Effect of deep brain stimulation on brain network and white matter integrity in Parkinson's disease

Abstract Aims The effects of subthalamic nucleus (STN)‐deep brain stimulation (DBS) on brain topological metrics, functional connectivity (FC), and white matter integrity were studied in levodopa‐treated Parkinson’s disease (PD) patients before and after DBS. Methods Clinical assessment, resting‐state functional MRI (rs‐fMRI), and diffusion tensor imaging (DTI) were performed pre‐ and post‐DBS in 15 PD patients, using a within‐subject design. The rs‐fMRI identified brain network topological metric and FC changes using graph‐theory‐ and seed‐based methods. White matter integrity was determined by DTI and tract‐based spatial statistics. Results Unified Parkinson's Disease Rating Scale III (UPDRS‐ III) scores were significantly improved by 35.3% (p < 0.01) after DBS in PD patients, compared with pre‐DBS patients without medication. Post‐DBS PD patients showed a significant decrease in the graph‐theory‐based degree and cost in the middle temporal gyrus and temporo‐occipital part‐Right. Changes in FC were seen in four brain regions, and a decrease in white matter integrity was seen in the left anterior corona radiata. The topological metrics changes were correlated with Beck Depression Inventory II (BDI‐II) and the FC changes with UPDRS‐III scores. Conclusion STN‐DBS modulated graph‐theoretical metrics, FC, and white matter integrity. Brain connectivity changes observed with multi‐modal imaging were also associated with postoperative clinical improvement. These findings suggest that the effects of STN‐DBS are caused by brain network alterations.

cognition. 9,10 The mechanisms underlying the DBS-induced amelioration of PD motor dysfunction and the effects of the combination of DBS with L-DOPA on the neuronal network are elusive.
Resting-state functional magnetic resonance imaging (rs-fMRI) is widely used to study abnormal patterns of functional connectivity (FC) in PD patients. Functional changes in the cerebello-thalamocortical circuit are a hallmark of PD and have been associated with the major motor symptoms of the disease. 11,12 Decreased connectivity in the posterior putamen is the most consistent functional alteration associated with motor impairment, 13 while the FC of the striatum with the cerebellum is reduced. 11 Pathological FC also affects the cerebellothalamic and basal ganglia circuits in tremor-dominant PD patients. 14 Moreover, increased connectivity between the STN and the sensorimotor cortex is found in levodopa-treated PD patients. 15 Some studies that investigated the effect of DBS on brain FC demonstrated that it could increase functional brain connectivity within the cerebello-thalamo-cortical network 2 and modulate effective connectivity within the cortico-striato-thalamo-cortical loop. 16 Another study showed that STN-DBS ameliorates PD symptoms through normalization of the human functional connectome in PD. 17 Most research focused on motor circuits in PD patients. 2,16 The brain areas mainly involved in these studies are the primary motor cortex (M1), 16,18,19 supplementary motor area (SMA), 15,19 premotor cortex, 20,21 basal ganglia, thalamus, caudal SMA, 19 cerebellum, and putamen. 22 However, the effect of STN-DBS on brain FC remains poorly understood, and little is known of its impact on the whole brain.
Graph theory is a powerful tool for characterizing the global topological organization of brain networks [23][24][25] and for investigating abnormal functional brain networks in different stages of PD. 26 A previous study found that PD patients exhibited lower global efficiency, higher clustering coefficients, and higher characteristic path lengths than healthy controls. 26 Early-stage drug-naïve PD patients exhibited disruption of the whole-brain topological organization (ie, decreased functional segregation and integration) of functional brain networks, 27 which may contribute to preclinical changes in the cognitive process. 28 At the node and connection level, PD patients exhibited reduced lengths of node centralities and connectivity, mainly not only in the temporal-occipital regions, but also in the sensorimotor regions. 28 No significant differences in intransitivity, characteristic path length, and degree were observed in the motor network related to the basal ganglia and cerebellum in PD patients, nor in the related network-level values compared with normal controls. 29 Levodopa modulates the global and local efficiency measures of small-world topology in PD patients. 30 However, the effect of STN-DBS treatment on brain network graph theory in PD patients remains unclear. Diffusion tensor imaging (DTI) allows the measurement of fractional anisotropy and similarity. Previous studies on the use of DTI in PD have demonstrated abnormal fractional anisotropy in multiple white matter regions, particularly in the dopaminergic nuclei and pathways. 31,32 Hence, it has been used to assess both disease progression and the effects of treatment. 5,33 In the last few years, DTI has been applied for preoperative target localization in DBS surgery. 34,35 A recent study reported that the fibers connecting the electrode with the left prefrontal areas were associated with a worsening of depressive symptoms with STN-DBS, 8 suggesting that DTI could provide an additional means of assessing the evolution of psychiatric symptoms after surgery. 8 Most rs-fMRI research on STN-DBS has focused on motor and non-motor circuits in PD patients. The effects of STN-DBS on brain FC and the topological metrics of graph theory in PD patients are unknown. Additionally, the impact of DBS on the fiber tract connectivity (integrity) profile of PD patients is unclear. We hypothesized that the benefit derived from STN-DBS in PD relies on distributed brain networks and anatomical connections. Consequently, this study aimed to use multi-modal imaging to compare the brain topological metrics, FC, and white matter integrity in levodopa-treated PD patients' pre-and post-DBS.

| Subjects
From December 2016 to July 2019, 15 PD patients who had undergone STN-DBS, had been followed-up for at least 4 months, and showed significant motor improvement from the DBS were enrolled in this study (nine men and six women, mean age at DBS surgery: 58.

| Clinical assessments
The UPDRS was used to evaluate the patient's mentation, behav-  We performed quality checking procedures before analyzing images. When patients exhibited head motion of more than 1.5 mm in any of the x, y, or z directions, more than 1.5° of any angular dimensions were discarded.

| Resting-state functional MRI data FC and graph-theory analysis
The correlation of FC networks and topological metrics was ana-

| Statistical analysis
All data were tested to ensure they were normally distributed. If a group of data did not exhibit a normal distribution, a Student t test or one-way ANOVA test was applied with a non-parametric equivalent. We use repeated measure analysis of variance to test UPDRS-III scores, including factors such as treatment method Whole-brain FA and MD images were compared with pre-DBS and post-DBS differences. The FA and MD were compared between the pre-and post-DBS conditions with paired t tests. The significance threshold was set at p < 0.05, FDR-corrected for multiple comparisons performed by permutation test with TFCE.

| Effect of DBS on motor disability and nonmotor symptoms of cognition and language
Deep brain stimulation on (DBS ON ) alone improved motor disability significantly, by 35.3% (UPDRS part III, Figure 1A). The LEDD was significantly decreased by 56.0% in the post-DBS compared with the pre-DBS state ( Figure 1B). In the medication-on state, there was no significant difference between pre-DBS and post-DBS ON in UPDRS-III (21.7 ± 6.7% vs. 24.9 ± 6.6%, respectively). Non-motor symptoms of depression were significantly improved after DBS (BDI-II, pre-DBS/post-DBS: 19.9 ± 16.5/12.1 ± 6.5, p < 0.05; Figure 1C). However, cognition (CASI,  Figure 1D, E) were not significantly changed after DBS, although the language showed a trend of deterioration.

| Graph theory at pre-DBS and post-DBS and correlation with clinical changes
In graph theory, the cost and degree were significantly decreased  Figure 3A, B). Figure 4 and Table 2 show four principal increases and four critical decreases in brain FC. Brain

| Diffusion tensor imaging TBSS in patients at pre-and post-DBS
We pre-DBS condition ( Figure 5, Table 3). The MD and radial and axial diffusivities of the patients in the post-DBS state showed no difference compared with those in the pre-DBS state.

| DISCUSS ION
Recent concepts of neurodegenerative disease have emphasized the importance of specific changes in brain connectivity in pathophysiology and effectiveness of treatment. 17,31,[42][43][44] In this study, we showed that "an increased connectivity of motor hubs with the brain stem and putamen" indicates that both the brain stem and putamen play a crucial role in the effects of STN-DBS on motor improvement in PD patients. 2,11,20 The FC changes in toMTG-R-aSTG L and SMA L-Networks-FrontoParietal-PPC L are correlated with motor improvement. Furthermore, "a major decrease in the degree and cost in the toMTG-R" indicated that connection density and centrality were decreased in the toMTG-R and correlated with depression improvement, which could be related to the deterioration in language and speech function in PD patients after DBS. 45 We also found a trend of cognition decline in PD patients after DBS, which could be observed in terms of topological metrics 46 and that the decreased FA in the white matter tract was more likely to be related to the emotion and executive attention decline in these patients. 47 In Figure 6, we schematically summarize the critical results of the study.

| Post-DBS motor improvement and FC
In line with the improved motor control and reduced LEDD that is widely noted in PD patients after DBS, 5 we also found significant improvements in the UPDRS-III motor score and a reduction in LEDD after DBS.

| Depression improvement and topological metrics: The toMTG-R and related findings
The cost and degree of topological metrics in toMTG-R were decreased after STN-DBS and were correlated with an improvement in depression as interpreted by the Beck Depression Intervention (BDI-Ⅱ) score ( Figure 3A,B). Ye et al. 48 used rs-fMRI and graph theory for analyzing patients with major depressive disorder and found that patients had higher node centralities than normal control group.
Our result showed STN-DBS improves patients' depression and decreased these pathologically higher parameters through analyzing graph-theory metrics. These findings indicate that the toMTG-R may be a biomarker for depression improvement when using topological metrics as diagnostic tools. Previous studies have demonstrated a gradual decline in cognitive function after long-term DBS. 3,49 The toMTG are related to higher-order cognitive function in visual perception. 50 Although our results showed a mild decline in cognitive function, the difference was not statistically significant. This may be due to the shorter follow-ups of STN-DBS.

| Deterioration in language and speech function and FC
Clinically, the worsening of PD symptoms, including akinesia, speech, postural stability, freezing of gait, and cognitive function, may occur after DBS. 3 We showed a decline in language and speech function, although this was not statistically significant, and may be attributed to the relatively short-term follow-up period.
Nevertheless, changes in FC may reflect the deterioration in their earlier stage.
While FC decreased in post-DBS PD patients in four brain regions, including the pMTG-R, toMTG-R, pPaHC-R, and Networks, language pSTG-R, and particularly the first three regions, may be a central hub in the decrease in FC after DBS. The MTG is related to language inputs. The pMTG is related to language 51 and semantic memory processing. 50 The IFG (Broca area) is related to language processing, speech production, and executive function. 50 Figure 5). The ACR is part of the limbic-thalamo-cortical circuitry associated with impaired top-down emotion regulation systems and the motor pathway. 47,57,58 It is also associated with executive attention network functions. 59  Using the rs-fMRI and behavioral assessments to explore the effects of DBS on emotion and executive attention is, therefore, both informative and valuable.
There are several limitations to this study. First, a major issue in fMRI with DBS is the occurrence of geometric distortions and drop-out of the EPI signal in the vicinity of the electrodes, particularly near the skull where electrodes are connected to the extension leads; these problems are currently unavoidable. 17 Furthermore, the impact of DBS-induced artifacts on the rs-fMRI signal has not been investigated in detail, and more methodological work is required to address potential issues. Second, according to our objective, we needed to turn DBS off 10 min before the scan due to safety concerns. 61 We defined this post-DBS ON/MedON state as the highest "on" state, in which DBS was switched on and the patient was on medications. Moreover, our included patients remained in their "on" state during the scan. This sustained effect observed after DBS was turned off could be due to synaptic plasticity within the stimulated neural network. 45 Third, the results could also be confounded by pharmacodynamic changes produced by the significant reduction in medication. Lastly, patients had different stimulation parameters, which could cause diversity of results. Furthermore, the length of the follow-up times after DBS was not uniform, making it difficult to compare the result with other studies.
In conclusion, we demonstrated that STN-DBS alters graphtheoretical metrics, FC, and white matter integrity, leading to a significant improvement in the motor and psychiatric symptoms of patients with PD. The changes in brain connectivity from this multimodal imaging were also associated with the extent of postoperative clinical improvement. These results suggest that changes in brain networks might explain the benefit of STN-DBS in PD.

ACK N OWLED G M ENTS
The authors are grateful to the Parkinson's disease patients who

CO N FLI C T O F I NTE R E S T
All authors declare that they have no competing interests.

AUTH O R CO NTR I B UTI O N S
Li-Chuan Huang performed the experiments, analyzed the rs-fMRI data, and wrote the manuscript. Li-Guo Chen analyzed the DTI data.
Ping-An Wu, Cheng-Yoong Pang, and Shinn-Zong Lin contributed conception and design of the study. Sheng-Tzung Tsai was responsible for the study design, data review, editing, and revision of the manuscript.
Shin-Yuan Chen reviewed and revised the manuscript and was responsible for the supervision of the study concept and design, critical revision of the manuscript, and study supervision. All authors contributed to manuscript revision and read and approved the submitted version.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.