Drooling disrupts the brain functional connectivity network in Parkinson's disease

Abstract Aims This study aimed to investigate the causal interaction between significant sensorimotor network (SMN) regions and other brain regions in Parkinson's disease patients with drooling (droolers). Methods Twenty‐one droolers, 22 PD patients without drooling (non‐droolers), and 22 matched healthy controls underwent 3T‐MRI resting‐state scans. We performed independent component analysis and Granger causality analysis to determine whether significant SMN regions help predict other brain areas. Pearson's correlation was computed between imaging characteristics and clinical characteristics. ROC curves were plotted to assess the diagnostic performance of effective connectivity (EC). Results Compared with non‐droolers and healthy controls, droolers showed abnormal EC of the right caudate nucleus (CAU.R) and right postcentral gyrus to extensive brain regions. In droolers, increased EC from the CAU.R to the right middle temporal gyrus was positively correlated with MDS‐UPDRS, MDS‐UPDRS II, NMSS, and HAMD scores; increased EC from the right inferior parietal lobe to CAU.R was positively correlated with MDS‐UPDRS score. ROC curve analysis showed that these abnormal ECs are of great significance in diagnosing drooling in PD. Conclusion This study identified that PD patients with drooling have abnormal EC in the cortico‐limbic‐striatal‐cerebellar and cortio‐cortical networks, which could be potential biomarkers for drooling in PD.


| INTRODUC TI ON
Drooling, also known as sialorrhea or ptyalism, is defined as excessive saliva that exceeds the edge of the lips. It is a devastating and debilitating complication of Parkinson's disease (PD) that can be caused by excessive saliva production or swallowing dysfunction. 1 Drooling can occur at any stage of PD, with a prevalence ranging from 9.26 to 70%. 2 It may cause feelings, such as embarrassment, malodor, and skin infection, and further lead to a lack of self-confidence, depression, and social isolation for PD patients. In addition to seriously reducing the quality of life of PD patients, these symptoms place a great burden on caregivers. More seriously, the accumulation of secretions in the oropharynx of PD patients considerably increases the risk of aspiration pneumonia, thereby contributing to the morbidity and mortality of aspiration pneumonia in PD patients. 3 Numerous researchers have recently focused on the prevalence of drooling, associated clinical symptoms, and treatment. 2,4 However, the cerebral mechanisms of drooling in PD remain unclear.
Recently, resting-state functional MRI (rsfMRI) has emerged as a novel noninvasive tool that greatly contributes to capturing changes in PD-related functional connectivity (FC) without requiring subjects to perform specific tasks. 5 This not only promotes a deeper understanding of PD pathogenesis but also further demonstrates the significant clinical value of neuroimaging biomarkers in PD diagnosis and monitoring. To date, only one rsfMRI study has explored the aberrant FC of the putamen within the bilateral sensorimotor cortices, superior and inferior parietal lobules, and areas in the right occipital and temporal lobes in early-stage PD patients with drooling (PD-DR; droolers). 6 Therefore, we are still far from fully understanding the functional changes in the brain associated with drooling in PD.
Swallowing saliva is a complex process. The somatosensory cortex plays an important role in the execution of swallowing as well as the integration of sensorimotor information related to preparation for swallowing. 7 The major cause of drooling in PD patients is swallowing dysfunction in the oral and/or pharyngeal phase of swallowing, which is also a manifestation of motor difficulties in PD. 4 It is reported that motor difficulties in PD are attributed to sensorimotor dysfunction caused by denervation along the cortex-striatum pathway following nigrostriatal dopaminergic loss. [8][9][10] Thus, the occurrence of drooling in PD patients is closely associated with sensorimotor areas of the brain. The sensorimotor network (SMN) consists of the primary sensorimotor cortex, secondary somatosensory cortices, and supplementary motor area. 5 It is crucial for detecting and processing sensory input as well as preparing and executing motor actions. Research has shown that FC within the SMN (premotor cortex, supplementary motor area, and primary motor cortex) is significantly increased in PD patients with dysphagia compared to those without dysphagia. 11 These findings suggest that altered FC in the SMN may be responsible for the occurrence of drooling in PD.
The present study aimed to investigate alterations in the SMN between droolers and non-droolers and the causal relationship between significant SMN regions and other brain regions. To extract the SMN, we employed independent component analysis (ICA), a mathematical technique that maximizes statistical independence among its components. In contrast to seed-based methods, the ICA approach proceeds without a priori seed selection, thus removing the bias from analysis. 12 Additionally, we performed Granger causality analysis (GCA) using the significant SMN regions as seeds to determine whether the SMN time series could be useful for predicting other brain areas. The GCA's capacity to predict the direction of information flow in brain networks has led to its widespread application. 13 We hypothesized that: (1) droolers would show abnormal intrinsic connectivity within the SMN relative to PD patients without drooling (PD-NDR; non-droolers) and (2) effective connectivity (EC) between significant SMN regions as seeds and other brain regions would be further detected, which correlates with clinical scores.

| Study design
Item number 2 of the Movement Disorder Society-Unified Parkinson's Disease Rating Scale (MDS-UPDRS) part II was used to evaluate drooling in PD. Patients with a score ≥3 were classified as "droolers," and those with a score <1 were classified as "non-droolers." Thus, all enrolled participants were divided into three groups, PD-DR (n = 21), PD-NDR (n = 22), and HCs (n = 22), and underwent clinical data collection in the "off" medication state after MRI scanning (Supplemental Materials). Two experienced neurologists blinded to the MRI data performed clinical examinations and recorded clinical data.

| fMRI data acquisition
MRI scans were performed using a 3.0-Tesla MR imaging system (MAGNETOM Prisma; Siemens Healthcare, Erlangen, Germany) with a 64-channel receiver array head coil. Functional imaging data based on BOLD were acquired using a gradient echo-planar imaging (EPI) sequence. The sequence parameters are depicted in the Supplemental Materials.

| Data preprocessing
The rsfMRI data were preprocessed using the Graph Theoretical Network Analysis Toolbox version 2.0 (GRETNA v2.0, http://www. nitrc.org/proje cts/gretn a/p/). The specific procedure is described in the Supplemental Materials.

| Independent component analysis
To identify the SMN in all PD patients, group spatial ICA was per-

| Granger causality analysis
The right caudate nucleus (CAU.R) and right postcentral gyrus (PoCG.R) were used as ROIs for GCA analysis. We further performed time-series-relevant GCA using REST software 15 to investigate the causal interaction between two ROIs and each voxel in the entire brain. We provide a brief description of GCA analysis in the Supplemental Materials.

| Statistical analysis
The demographic and clinical data were analyzed using SPSS ver-   analysis of variance was conducted on all continuous variables (p < 0.05).
We used REST to analyze the EC differences between two ROIs (CAU.R and PoCG.R) and other brain regions. Two-tailed t-tests were performed to identify between-group differences in the significant EC, with age, gender, and education years as covariates. A cluster threshold of p < 0.01 with AlphaSim correction (single voxel p < 0.01, cluster size >40 voxels) was set.
The Additionally, receiver operating characteristic (ROC) curves were plotted to assess the diagnostic performance of EC in droolers, and Youden's J parameter was measured to find the optimum threshold.

| Demographic and clinical characteristics
Sixty-five subjects were included in this study, including 21 droolers,  Table 1.

| ICA results
According to the ICA analysis, we observed the SMN in the PD-DR and PD-NDR groups ( Figure 1A). Compared to non-droolers, droolers showed significantly increased FC values in the CAU.R and significantly reduced FC values in the PoCG.R ( Figure 1B). These two brain regions within the SMN extracted from the ICA were selected as ROIs for further ROIwise GCA investigation. The corresponding ROIs are listed in Table 2. CAU.R to the left lingual gyrus (LING.L) and right cerebellum (CB.R) ( Figure 2A, Table 3). However, significantly increased EC from the CAU.R to the MTG.R and right inferior parietal lobe (IPL.R) and significantly decreased EC from the CAU.R to the left anterior cingulate and paracingulate gyri (ACG.L) were found in droolers compared with non-droolers ( Figure 2B, Table 3). Additionally, non-droolers showed increased EC from the CAU.R to the right anterior cingulate and paracingulate gyri (ACG.R) and decreased EC from the CAU.R to the left olfactory cortex (OLF.L) relative to healthy controls ( Figure 2C, Table 3).

| EC to the CAU.R
Compared to healthy controls, we found significantly weakened EC from the STG.R to the CAU.R in both non-droolers and droolers ( Figure 3a,c, Table 4). Compared with non-droolers, droolers showed significantly strengthened EC from the left cerebellum (CB.L) to the CAU.R; but weakened EC from the IPL.R to the CAU.R (Figure 3b, Table 4).

| EC from the PoCG.R
For the causal connectivity from the PoCG.R to other brain regions, droolers exhibited significantly increased causal connectivity from the PoCG.R to the left precentral gyrus (PreCG.L) and left putamen (PUT.L) relative to healthy controls ( Figure 4A, Table 5). Additionally, droolers showed significantly decreased causal connectivity from the PoCG.R to the MTG.R compared to non-droolers ( Figure 4B,  Figure 5A,C, Table 6). Compared with non-droolers, droolers showed significantly decreased inflow from the right Rolandic operculum (ROL.R) to the PoCG.R ( Figure 5B, Table 6).

| Correlations between abnormal connectivity and clinical characteristics
Pearson correlation coefficients were used to examine the asso-  Figure 6F). After correction for age, sex, and education, significant correlations remained (Table S1). There were no significant associations between abnormal connectivity and these clinical characteristics after the Bonferroni correction. Disrupted EC was not correlated with the NMSS-19 domain, SCS-PD, MMSE, or HAMA scores.  Figure S1). The area under the ROC curve (AUC), 95% confidence interval (CI), p value, optimal cutoff value, sensitivity, specificity, and

| DISCUSS ION
By comparing a cohort of PD patients with and without drooling and healthy controls using rsfMRI, we observed abnormal connectivity in the cortico-striatal, cerebellar-striatal, limbic-striatal, and cortiocortical loops in droolers. Increased inflow was positively correlated with the severity of motor symptoms, nonmotor symptoms, and depression in droolers. These abnormal directional connectivities may be of great significance in the differential diagnosis of drooling in PD.
We speculated that drooling disrupted the functional integration of the cortico-limbic-striatal-cerebellar and cortio-cortical networks, resulting in motor function and neuropsychological deficits in PD patients with drooling. These findings provide novel insight into the large-scale functional reorganization of drooling in PD.
Drooling is a devastating and debilitating complication of PD and other neurological diseases including cerebral palsy and amyotrophic lateral sclerosis. Due to dopamine deficiency, an accumulating body of evidence has indicated that PD patients produce less saliva than healthy controls. 16,17 Consequently, decreased salivary clearance due to oropharyngeal bradykinesia may be the major cause of drooling in PD patients. 18

Convergent evidence from prior observations has implicated
the striatum in the pathophysiology of drooling in PD. 6,20 A 123 I-FP-CIT SPECT imaging study revealed that drooling in PD was associated with reduced striatal DAT availability. 20 In priori ROI analyses, PD patients complaining of drooling exhibited functional disconnection between the putamen and cortical, with remarkable alterations, which indicated that interrupt FC in the cortico-striatal loop may be a cardinal contributor to the "drooling network" in PD. 6 As an essential component of the dorsal striatum structure, the caudate nucleus together with the putamen form the primary input nucleus of the basal ganglia in the cortico-striatal dopaminergic circuitry. However, Hou et al.

TA B L E 3
Effective connections from the right caudate nucleus among the PD-DR, PD-NDR, and HCs groups.    45 These results demonstrate that droolers may experience dysarthria or swallowing difficulty due to alterations in ROL function.

PD-DR > HCs
Several limitations of the present study should be acknowledged.
First, this study was a cross-sectional study and no follow-up study was conducted to analyze the dynamic changes in brain functions in droolers and non-droolers. Second, the sample size of the study was not large, and there may be statistical bias. Third, although all subjects stopped taking anti-PD drugs for at least 12 h before undergoing an imaging scan, it is not fully determined whether long-term drug use could potentially affect the experimental results.
In conclusion, this study discovered that PD patients with drooling have changes in the cortico-limbic-striatal-cerebellar and cortiocortical networks, which might reflect that droolers have abnormal brain integration functions, neurological, and downregulation or compensation mechanisms in specific areas. Drooling disrupts the information flows in the cortico-striatal, cerebellar-striatal, limbicstriatal, and cortio-cortical loops of PD patients. These FC abnormalities in extensive brain regions not only form a "drooling network" in PD, but may also be used to explain the neural underpinnings of cognitive impairment, hallucinations, depression, dysphagia, dysarthria, and other symptoms in PD patients with drooling. Moreover, abnormal FC could predict the severity of motor symptoms, nonmotor symptoms, and depression in droolers and improve the discrimination between non-droolers and non-droolers. This study provides new ideas for further exploring the neuropathological mechanism of drooling in PD.

AUTH O R CO NTR I B UTI O N S
YDZ, YYT, CYC, and TH were involved in study design and manuscript draft. TH, LLT, and YJZ were involved in recruitment of subjects and data collection. TH, LLT, YJZ, and SAS carried out data analysis and discussion. All authors have read and approved the final manuscript.

ACK N OWLED G M ENTS
The authors thank all participants who contributed to the present research. The authors also appreciate the editor and reviewers for their kind help and constructive suggestions.

CO N FLI C T O F I NTER E S T S TATEM ENT
The authors have declared that no conflict of interest, financial or otherwise, exists.

DATA AVA I L A B I L I T Y S TAT E M E N T
All data generated or analyzed during this study are included in this published article and its supplementary information files.