This single-center, double-blind, randomized, placebo-controlled trial was conducted at the Neurology and Clinical Nutrition Department of the Second Affiliated Hospital of Guangzhou Medical University. The study was approved by the hospital’s Ethics Committee (Project No. 2019-hs-42), and is registered in the Chinese Clinical Trial Registry (http://www.chictr.org.cn/; Identification No: ChiCTR2000035122). Participants provided written informed consent before starting the trial.

A 12-week Randomized Controlled Trial (RCT) was performed, and metagenomic changes in the gut microbiota of some patients were detected. Figure 1 illustrates the trial process. Participants were enrolled from September 1, 2020 to August 31, 2021. The inclusion criteria were: (1) meeting the diagnostic criteria of the Movement Disorder Society (MDS) Clinical Diagnostic Criteria and Hoehn and Yahr stages 1–3 (16, 17), indicating clinically established PD; (2) the ability to independently or with assistance from family members complete examinations, questionnaires, and provide medical history; (3) use of anti-parkinsonian drugs such as levodopa, dopadecarboxylase-inhibitor, dopamine agonists (DA), monoamine oxidase type B inhibitor, catechol-o-methyltransferase inhibitor, amantadine hydrochloride, and trihexyphenidyl provided that the dosage remained unchanged in the 30 days leading up to enrollment. The exclusion criteria were: (1) severe cognitive impairment or aphasia and dysarthria leading to communication difficulties; (2) severe liver or kidney dysfunction or tumors; (3) use of immunosuppressants, antibiotics, or other probiotics, or fecal microbiotic transplantation 30 days prior to enrollment; (4) history of severe and uncontrolled hypertension; and (5) spontaneous bleeding, coagulopathy, or long-term use of anticoagulants. The exit criteria were: (1) lost to follow-up; (2) unacceptable or serious adverse events; (3) adjustment of anti-parkinsonian drug dosage during the trial; and (4) using <20% of the trial supplement dose. All patients had primary PD without family history of PD. However, genetic testing was not performed, and the exclusion of patients with multifactorial PD was based primarily on the absence of established concomitant neurologic conditions, such as dementia, brain atrophy, stroke.

An independent neurologist is responsible for the assessment of patients. The assessment scales include the UPDRS scale, Hamilton Depression Scale-24 (HAMD-24), Self-Rating Anxiety Scale (SAS), Constipation Scoring System Scale (CSS), and Minimum Mental State Examination (MMSE). The patients were evaluated at day 0, 6 weeks and 12 weeks. Each evaluation was performed at the same time of day for each patient before and after enrollment, thus basically excluding the influence of motor fluctuations and dyskinesia on the results. In fact, four patients in this trial experienced motor fluctuations, yet they were in the ‘ON’ state during each evaluation. The response rate was defined as the proportion of Minimal Clinically Important Difference change in patient scores from baseline to 12 weeks (Supplementary Data 1.1).

Per previous study results (18), the total UPDRS reduction from baseline with placebo as adjunctive therapy was 8.5% in Hoehn-Yahr I–III patients. In our preclinical trial data, we observed a substantial decrease of 45% in the UPDRS. We set the alpha value at 0.05, power effect value at 80%, and a ratio of 1, and calculated the required sample size to be 40 (20 per group) using pass 11.0 software (NCSS, Inc., USA). Considering a 15% dropout rate, 46 patients (23 patients per group) were enrolled.

Participants were allocated (1:1) to the trail or placebo group using a concealed random allocation from a computer-generated random numbers table produced using the Python programming environment. A nurse dispensed the trial supplement, but she did not know which trial supplement the patient received. BF839 + earthworm protein supplements and placebo were prepared and packaged by Guangzhou Totem Life Medical Research Co. Ltd., China. The two products are almost identical and cannot be distinguished. To ensure the authenticity of the statistical results, both the participants and researchers (including people of trial organization, follow-up, evaluation, data entry and statisticians) were blinded. Unblinding was only indicated upon serious adverse events, with approval of a Steering.

10 g BF839 + earthworm protein supplement trial solution dissolved in 200 mL water, twice daily, or matching placebo as an adjunctive therapy for 12 weeks. 10 g trial supplement contains 10⁶ BF839 and 0.3 g earthworm protein. More details about BF839 and earthworm protein production can be seen in Supplementary material 1.2. The placebo was made from maltodextrin and had a similar odor and taste than did the BF839 + earthworm protein supplement with identical packaging. Concomitant anti-parkinsonian medications were retained, and no changes were allowed throughout the trial.

Patients were monitored for gastrointrial inal-related diseases, including common adverse events such as nausea, vomiting, diarrhea, constipation, increased exhaust, and rash. Unexpected adverse events were also recorded.

Stool samples were collected from 20 randomly selected patients on both day 0 and week 12. Ten patients were selected from the placebo group; two failed to provide stool samples at week 12, resulting in a total of 18 samples. In the test group, 10 patients were selected, resulting in a total of 20 samples. Samples were collected at home and shipped to Shenzhen 01 Life Institute Co. Ltd. at room temperature for testing. R3.6.3 was used to calculate species count and Shannon diversity to assess microbiota diversity. Bray-Curtis distance and PCoA were applied to analyze microbiota composition changes, with a permutational multivariate analysis of variance performed to identify temporal and group differences in microbiota composition. Gene set alignment and intestinal metabolic modules prediction were performed (Supplementary material 1.3–1.8).

Clinical data were analyzed using per-protocol and intention-to-treat analyses. Statistical analyses were performed using SPSS statistical software (version 22.0; SPSS Inc., Chicago, IL, USA). Measurement data were expressed as mean ± SD, and the comparison between groups was analyzed using t-test. Count data are expressed as number of cases (%), and the chi-square or Fisher’s exact test was used for between-group comparison. Statistical significance was set at p < 0.05.

Of 50 patients screened between September 1, 2020, and August 31, 2021, 46 were included. In the trial group, one patient exited because of a change in anti-parkinsonian drugs, and another because of an adverse event. In the placebo group, one patient withdrew consent because he participated in another trial and another could not return to the hospital due to the COVID-19 prevention policy. Forty-two patients completed the trial. Table 1 presents their baseline characteristics.

At 6 weeks, in the intention-to-treat analysis, the trial group exhibited a significant reduction in the UPDRS total and UPDRS part IV scores compared with the placebo group. Additionally, the trial group showed a non-significant downward trend in the UPDRS part I, II, and III scores than the placebo group. Compared with the placebo group, the trial group exhibited significant reduction in the HAMD-24, SAS, and CSS scores. A similar trend was observed in the per-protocol analysis (Table 2).

At 12 weeks, in the intention-to-treat analysis, the trial group demonstrated significant decreases in the UPDRS total, UPDRS part I, UPDRS part II, UPDRS part III, and UPDRS part IV scores compared to the placebo group. The trial group showed better improvement than the placebo group, especially in the leg symptom (Leg agility、Arising from Chair) of UPDRS part III score. Other symptoms assessed by UPDRS part III (Table 2), also showed a non-significant downward trend in the trial group compared to the placebo group. Compared to the placebo group, the trial group showed significant decreases in the HAMD-24, SAS, and CSS scores. A similar trend was observed in the per-protocol analysis (Table 2).

In intention-to-treat analysis n, at 12 weeks, compared with the placebo group, the trial group had a significantly higher response rate in patients with score decrease of total UPDRS≥8 point (47.82% Vs 4.34%, p = 0.001), UPDRS part II ≥ 2 point (60.87% Vs 17.39%, p = 0.003), UPDRS part III ≥ 5 point (30.43% Vs 4.34%, p = 0.02), and CSS >30% (95.65% Vs 0%, p < 0.001). Per-protocol analysis showed similar results (Supplementary Table S1). Two special cases in the trial group did not receive levodopa and took only rasagiline (1 mg/day). Their UPDRS scores reduced substantially from 13 and 14 at baseline to 5.5 and 4.5 at 12 weeks, respectively. Both patients regained their sense of smell at 12 weeks. Additionally, one patient in the trial group reported regaining his sense of smell.

In the trial group, 4.3% (1 of 23) of the patients reported mild eczema at 3 weeks, which resolved spontaneously after one week and did not lead to an exit from the study. Additionally, 4.3% (1 of 23) of the patients in the trial group experienced dizziness and low blood pressure (80–105 mmHg/50–60 mmHg), resulting in withdrawal from the study. However, this patient was concurrently using other Chinese medicines, including sodium aescinate tablets, ginkgo ketone drop pills, and compound Xueshuantong capsules containing notoginseng, Astragalus, Salvia miltiorrhizae, and Radix scungshentch, to improve circulation. Therefore, it is difficult to determine the correlation between these adverse events and the use of the trail medication. Additionally, two cases of bloating were reported in the placebo group; however, these did not meet the criteria for study withdrawal.

There was no significant difference in diversity (Figures 2A,B) and species-level abundance (Figures 2C,D) between the placebo and trial groups on day 0.

At 12 weeks, there was a greater decrease in microbiome diversity in the trail than placebo group; however, the difference was not significant (Figure 2A). No clustering was observed for the Bray–Curtis dissimilarity in principal coordinates analysis (PCoA) (Figure 2B).

Compared with the placebo group, the trial group exhibited a marked increase in the abundance of Enterococcus faecium at 12 weeks. Conversely, the abundances of Klebsiella variicola, Roseburia intestinalis, Sellimonas intestinalis, and Klebsiella aerogenes considerably decreased in the trial group (Figure 2C). When compared with day 0, the placebo group showed higher levels of Eubacterium eligens and Bacteroides uniformis at week 12. However, the trial group exhibited a higher abundance of Enterococcus faecium and a lower abundance of Eubacterium eligens at week 12 compared to day 0 (Figure 2D).

The metabolic functions of the neuroactive compounds encoded by gut microbes were analyzed. No significant differences were observed between the placebo and trial groups on day 0 and at 12 weeks. The bifidobacterial shunt, lactate production, and pentose phosphate pathway (non-oxidative branch) metabolic functions in the trial group were remarkably increased compared to those in the placebo group. Meanwhile, glutamine degradation II, nitrate reduction (dissimilatory), and fucose degradation considerably decreased (Figure 2E). At 12 weeks, succinate production, nitrate reduction (dissimilatory), and lysine degradation I-related functions were markedly enhanced in the placebo group compared to those at day 0. Contrastingly, glycerol degradation II, cysteine biosynthesis/homocysteine degradation, and homoacetogenesis considerably decreased. At 12 weeks, in the trial group, the pentose phosphate pathway (oxidative phase) and pyruvate: formate lyase increased compared to those at day 0. Meanwhile, glutamate degradation I and pyruvate: ferredoxin oxidoreductase substantially decreased (Figure 2F).

This is only a small sample size pilot trial, the results need to be validated in more samples. We did not employ specifically non-motor symptoms scales for PD patients to assess their cardiovascular symptoms, sleep and gastrointestinal function et al., which leading to the result is imperfection. We tested only 20 stool samples and can not exclude the possibility that maltodextrin could affect the gut microbiota, which resulted in insufficient confidence in the results of gut microbiota testing. We do not record the levodopa equivalent daily dose, also a limitation of the experiment.