Moreover, there is limited research combining the functional features within WMFNs and the underlying WM loss process.

To fill the current research gaps, the present study aimed to comprehensively explore changes in neural activity and interactions within WMFNs in SCA3, and to determine whether these functional features are associated with structural degeneration within WM and clinical ataxia severity. We hypothesized that alterations in neural activity and connectivity patterns would exist within WMFNs in SCA3 patients compared to healthy controls (HCs), potentially associated with underlying structural degeneration and clinical severity. We hope that this study provides deeper insights into the neural mechanisms underlying SCA3 from the perspective of WMFNs. We present this article in accordance with the STROBE reporting checklist (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-736/rc).

Methods

Participants

All participants were enrolled from the Department of Neurology of Xiangya Hospital, Central South University, Changsha, China. In total, 200 participants, including 98 patients with genetically confirmed SCA3 and 102 demographically matched HCs were recruited. The ataxia severity of each patient was assessed using the Scale for the Assessment and Rating of Ataxia (SARA). The exclusion criteria for all participants were as follows: (I) pregnancy or breastfeeding; (II) neurological diseases, psychiatric deficits, metabolic diseases, or tumors; (III) any contraindications for magnetic resonance imaging (MRI) examination; and (IV) any history of neuromodulation therapy or long-term medication use. Due to the susceptibility of WM BOLD signals to motion-related artifacts, we applied a strict motion exclusion criterion. Specifically, participants with head motion exceeding 2 mm in translation or 2° in rotation were excluded, which eliminated 32 cases (SCA3 =28, HCs =4). Importantly, the demographic matching between the SCA3 and HC groups was maintained both before and after the application of these exclusion criteria. The final analyzed sample consisted of 70 patients with SCA3 and 98 HCs. Demographic and clinical details are listed in Table 1.

The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. This prospective study was approved by the Medical Research Ethics Committee of Xiangya Hospital of Central South University (IBM No. 201601010). Written informed consent was provided by all individuals before participation.

Image acquisition

Imaging data were acquired on a Siemens Magnetom Prisma 3.0T MR scanner (Siemens, Erlangen, Germany) using a 64-channel head coil. Routine imaging sequences, including axial T1-weighted images, T2-weighted images, and T2-weighted fluid-attenuated inversion recovery (FLAIR) images, were obtained for every participant to rule out gross brain pathology. Functional images were acquired by using a gradient-echo echo-planar functional imaging sequence [time points = 240; repetition time/echo time (TR/TE) =2,400/30 ms; flip angle =90°; matrix size =64×64; field of view (FOV) =192 mm × 192 mm, number of axial slices =46; slice thickness =3 mm with no slice gap; and voxel size =3×3×3 mm³]. The participants were asked to stay still and close their eyes, but avoid thinking of anything or falling asleep during scanning. High-resolution T1-weighted images were acquired by a three-dimensional (3D) fast spoiled gradient-echo sequence (thickness/gap =0.90/0 mm; TR =2,300 ms; TE =2.32 ms; inversion time =900 ms; 192 sagittal slices; FOV = 240 mm ×240 mm; matrix size =128×128; flip angle =8°; voxel size =0.9×0.9×0.9 mm³).

Image preprocessing

Functional images were preprocessed using the software Data Processing Assistant for Resting-State fMRI (DPARSF, http://rfmri.org/DPARSF) (20) and the SPM12 (http://www.fil.ion.ucl.ac.uk/spm/software/spm12). For each participant’s fMRI data, the preprocessing steps were as follows: (I) Delete the first five functional volumes. (II) slice-timing correction. (III) Head movement correction and realignment to the mean functional image. (IV) Co-registration of T1 images with the above preprocessed functional images, and segmentation into GM, WM, and cerebrospinal fluid (CSF) by Diffeomorphic Anatomical Registrations through Exponentiated Lie Algebra (DARTEL) (21). (V) Regression of average CSF signal and 24 rigid-body motion parameters. In this step, WM signals and whole-brain signals are not regressed in order to avoid excluding signals of interest. (VI) Temporal scrubbing using motion spikes [framewise displacement (FD) >1 mm] as separate repressors was performed (22). (VII) Application of a band-pass filter (0.01–0.10 Hz) to minimize high-frequency physiological noise. (VIII) Use of the segmented T1 image to identify GM and WM masks (threshold 0.5), with the GM and WM smoothed separately [4 mm full width at half maximum (FWHM)] (14). (IX) Normalization into the MNI space by DARTEL and resampled to 3×3×3 mm³ voxels.

WMFN clustering

The analysis codes and details were described in the original research (14). Briefly, the steps are as follows. Firstly, a unified group-level WM mask was created using the segmentation results of T1 images. Specifically, the group WM mask was first defined as voxels showing spatial overlap in at least 60% of the participants. The thalamus, caudate nucleus, putamen, globus pallidus, and nucleus accumbens (based on the Harvard-Oxford Atlas) were removed from the WM mask. The resulting mask was then compared with the preprocessed functional images of each participant; only voxels covered by functional data from more than 80% of the participants were preserved. Subsequently, WMFNs were identified by K-means clustering approach (distance metric-correlation) based on the Pearson’s correlation matrices between WM voxels that group masks restricted. In order to evaluate the stability of the number of networks obtained from clustering within the range of 2 to 22 clusters, the functional connectivity matrix was randomly divided into four submatrices, and each submatrix was clustered separately (100 replicates for each submatrix). To determine the similarity among the clustering results for each submatrix, the adjacency matrices were computed for every pair of submatrices and then compared using the Dice coefficient (the threshold was set at 0.90). The average Dice coefficient was subsequently used to assess the stability of the chosen number of clusters. Finally, the number of clusters that yielded the most detailed and stable WMFNs was 14. To determine the correlations between each WMFN and superficial GM functional network (GMFN), we further calculated their functional connectivity with GMFNs defined by Yeo et al.’s previous study (23).

Neural activity and inter-network coefficient Granger causality analyses (GCAs)

The neural activity of WMFNs was characterized using power spectrum analysis (24,25). Specifically, amplitudes at each frequency were extracted from each WM network for each participant through Fourier transformation. Frequency-power plots for each network were generated by averaging the amplitudes of all participants within each group.

The interactions among WMFNs were explored by using coefficient Granger causality analysis (cGCA) (26). This approach is a type of GCA method, which can yield information concerning excitatory or inhibitory causal influences among functional networks by using the regression coefficient. Therefore, it may be an effective method to detect the directed interactional connections among WMFNs. For each network, the individual preprocessed fMRI time series was extracted by averaging the time series of all voxels within. The Granger causality (GC) strength among WMFNs was calculated using the software REST (v1.8; www.restfmri.net). Each regression coefficient characterizes the signed strength and direction of the relationship between each two networks, in similarity with previous studies (17,27). The positive/negative causal coefficients that source activity predicts subsequent increases/decreases in target activity were interpreted as excitatory/inhibitory paths. Finally, a directed asymmetric matrix (14×14 regression coefficient matrix) was obtained for each participant. We further defined the in/out strength value as the cGCA graph-theoretic metric to describe the inflow/outflow influence strengths of each network: sum of absolute regression coefficients of certain networks where the network is the source/target variable to significantly predict other networks, respectively, denotes global GC inflow/outflow connections of each network.

Statistical analyses

For each WMFN, the average amplitudes of all participants were used to represent the degree of network activity, and a two-sample t-test was performed to detect neural activity differences between the two groups, controlling for sex and age as confounding variables. The within-group influence patterns among WMFNs were examined using one-sample t-tests on the influence coefficient matrices across the participants within each group. Meanwhile, the network-level between-groups difference patterns for each directed edge and the in/out strength of each network were obtained using two-sample t-tests. For the GCA analyses, Euclidean distances between each two networks were regressed from the coefficient matrices to exclude spatial effect and correlation disturbance (17), then sex and age were controlled for as confounding variables. The Euclidean distance was calculated using the L2 norm of any pair of node centroids in the Montreal Neurological Institute (MNI) space (28). For all the above analyses, FDR correction was used to resolve multiple comparisons among multiple WM networks. A P value <0.05 was used to determine significance. To assess the relationships between WMFNs and the SARA scores of SCA3, we used Pearson’s correlation coefficient, and correlations were considered significant at P<0.05.

Exploratory analyses of associations between functional features and structural atrophy within WMFN

To further explore the anatomical neural basis of neural activity and interactions within WMFNs, the WMV of the 14 WMFNs was calculated by using Computational Anatomy Toolbox (CAT12, http://www.neuro.uni-jena.de/cat/) following standard, widely used methods, the details of which are clarified in Appendix 1. Between-group differences of WMV value in WMFNs were calculated using a two-sample t-test, with sex and age as confounding variables. Pearson’s correlation analyses were then performed between the WMV value, the average amplitude value, and in/out strength of networks.

Results

Organization of WMFNs

Using K-means clustering on the WM voxel-wise correlation matrices, we identified 14 stable and detailed WMFNs (Dice coefficient >0.9, Figure 1). The resulting networks could be divided into two groups: supratentorial and infratentorial. Supratentorial networks could be further divided into three layers: superficial, middle, and deep, where superficial networks mostly correspond to functional subdivisions of the overlying GM. We qualitatively labeled these superficial WMFNs based on their correspondence with known resting-state GM networks (GMNs). Other networks were then named by the spatial locations of established WM regions. The 14 networks were named: WM1, the default-mode WM network; WM2 the executive control WM network; WM3, the temporo-motor WM network; WM4, the somatomotor WM network; WM5, the limbic WM network; WM6, the visual WM network; WM7, the dorsal frontoparietal WM network; WM8, the forceps minor network; WM9, the internal capsule network; WM10, the deep WM network; WM11, the deep occipital WM network; WM12, the midbrain network; WM13, the cerebellar WM network; and WM14, the pons network. WM1 to WM6 were superficial networks, WM7 to WM9 were middle, WM10 and WM11 were deep, and WM12 to WM14 were infratentorial WMFNs. The detailed organization of the networks is presented in Figure 1. For each WMFN, its detailed information is listed in Table 2.

Figure 1 Based on the Dice coefficient, 14 reliable and stable clustered WMFNs consisting of 4 groups were obtained for SCA3 and HCs: WM1, the default-mode white matter network; WM2, the executive control white matter network; WM3, the temporo-motor white matter network; WM4, the somatomotor white matter network; WM5, the limbic white matter network; WM6, the visual white matter network; WM7, the dorsal frontoparietal white matter network; WM8, the forceps minor network; WM9, the internal capsule network; WM10, the deep white matter network; WM11, the deep occipital white matter network; WM12, the midbrain network; WM13, the cerebellar white matter network; and WM14, the pons network. WM1 to WM6 were superficial networks, WM7 to WM9 were middle, WM10 and WM11 were deep, and WM12 to WM14 were infratentorial WMFNs. WM, white matter; WMFNs, white matter functional networks.

Activity of WMFNs

Frequency amplitudes in WMFNs are shown in Figure 2. For the two groups, most of the networks displayed a decreasing trend in amplitude with increased frequency, suggesting greater activity in WM at lower frequencies. On the contrary, the WMFNs located in deep layers had increasing trends in amplitude with frequency. In addition, compared with the HCs, significantly decreased amplitudes in SCA3 were shown in WM6, WM7, and WM9–14. These dysfunctional WMFNs were located in middle-layer (WM7, WM9), deep-layer (WM10, WM11), infratentorial group (WM12–14), and occipital region (WM6), whereas normal WMFNs were located in superficial layer (WM1–5) and frontal region (WM8), as shown in Figure S1.

Figure 2 Averaged amplitude in the WMFNs of the SCA3 and HC group. (A) Frequency plots showing the activity frequency within each WMFN. Asterisk indicates a significant difference (*, P<0.05, false discovery rate corrected). WMFNs without significant between-group differences (normal WMFNs) were mainly located in superficial-layer (WM1–5) and frontal region (WM8). WMFNs with significantly decreased amplitudes in SCA3 (dysfunctional WMFNs) were mainly located in middle-layer (WM7, WM9), deep-layer (WM10, WM11), infratentorial group (WM12–14) and occipital region (WM6). (B) Plots of the between-group differences in the mean amplitude for each WMFN. The upper figure shows WMFNs without significant between-group differences, whereas the lower presents WMFNs with significantly decreased amplitudes in the SCA3 group. WM1, the default-mode white matter network; WM2, the executive control white matter network; WM3, the temporo-motor white matter network; WM4, the somatomotor white matter network; WM5, the limbic white matter network; WM6, the visual white matter network; WM7, the dorsal frontoparietal white matter network; WM8, the forceps minor network; WM9, the internal capsule network; WM10, the deep white matter network; WM11, the deep occipital white matter network; WM12, the midbrain network; WM13, the cerebellar white matter network; WM14, the pons network. f, frequency (Hz); HC, healthy controls; PA, patients with SCA3; SCA3, spinocerebellar ataxia type 3; WMFNs, white matter functional networks.

Within-group GC patterns of WMFNs

Within-group WMFNs influence patterns of the SCA3 and HCs groups are shown in Figure 3A,3B. The WMFNs in HCs demonstrated a complex pattern of directed information flow across extensive superficial-superficial and superficial-middle/infratentorial connections, indicating a relatively balanced and polycentric mode of network control. However, a mismatched and overall sparser effective connection pattern could be observed in SCA3 compared with HCs.

Figure 3 Within-group white matter functional networks causal influence patterns across SCA3 and HCs were determined by the one-sample t-test (FDR corrected P<0.05). Details of the one-sample t-test matrix are shown on the left, where positive/negative t-values denote excitatory or inhibitory influence, respectively. The right circular plots present the overall effective connections, red lines represent significant excitatory influence while blue ones represent significant inhibitory influence. The color value of each line becomes darker with the strength of the connection. WM1, the default-mode white matter network; WM2, the executive control white matter network; WM3, the temporo-motor white matter network; WM4, the somatomotor white matter network; WM5, the limbic white matter network; WM6, the visual white matter network; WM7, the dorsal frontoparietal white matter network; WM8, the forceps minor network; WM9, the internal capsule network; WM10, the deep white matter network; WM11, the deep occipital white matter network; WM12, the midbrain network; WM13, the cerebellar white matter network; WM14, the pons network. FDR, false discovery rate; HCs, healthy controls; SCA3, spinocerebellar ataxia type 3.

Between-group differences in GC patterns of WMFNs

Between-group differences in GC patterns are shown in Figure 4A. Compared with HCs, patients with SCA3 showed significantly enhanced influence strength from dysfunctional WMFNs to superficial-layer normal WMFNs (especially WM3 and WM4), including influences from: the middle-layer networks (WM9→WM3, WM7/WM9→WM4), the deep-layer networks (WM10→WM2, WM10/WM11→WM4, WM11→WM6), and the infratentorial networks (WM13→WM3/WM4). Meanwhile, significantly enhanced influences were observed within dysfunctional WMFNs, and were positive outflows from the infratentorial networks (WM12→WM9/WM10, WM12/WM13/WM14→WM11) and negative outflows from the deep-layer network (WM10→WM14). In contrast, outflow connections from normal WMFNs showed significantly weaker strength in SCA3 compared with HCs, including positive influences (WM1, 3→WM10, WM5→WM8) and negative influences (WM4, WM8→WM9).

Figure 4 Between-group differences of the interactions and in/out strength of WMFNs. (A) Between-group differences of the interactions within WMFNs. The upper plots list the significantly enhanced and declined connections in SCA3 (FDR-corrected P<0.05). Red/blue color denote excitatory or inhibitory influence respectively. The lower diagrams summarized the altered interaction pattern in SCA3. Compared with HCs, patients with SCA3 showed hyper-interactions from dysfunctional WMFNs to normal WMFNs and within dysfunctional WMFNs. Meanwhile, the strength of outflow interactions from normal WMFNs to dysfunctional WMFNs and within normal WMFNs decreased in SCA3. (B) The between-group differences of in/out strength within 4 layers of the WMFNs and within normal and dysfunctional networks. Compared to the HCs, the out strength of deep-layer and infratentorial WMFNs were significantly increased in SCA3, while the out strength of superficial-layer WMFNs was significantly decreased. Asterisk indicates a significant difference (*, P<0.05, FDR corrected). WM1, the default-mode white matter network; WM2, the executive control white matter network; WM3, the temporo-motor white matter network; WM4, the somatomotor white matter network; WM5, the limbic white matter network; WM6, the visual white matter network; WM7, the dorsal frontoparietal white matter network; WM8, the forceps minor network; WM9, the internal capsule network; WM10, the deep white matter network; WM11, the deep occipital white matter network; WM12, the midbrain network; WM13, the cerebellar white matter network; WM14, the pons network. FDR, false discovery rate; GC, Granger causality; HCs, healthy controls; SCA3, spinocerebellar ataxia type 3; WMFNs, white matter functional networks.

Compared to the HCs, the out strength of deep-layer and infratentorial WMFNs was significantly increased in SCA3, whereas the out strength of superficial-layer WMFNs was significantly decreased (FDR-corrected P<0.05, Figure 4B).

Correlation analyses between WMFN characteristic and the SARA scores

Relationships between functional properties in WMFNs and the SARA scores are shown in Figure 5. The average amplitude of WM9 (the internal capsule network, FDR-corrected P=0.0476, r=−0.3206), WM12 (the midbrain network, FDR-corrected P=0.0492, r=−0.3039), and WM13 (the cerebellar WM network, FDR-corrected P=0.0248, r=−0.3671) were significantly negatively correlated with the SARA scores. Meanwhile, influences strength from these networks to WM4 (somatomotor WM network) were associated with the SARA scores: WM9→WM4 (r=−0.306, P=0.01), WM12→WM4 (r=−0.315, P=0.008), and WM 13→WM4 (r=−0.349, P=0.003).

Figure 5 Relationships between functional properties in WMFNs and the SARA scores. Significant correlations were observed between SARA scores and (A) amplitude and (B) GC strength within WMFNs. *, uncorrected P<0.05; **, FDR corrected P<0.05. FDR, false discovery rate; GC, Granger causality; SARA, Scale for the Assessment and Rating of Ataxia; WM4, somatomotor white matter network; WM9, internal capsule network; WM12, midbrain network; WM13, cerebellar white matter network; WMFNs, white matter functional networks.

WMV of WMFNs and its relationship with functional characteristics

Between-group differences of WMV of each WMFNs among SCA3 and HCs are shown in Table S1. The SCA3 group exhibited significantly reduced WMV in WM9 and infratentorial networks (WM12, WM13, and WM14) compared with the HC group (FDR-corrected P<0.05). Furthermore, the averaged amplitude values in WM9 (r=0.325, FDR-corrected P<0.001), WM10 (r=0.222, FDR-corrected P=0.02), WM12 (r=0.512, FDR-corrected P<0.001), WM13 (r=0.462, FDR-corrected P<0.001), and WM14 (r=0.384, FDR-corrected P<0.001) were significantly associated with corresponding WMV values. Meanwhile, the out strengths of WM12 (r=−0.37, FDR-corrected P<0.001), WM13 (r=−0.29, FDR-corrected P<0.001), and WM14 (r=−0.23, FDR-corrected P<0.001) were significantly correlated with WMV values. Detailed results are listed in Table 3.

Discussion

In this prospective and large-sample study, a comprehensive exploration was performed into the functional organization, neural activity, and inter-network causal interactions within the WMFNs in SCA3 and HCs, based on BOLD signals from WM. Identified 14 WMFNs that could be divided into 4 groups (superficial-, middle-, deep-layer, and infratentorial), with independent functional features of each group. Dysfunctional WMFNs in SCA3 were mainly infratentorial, middle-layer, and deep-layer, with significantly decreased amplitudes in comparison with HCs. In addition, the effective connectivity pattern within WMFNs in SCA3 was overall sparser than it was in HCs, whereas the directed connections from the dysfunctional WMFNs to the normal superficial-layer WMFNs and connections within the dysfunctional WMFNs were enhanced in SCA3. Concurrently, the normal WMFNs showed reduced out-strength of inter-network connections, whereas the dysfunctional WMFNs exhibited elevated out-strength. Furthermore, the declined neural activity and altered interactions observed can be partially attributed to the extent of WMV loss within the WMFNs, and are associated with the ataxia severity in SCA3.

In the current study, a total of 14 stable WMFNs consisting of four groups (superficial-, middle-, deep-layer, and infratentorial) were identified. The superficial-layer WMFNs presently displayed a tight correlation with their overlying GMFNs, and they showed obviously decreasing trend in amplitude with increased frequency. In contrast, the deep-layer WMFNs had a weak association with the GMFNs, and showed increasing trends in amplitude with frequency. Furthermore, the infratentorial WMFNs, which had also been defined as superficial-layer previously, showed visibly more moderate decreasing trends in amplitude than the superficial-layer WMFNs, similar to the middle-layer WMFNs. Considering their different feature of neuro-activity from the superficial-layer WMFNs, we divided them into a new group. Of note, the specific patterns of neuro-activity of each group were highly similar to previous studies (14,29,30), which enhanced the feasibility of the K-means clustering algorithm for clustering the WMFNs in SCA3.

Moreover, the hierarchical structure of WMFNs observed presently, identified through their distinct spectral characteristics, is consistent with previous research and may reflect the brain’s overall hierarchical organization of neural information processing (14). Superficial WMFNs are tightly coupled with cortical GM functional networks, suggesting that they function as critical integration hubs at the GM-WM interface, thereby playing a significant role in synchronizing neural information across different brain regions. Spatially, the distribution of superficial WM networks closely aligns with that of axonal projection pathways originating from GMNs, notably short U-fibers beneath the cerebral cortex, likely facilitating the integration of spontaneous cortical activity at rest. Middle-layer WMFNs, such as WM7 (dorsal frontoparietal network), comprise projection fibers including the corona radiata, whereas WM8 (forceps minor network) contains commissural fibers crossing through the genu of the corpus callosum to interconnect bilateral frontal regions. WM9 (internal capsule network) involves long association and projection fibers traversing the internal capsule. Functional activity within these networks likely facilitates communication across hemispheres and long-range interactions among distant brain areas. Deep-layer WM networks exhibit distinctive spectral patterns, characterized by increased neural activity amplitude at higher frequencies, suggesting their role as pivotal hubs in coordinating whole-brain information integration. Additionally, infratentorial WMFNs, composed predominantly of cerebellar pathways and their afferent/efferent connections with the brainstem and midbrain, reflect functional characteristics unique to cerebellar circuits. Collectively, this hierarchical model of WMFNs provides a novel framework for understanding the spectral specificity and functional roles of WM regions, emphasizing that neural activities at distinct frequency bands may support specialized functional roles and facilitate coordinated brain-wide communication through specific hierarchical WMFN levels.

Furthermore, compared with HCs, hypoactivity was shown in multiple WMFNs in SCA3, including WM6, WM7, and WM9–14. This distribution pattern of dysfunctional WMFNs in SCA3 follows a rostral-caudal distribution similar to that seen in WM microstructural lesions (4,5): low-activity WMFNs are mainly located in infratentorial, deep-layer, and occipital regions, whereas other superficial-layer as well as the frontal middle-layer WMFNs retain relatively preserved neural activity. These findings suggest that the functional abnormalities in patients with SCA3 were organized in a regionally specific manner across large-scale brain networks, indicating that the pathological progression of the disease may follow a characteristic neuroanatomical trajectory.

In addition, exploratory analyses indicated that abnormalities in neural activity within specific WMFNs were associated with structural atrophy. Currently, the relationship between WM structure and function remains unclear. Although some studies have reported significant associations between WM functional activation and structural characteristics (e.g., WM and fractional anisotropy), other studies have failed to observe clear structure-function correlations (31-33). At the microscopic level, WM predominantly consists of neuronal axons and serves as a critical pathway for signal transmission (34). Myelin sheaths, formed by oligodendrocytes, substantially enhance axonal conduction velocity (35). Furthermore, WM contains additional glial cells, including microglia and astrocytes. Microglia participate in immune surveillance and modulate neuroinflammatory responses, whereas astrocytes provide metabolic and oxygen support to axons and mediate neurovascular coupling, as previously demonstrated in GM (36-39). Collectively, these glial components provide the neurobiological basis for the generation and maintenance of BOLD signals in WM. Overall, our findings support the concept that WM BOLD signals have well-defined structural underpinnings, wherein the organization of fiber tracts and associated glial cells play critical neurophysiological roles in initiating and sustaining functional network activity. The present results highlight the intricate and interdependent relationship between WM structure and function.

To further understand the functional interactions within WMFNs in SCA3 and HCs, we adopted the GCA to quantify the macro information flow within WMFNs. It is notable that despite the overall sparser effective connection pattern observed in SCA3, interaction from dysfunctional WMFNs to superficial-layer WMFNs and within dysfunctional WMFNs in SCA3 showed increased absolute strength in comparison with HCs. This piques our awareness of the neural substrate of functional interactions within WM networks. Anatomically, WM tracts play a critical role in facilitating the transport of a substantial amount of functional data between spatially separated brain regions (34). Intricate interactions across WM networks facilitate the swift transfer of complex information between GM regions, thereby sustaining well-coordinated neural activities (40). In contrast, under neurodegenerative pathological or aging states, both WMFNs or GMNs interactions tend to degenerate from intricate connectomes to a simpler mode of communication (17,18,41). For instance, there are also sparser patterns of the effective connectomes within WMFNs in PD patients than HCs (17). These degenerations can be partly explained by the structural atrophy, yet they may also be attributed to the functional regulation within networks (42). Moreover, a common phenomenon observed in aging and neurodegenerative conditions is the weakening of connectivity within higher-order cognitive networks (such as those related to cognition and imagination) and the enhancement of connectivity within basic sensory-motor networks (such as visual and motor networks) (41-44). These alterations may serve to compensate for overall functional impairments and preserve basic functions necessary for maintaining activities of daily living, and have also been observed in SCA3 (45,46). For instance, a recently functional gradient study has shown the compressed principal gradient in SCA3, but with increased gradient scores between cerebellum and sensorimotor network (45). The compressed principal gradient suggests a dedifferentiation of the global GMN functions, whereas the increased gradient scores between cerebellum and sensorimotor network may indicate the compensation of the dysfunctional cerebellum in SCA3 (45). The notable similarity between functional gradient results and the WMFNs interactions enhanced the reliability of the present findings and may reflect a common pathological mechanism in SCA3. In summary, the present findings indicate that the disease-induced interaction pattern among WMFNs in SCA3 transitions from complex, balanced interplays into simple, dedifferentiated, and specific-enhanced circuits.

Additionally, both amplitudes in the cerebellar WM, the midbrain, and internal capsule networks and strength of influences from these WMFNs to somatomotor WM network were associated with the severity of ataxia symptoms, although the correlations between strength of influences and the SARA scores did not survive statistical correction and thus should be interpreted cautiously. In other words, the greater the ataxia symptoms, the worse the neural activity of these WMFNs, and the more pronounced their negative influences on the somatomotor WM network become. The WM fibers involved in these WMFNs are associated with motor functions physiologically and are known to be affected in SCA3 (8,47,48). Specifically, the damaged WM structure in cerebellar WM and brainstem in SCA3 patients has been consistently verified by previous studies, and was thought to be the key disease feature in SCA3 (2,48). Another recent study exploring the longitudinal changes of regional spontaneous activity in WM tracts has revealed the progressive impaired activity over time in the internal capsule in SCA3 (8). Anatomically, the pathway from the cerebellum through the midbrain and thalamus, then via the internal capsule to the motor cortex, constitutes the cerebellum’s principal ascending projection route (49,50). In this cerebellar–cerebral circuit, the cerebellum’s feedforward commands are conveyed via the midbrain, thalamus, and internal capsule to the sensorimotor cortex, enabling rapid and precise correction of the motor programs (51,52). Therefore, the dysfunction in the related WMFNs might reflect the impairment of the information processing between cerebellar-cerebral pathway, resulting in an increased demand for actual movement correction, which may present as the increased inflow influence from the affected WMFNs to somatomotor WM networks. These findings suggest that the alterations in the neural activity and interactions within WMFNs related to cerebellar-cerebral pathway may underlie ataxic symptoms in SCA3.

There are still limitations to the present study. Firstly, the cross-sectional nature of the data limits the ability to assess the changes in neural activity and inter-network causal interactions within the WMFNs throughout the progression of the disease. In addition, due to practical constraints, the present study did not include comparative analyses between preataxic and ataxic individuals. Such comparisons could provide novel insights into the temporal and causal relationship between structural degeneration and functional network abnormalities, thus representing an important direction for future research. Future longitudinal studies involving both preataxia and ataxia patients could investigate how neural activity and interactions within WMFNs in SCA3 change in different disease stages. Secondly, the current cohort consisted predominantly of patients in mild-to-moderate disease stages, as indicated by SARA scores lower than 20 points (total possible score: 0–40). This may partially explain why we predominantly observed enhanced dysfunctional interactions rather than extensive network disruption, potentially reflecting compensatory or maladaptive mechanisms occurring in earlier disease stages. Future research, including patients at advanced disease stages, is required to clarify the full spectrum and significance of WMFN abnormalities throughout the disease course. Thirdly, sub-scores for tremor, rigidity, bradykinesia, and posture were not included in our data, which prevented us from conducting a more detailed investigation into the associations between distinct motor symptoms and WMFN characteristics. Future work should take these into account and consider specific features in more detail. Finally, this research focused on primary functional changes within WMFNs; the issue of how WMFNs were influenced by GMFNs is also worthy of further investigation.

Conclusions

The present study provides a comprehensive exploration of the functional organization, neural activity, and inter-network causal interactions within the WMFNs in SCA3. The current findings suggest that the WM dysfunction in SCA3 is featured by widespread hypo-active WMFNs, overall impaired interactions, and enhanced outflow connections from dysfunctional WMFNs. These altered functional properties within WMFNs appear to be associated with underlying WM structural atrophy and are related to the SARA scores. In conclusion, the present study uncovered impaired neural activity and altered interactions among the WMFNs in SCA3 and their association with the ataxia severity, thereby indicating that the ataxia mechanism in SCA3 may also lie in WM functional abnormalities.

Acknowledgments

The authors thank all of participants for their involvement in this study.

Footnote

Reporting Checklist: The authors have completed the STROBE reporting checklist. Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-736/rc

Data Sharing Statement: Available at https://qims.amegroups.com/article/view/10.21037/qims-2025-736/dss

Funding: This study was supported by the National Natural Science Foundation of China (Nos. 82471984, 82071894, and 91959117), the China Postdoctoral Science Foundation (No. 2022TQ0377), National Natural Science Foundation of Hunan Province (Nos. 2023JJ30927, 2022JJ40820 and 2024JJ6687), and Natural Science Foundation of Changsha (No. KP2403036).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://qims.amegroups.com/article/view/10.21037/qims-2025-736/coif). All authors report that this study was supported by the National Natural Science Foundation of China (Nos. 82071984, 82071894, and 91959117), the China Postdoctoral Science Foundation (No. 2022TQ0377), National Natural Science Foundation of Hunan Province (Nos. 2023JJ30927, 2022JJ40820 and 2024JJ6687), and Natural Science Foundation of Changsha (No. KP2403036). The authors have no other conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. This prospective study was approved by the Medical Research Ethics Committee of Xiangya Hospital of Central South University (IBM No. 201601010). Written informed consent was obtained from all participants before participation.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Klockgether T, Mariotti C, Paulson HL. Spinocerebellar ataxia. Nat Rev Dis Primers 2019;5:24. [Crossref] [PubMed]
2. Rüb U, Schöls L, Paulson H, Auburger G, Kermer P, Jen JC, Seidel K, Korf HW, Deller T. Clinical features, neurogenetics and neuropathology of the polyglutamine spinocerebellar ataxias type 1, 2, 3, 6 and 7. Prog Neurobiol 2013;104:38-66. [Crossref] [PubMed]
3. Chandrasekaran J, Petit E, Park YW, du Montcel ST, Joers JM, Deelchand DK, et al. Clinically Meaningful Magnetic Resonance Endpoints Sensitive to Preataxic Spinocerebellar Ataxia Types 1 and 3. Ann Neurol 2023;93:686-701. [Crossref] [PubMed]
4. Qiu H, Wu C, Liang J, Hu M, Chen Y, Huang Z, Yang Z, Zhao J, Chu J. Structural alterations of spinocerebellar ataxias type 3: from pre-symptomatic to symptomatic stage. Eur Radiol 2023;33:2881-94. [Crossref] [PubMed]
5. Rezende TJR, de Paiva JLR, Martinez ARM, Lopes-Cendes I, Pedroso JL, Barsottini OGP, Cendes F, França MC Jr. Structural signature of SCA3: From presymptomatic to late disease stages. Ann Neurol 2018;84:401-8. [Crossref] [PubMed]
6. Piccinin CC, Rezende TJR, de Paiva JLR, Moysés PC, Martinez ARM, Cendes F, França MC Jr. A 5-Year Longitudinal Clinical and Magnetic Resonance Imaging Study in Spinocerebellar Ataxia Type 3. Mov Disord 2020;35:1679-84. [Crossref] [PubMed]
7. Rezende TJR, Petit E, Park YW, Tezenas du Montcel S, Joers JM, DuBois JM, et al. Sensitivity of Advanced Magnetic Resonance Imaging to Progression over Six Months in Early Spinocerebellar Ataxia. Mov Disord 2024;39:1856-67. [Crossref] [PubMed]
8. Tang J, Xing W, Wang D, Qin Y, Li J, Zhang Y, Yang F, Zhou G, Jiang H, Liao W. White matter functional and structural alterations of spinocerebellar ataxia type 3: A longitudinal MRI study. Neuroscience 2025;567:77-85. [Crossref] [PubMed]
9. Schilling KG, Li M, Rheault F, Gao Y, Cai L, Zhao Y, Xu L, Ding Z, Anderson AW, Landman BA, Gore JC. Whole-brain, gray, and white matter time-locked functional signal changes with simple tasks and model-free analysis. Proc Natl Acad Sci U S A 2023;120:e2219666120. [Crossref] [PubMed]
10. Gonzalez-Castillo J, Hoy CW, Handwerker DA, Roopchansingh V, Inati SJ, Saad ZS, Cox RW, Bandettini PA. Task Dependence, Tissue Specificity, and Spatial Distribution of Widespread Activations in Large Single-Subject Functional MRI Datasets at 7T. Cereb Cortex 2015;25:4667-77. [Crossref] [PubMed]
11. Wang H, Wang X, Wang Y, Zhang D, Yang Y, Zhou Y, Qiu B, Zhang P. White matter BOLD signals at 7 Tesla reveal visual field maps in optic radiation and vertical occipital fasciculus. Neuroimage 2023;269:119916. [Crossref] [PubMed]
12. Li J, Wu GR, Shi M, Xia J, Meng Y, Yang S, Chen H, Liao W. Spatiotemporal topological correspondence between blood oxygenation and glucose metabolism revealed by simultaneous fPET-fMRI in brain's white matter. Cereb Cortex 2023;33:9291-302. [Crossref] [PubMed]
13. Huang Y, Wei PH, Xu L, Chen D, Yang Y, Song W, Yi Y, Jia X, Wu G, Fan Q, Cui Z, Zhao G. Intracranial electrophysiological and structural basis of BOLD functional connectivity in human brain white matter. Nat Commun 2023;14:3414. [Crossref] [PubMed]
14. Peer M, Nitzan M, Bick AS, Levin N, Arzy S. Evidence for Functional Networks within the Human Brain's White Matter. J Neurosci 2017;37:6394-407. [Crossref] [PubMed]
15. Wen K, Zhao Y, Zhang F, Lui S, Kemp GJ, Gong Q. Large-scale dysfunctional white matter and grey matter networks in patients with social anxiety disorder. iScience 2022;25:105094. [Crossref] [PubMed]
16. Yang Y, Wang S, Liu J, Zou G, Jiang J, Jiang B, Cao W, Zou Q. Changes in white matter functional networks during wakefulness and sleep. Hum Brain Mapp 2022;43:4383-96. [Crossref] [PubMed]
17. Meng L, Wang H, Zou T, Wang X, Chen H, Xie F, Li R. Attenuated brain white matter functional network interactions in Parkinson's disease. Hum Brain Mapp 2022;43:4567-79. [Crossref] [PubMed]
18. Xu L, Zhao Y, Choi S, Li M, Schilling KG, Zu Z, Rogers BP, Ding Z, Anderson AW, Landman BA, Gore JC, Gao YAlzheimer's Disease Neuroimaging Initiative. Reductions in the white-gray functional connectome in preclinical Alzheimer's disease and their associations with amyloid and cognition. Alzheimers Dement 2024;20:8317-30. [Crossref] [PubMed]
19. Gao Y, Zhao Y, Li M, Lawless RD, Schilling KG, Xu L, Shafer AT, Beason-Held LL, Resnick SM, Rogers BP, Ding Z, Anderson AW, Landman BA, Gore JCAlzheimer's Disease Neuroimaging Initiative. Functional alterations in bipartite network of white and grey matters during aging. Neuroimage 2023;278:120277. [Crossref] [PubMed]
20. Yan CG, Wang XD, Zuo XN, Zang YF. DPABI: Data Processing & Analysis for (Resting-State) Brain Imaging. Neuroinformatics 2016;14:339-51. [Crossref] [PubMed]
21. Mak HK, Zhang Z, Yau KK, Zhang L, Chan Q, Chu LW. Efficacy of voxel-based morphometry with DARTEL and standard registration as imaging biomarkers in Alzheimer's disease patients and cognitively normal older adults at 3.0 Tesla MR imaging. J Alzheimers Dis 2011;23:655-64. [Crossref] [PubMed]
22. Power JD, Barnes KA, Snyder AZ, Schlaggar BL, Petersen SE. Spurious but systematic correlations in functional connectivity MRI networks arise from subject motion. Neuroimage 2012;59:2142-54. [Crossref] [PubMed]
23. Yeo BT, Krienen FM, Sepulcre J, Sabuncu MR, Lashkari D, Hollinshead M, Roffman JL, Smoller JW, Zöllei L, Polimeni JR, Fischl B, Liu H, Buckner RL. The organization of the human cerebral cortex estimated by intrinsic functional connectivity. J Neurophysiol 2011;106:1125-65. [Crossref] [PubMed]
24. Yang H, Long XY, Yang Y, Yan H, Zhu CZ, Zhou XP, Zang YF, Gong QY. Amplitude of low frequency fluctuation within visual areas revealed by resting-state functional MRI. Neuroimage 2007;36:144-52. [Crossref] [PubMed]
25. Zuo XN, Di Martino A, Kelly C, Shehzad ZE, Gee DG, Klein DF, Castellanos FX, Biswal BB, Milham MP. The oscillating brain: complex and reliable. Neuroimage 2010;49:1432-45. [Crossref] [PubMed]
26. Chen G, Hamilton JP, Thomason ME, Gotlib IH, Saad ZS, Cox RW. Granger Causality via Vector AutoRegression Tuned for FMRI Data Analysis. Proc Int Soc Mag Reson Med 2009;17:1718.
27. Fan YS, Li Z, Duan X, Xiao J, Guo X, Han S, Guo J, Yang S, Li J, Cui Q, Liao W, Chen H. Impaired interactions among white-matter functional networks in antipsychotic-naive first-episode schizophrenia. Hum Brain Mapp 2020;41:230-40. [Crossref] [PubMed]
28. Salvador R, Suckling J, Coleman MR, Pickard JD, Menon D, Bullmore E. Neurophysiological architecture of functional magnetic resonance images of human brain. Cereb Cortex 2005;15:1332-42. [Crossref] [PubMed]
29. Bu X, Gao Y, Liang K, Chen Y, Guo L, Huang X. Investigation of white matter functional networks underlying different behavioral profiles in attention-deficit/hyperactivity disorder. Psychoradiology 2022;2:69-77. [Crossref] [PubMed]
30. Wang J, Xue T, Song D, Dong F, Cheng Y, Wang J, Ma Y, Zou M, Ding S, Tao Z, Xin W, Yu D, Yuan K. Investigation of white matter functional networks in young smokers. Neuroimage 2024;303:120917. [Crossref] [PubMed]
31. Ji GJ, Liao W, Chen FF, Zhang L, Wang K. Low-frequency blood oxygen level-dependent fluctuations in the brain white matter: more than just noise. Sci Bull (Beijing) 2017;62:656-7. [Crossref] [PubMed]
32. Jiang Y, Duan M, Li X, Huang H, Zhao G, Li X, Li S, Song X, He H, Yao D, Luo C. Function-structure coupling: White matter functional magnetic resonance imaging hyper-activation associates with structural integrity reductions in schizophrenia. Hum Brain Mapp 2021;42:4022-34. [Crossref] [PubMed]
33. Jiang Y, Yao D, Zhou J, Tan Y, Huang H, Wang M, Chang X, Duan M, Luo C. Characteristics of disrupted topological organization in white matter functional connectome in schizophrenia. Psychol Med 2022;52:1333-43. [Crossref] [PubMed]
34. Fields RD. White matter matters. Sci Am 2008;298:42-9.
35. Dutta DJ, Woo DH, Lee PR, Pajevic S, Bukalo O, Huffman WC, Wake H, Basser PJ. SheikhBahaei S, Lazarevic V, Smith JC, Fields RD. Regulation of myelin structure and conduction velocity by perinodal astrocytes. Proc Natl Acad Sci U S A 2018;115:11832-7. [Crossref] [PubMed]
36. Lee J, Hamanaka G, Lo EH, Arai K. Heterogeneity of microglia and their differential roles in white matter pathology. CNS Neurosci Ther 2019;25:1290-8. [Crossref] [PubMed]
37. Lundgaard I, Osório MJ, Kress BT, Sanggaard S, Nedergaard M. White matter astrocytes in health and disease. Neuroscience 2014;276:161-73. [Crossref] [PubMed]
38. Lia A, Di Spiezio A, Speggiorin M, Zonta M. Two decades of astrocytes in neurovascular coupling. Front Netw Physiol 2023;3:1162757. [Crossref] [PubMed]
39. Gore JC, Li M, Gao Y, Wu TL, Schilling KG, Huang Y, Mishra A, Newton AT, Rogers BP, Chen LM, Anderson AW, Ding Z. Functional MRI and resting state connectivity in white matter - a mini-review. Magn Reson Imaging 2019;63:1-11. [Crossref] [PubMed]
40. Schmahmann JD, Pandya DN. Fiber Pathways of the Brain. Oxford: Oxford University; 2009.
41. Ferreira LK, Busatto GF. Resting-state functional connectivity in normal brain aging. Neurosci Biobehav Rev 2013;37:384-400. [Crossref] [PubMed]
42. Sala-Llonch R, Bartrés-Faz D, Junqué C. Reorganization of brain networks in aging: a review of functional connectivity studies. Front Psychol 2015;6:663. [Crossref] [PubMed]
43. Wang L, Zhang Y, Zhang J, Sang L, Li P, Yan R, Qiu M, Liu C. Aging Changes Effective Connectivity of Motor Networks During Motor Execution and Motor Imagery. Front Aging Neurosci 2019;11:312. [Crossref] [PubMed]
44. Guo S, Feng L, Ding R, Long S, Yang H, Gong X, Lu J, Yao D. Functional gradients in prefrontal regions and somatomotor networks reflect the effect of music training experience on cognitive aging. Cereb Cortex 2023;33:7506-17. [Crossref] [PubMed]
45. Liu X, Guo J, Jiang Z, Liu X, Chen H, Zhang Y, Wang J, Liu C, Gao Q, Chen H. Compressed cerebellar functional connectome hierarchy in spinocerebellar ataxia type 3. Hum Brain Mapp 2024;45:e26624. [Crossref] [PubMed]
46. Wang X, Chen H, Wen R, Ou P, Huang Y, Deng L, Shi L, Chen W, Chen H, Wang J, He C, Liu C. Exploring functional and structural connectivity disruptions in spinocerebellar ataxia type 3: Insights from gradient analysis. CNS Neurosci Ther 2024;30:e14842. [Crossref] [PubMed]
47. Inada BSY, Rezende TJR, Pereira FV, Garcia LÁL, da Rocha AJ, Neto PB, Barsottini OGP, França MC Jr, Pedroso JL. Corticospinal tract involvement in spinocerebellar ataxia type 3: a diffusion tensor imaging study. Neuroradiology 2021;63:217-24. [Crossref] [PubMed]
48. Faber J, Berger M, Wilke C, Hubener-Schmid J, Schaprian T, Santana MM, et al. Stage-Dependent Biomarker Changes in Spinocerebellar Ataxia Type 3. Ann Neurol 2024;95:400-6. [Crossref] [PubMed]
49. Schmahmann JD. The cerebrocerebellar system: anatomic substrates of the cerebellar contribution to cognition and emotion. International Review of Psychiatry 2001;13:247-60.
50. Kelly RM, Strick PL. Cerebellar loops with motor cortex and prefrontal cortex of a nonhuman primate. J Neurosci 2003;23:8432-44. [Crossref] [PubMed]
51. Friston K, FitzGerald T, Rigoli F, Schwartenbeck P, Pezzulo G. Active Inference: A Process Theory. Neural Comput 2017;29:1-49. [Crossref] [PubMed]
52. Hutchinson JB, Barrett LF. The power of predictions: An emerging paradigm for psychological research. Curr Dir Psychol Sci 2019;28:280-91. [Crossref] [PubMed]