This study was conducted in the Department of Neurology, Guangdong Provincial People's Hospital, Guangzhou, People's Republic of China. A diagnosis of idiopathic PD was assigned according to the MDS-Sponsored Revision of the Unified Parkinson's Disease Rating Scale (MDS-UPDRS) (7) by 2 neurologists with experience in movement disorders. Exclusion criteria included: (1) patients with secondary Parkinsonism or Parkinson-plus syndrome; (2) patients with hyperlipidemia or diabetes mellitus; (3) patients with a tumor, heart failure, chronic obstructive pulmonary disease, nephritis, infectious disease, or other severe chronic diseases; (4) a history of cerebrovascular disease, intracranial surgery, head trauma, AD, epilepsy, mental illness, or other conditions that may affect cognitive function; (5) patients failing to cooperate or complete the trial.

Demographic information was collected from each subject, including sex, gender, age, education level, age of the onset, duration, medications, smoking, and drinking history. A detailed medical history and physical examination were collected from all the participants. Unified Parkinson's Disease Rating Scale Part III (UPDRS III) and Hoehn-Yahr Scale (H-Y) were used to assess motor function. Mental status was assessed by Hamilton Anxiety Scale (HAMA) and Hamilton Depression Scale (HAMD).

Our neuropsychological protocol included Mini Mental State Examination (MMSE) and Montreal Cognitive Assessment (MoCA) to assess general cognitive functioning, memory impairment, and executive dysfunction. The attention/working memory domain was tested by Symbol Digit Modalities Test (SDMT) and Digit Span Test (DST). Executive functions were evaluated by the Verbal Fluency Test (VFT) and Clock Drawing Test. Memory was assessed by visual reproduction (VR) and logical memory. Language function was tested by the Similarities Test and Category Fluency Task. Visuo-spatial and visuo-perceptive functions were assessed by the Block design Test. The patients were diagnosed with PD-MCI based on Movement Disorders Society Task Force Level II 20 using a cut-off of 1.5 standard deviations below standard scores in one or more cognitive domains in neuropsychological tests as previously described (7).

A fasting venous blood sample was collected with a lithium heparin tube from all the subjects in the morning. The serum was separated and stored in the refrigerator at −80°C until experimental analysis. Thirty BA metabolites in serum were detected by ultra-high performance liquid chromatography-MS/MS (UHPLC-MS/MS). After quality-control checks, the data set included 19 BAs (11 were too few for statistical analysis) for a total of 67 subjects.

Statistical analyses were performed using SPSS25 software. Between-group (PD-MCI vs. PD-NC) differences in terms of demographics, clinical data, and BA characteristics were addressed by independent-sample t-tests or Mann–Whitney U-test for continuous variables and χ² for categorical variables. False discovery rate (FDR)-based multiple comparison adjustment with the Benjamini-Hochberg procedure was used for BAs comparison to increase the significance of the findings.

To simplify the complexity of the high-dimensional metabolome data, orthogonal projections to latent structures—discriminant analysis (OPLS-DA) was implemented to identify the differential expression of patients with BAs in PD with or without cognitive impairment. Two iterations of the permutation test verified the model. Through OPLS-DA, a variable importance in projection (VIP) value could be obtained for each metabolite. The larger the VIP is, the more significant the contribution of the metabolite to distinguish the two groups. Differential metabolites were defined as VIP >1.0 and FDR-corrected p-values < 0.05. Then, according to the HMDB database, these differentially expressed metabolites were chemically classified. The area under the receiver-operating-characteristic curve (AUC-ROC) was also calculated.

This study was performed following the principles outlined in the Declaration of Helsinki. Informed consent was obtained from all the individual participants included in the study.

A total of 33 patients with PD-NC and 34 patients with PD-MCI were enrolled in this study. Demographic and clinical characteristics for all the subjects are shown in Table 1. Categorical data are displayed as absolute numbers and continuous data as mean ± standard error. The demographic characteristics of the two groups were compared, including gender, age, and education level, and the differences were not statistically significant (p > 0.05), indicating comparability (Table 1).

Mean and standard errors of 19 detected BAs are presented in Table 2. After controlling for multiple testing using FDR, lower levels of cholic acid (CA), chenodeoxycholic acid (CDCA), ursodeoxycholic acid (UDCA), and 7-Ketolithocholic acid were significantly associated with PD-MCI compared to patients with PD-NC (corrected p < 0.01). The OPLS-DA model was also constructed to show statistical differences in CDCA (VIP ≈ 2.67), Glycochenodeoxycholic acid (GCDCA, VIP ≈ 1.846), CA (VIP ≈ 1.658), glycoursodeoxycholic acid (GUDCA, VIP ≈ 1.298), and UDCA (VIP ≈ 1.26) levels between the PD-MCI and PD-NC groups. We identified CA, CDCA, and UDCA with VIP > 1 and corrected p-value < 0.05 as the differential BA metabolites in patients with PD-MCI (Figure 1; Table 2). The data distribution characteristics of these 3 BAs are shown in Figures 2A–C. AUC-ROC for CA, CDCA, and UDCA were 74.2, 78.1, and 74.5%, respectively (Figures 2D–F; Table 2).

After determining the differential bile acids between the PD-MCI and PD-NC groups, we conducted correlation analysis between these bile acid levels and the scores of various cognitive scales to find the specific cognitive domains affected by bile acids. Unfortunately, we found no significant correlation between bile acid levels and specific cognitive domains.

The existence of BA is an emerging research topic in neurodegenerative disease (9). CA and CDCA are the only primary bile acids synthesized in humans, impacting fat digestion and energy metabolism signaling (22, 23). A previous study has shown that CBJC, a combination of Chinese herb active components, including cholic acid, can ameliorate the IBO-induced dementia in rats (24). CDCA has been shown to play a significant neuroprotective role in cerebrotendinous xanthomatosis (25, 26). Besides, CDCA significantly attenuates the cognitive dysfunction and spatial deficits induced by AlCl3 in the rat model (27). UDCA, another critical bile acid we discovered, is converted from CDCA catalyzed by 7α-hydroxysteroid dehydrogenase (7α-HSDH) and 7β-HSDH (28, 29). UDCA has been widely used in treating primary biliary cirrhosis and proved to have good safety and tolerability (30). It crosses the blood-brain barrier in a dose-dependent manner and improves mitochondrial function with anti-apoptotic effects in AD and PD models.

At the micro-level, BAs might be associated with cognitive functions in PD through the theory of neuroinflammation and apoptosis. CDCA could bind with Takeda G-protein-coupled receptor 5 (TGR5), a membrane receptor that alleviates neuroinflammation and inhibits cell apoptosis (31, 32). Besides, daily intraperitoneal injection of UDCA can significantly reduce the levels of nuclear factor-κB (NF-κB), Bcl2-related X apoptotic regulator (Bax), and caspase-9 mRNA, thus exerting anti-apoptotic effects (32). The interaction of BAs with anti-apoptotic substances may explain the relationships between BA and PD-MCI in some ways.

At a macro-level, abnormalities in BA metabolism may affect disease progression by altering structural networks in specific brain regions. Researchers have found that changes in BA of AD and patients with MCI were associated with changes in particular areas in brain imaging (21), which may also happen in patients with PD-MCI. We would combine metabolomics and radiomics to explore the association between BA, specifically CA, CDCA, and UDCA, and brain imaging changes in future studies.

Bidirectional biochemical communication between the brain and the gut is associated with various neurodegenerative diseases, including Parkinson's disease. α-synuclein, a pathological marker of PD, was found to accumulate in the gastrointestinal tract of patients with early PD (33). After intestinal flora obtained from feces of patients with PD was transplanted into the intestinal tract of germ-free rats, all the previously disease-free rats developed PD symptoms (34). This suggests that the gut microbiome might be a vital contributor to Parkinson's disease. In terms of cognitive impairment, our previous studies have found that patients with PD-MCI have unique gut microbial characteristics, suggesting that intestinal flora may be related to the pathogenic mechanism of PD-MCI.

Intestinal flora is an integral part of bile acid metabolism. In rats with PD symptoms that overexpress α- synuclein, intestinal microbial diversity was changed, and the levels of related primary bile acids were significantly increased and thus affecting motor symptoms (16). Another study has found that intestinal bacteria adjust the primary bile acid metabolism in patients with AD and, in turn, affect cognitive function (21). Increased levels of metabolites of intestinal microbiota, including deoxycholic acid and glycocholic acid, were significantly associated with thinning of cortical regions in bilateral frontal, parietal, and temporal lobes (21). This suggests that bile acids may serve as a “bridge” between gut microbiome disorder and the occurrence and development of cognitive impairment in PD.

Because the inclusion criteria were relatively strict, the total number of samples included was relatively small, which might decrease statistical power and cause some false-negative results. In addition, only the patients with PD-MCI and the patients with PD-NC were included in this study to explore the bile acid metabolites in PD cognitive impairment preliminarily. We will recruit more patients with PD with cognitive impairment in future studies, including those with dementia.