Introduction

Moyamoya disease (MMD) is a rare cerebrovascular disease characterized by stenosis in the terminal portions of the bilateral carotid arteries and collateral blood vessels due to compensatory mechanisms corresponding to that stenosis1. Inadequate collateral blood vessel development causes cerebral ischemia in pediatric patients with MMD, and fragile collateral blood vessels drive intracranial hemorrhage in adult patients with MMD2. The only established treatment to reduce the risk of cerebral ischemia or intracranial hemorrhage is revascularization surgery performed by neurovascular surgeons3.

Although Ring Finger Protein 213 (RNF213) was identified as a susceptibility gene for MMD in 20114,5, RNF213 alone cannot explain its pathophysiology. The RNF213 p.R4810K variant is estimated to be present in 2.5% of the general Japanese population; however, less than 1% of these individuals develop MMD6. Therefore, environmental factors are thought to be involved in the pathogenesis of MMD, and an immune response is the most promising candidate. Pathology findings indicate that proliferating smooth muscle cells and inflammatory cells (macrophages and T cells) are colocalized in the thickened intima of occlusive major intracranial arteries7. In addition, transcriptome-wide analysis of blood and intracranial artery samples from patients with MMD shows that the pathophysiology of MMD is related to the immune response8,9.

The oral and gut microbiomes have attracted attention because of their association with vascular and autoimmune diseases10,11,12,13,14,15. The gut microbiome of patients with atherosclerotic cardiovascular disease deviates from that of healthy controls10. In cerebrovascular diseases, CNM-positive Streptococcus mutans in the oral microbiome is associated with an increased incidence of cerebral microbleeds11, and Campylobacter ureolyticus in the gut microbiome may be related to the rupture of cerebral aneurysms12. Regarding autoimmune diseases, the status of inflammatory bowel disease is strongly correlated with the intestinal immune system, which is activated by intestinal colonization of bacteria from oral origin13,14. In rheumatoid arthritis, dysbiosis of the oral and gut microbiomes promotes the production of autoantibodies that migrate to the joints and contribute to disease onset15.

MMD is a cerebrovascular disease in whose pathophysiology the immune response plays an essential role. However, the relationship between MMD and these microbiomes remains unelucidated16. We suspect that a particular bacterium in the oral or gut microbiome drives the immune response that causes MMD. This study aimed to investigate whether there is a difference in the oral and gut microbiome profiles between patients with MMD and healthy individuals, and to identify bacterial species specific to patients with MMD by 16S rRNA sequencing.

Results

Characteristics of the study cohort

The characteristics of the study cohort for each sample are summarized in Table 1. For saliva samples, the median patient age was 32 years (IQR 25–39; range 7–58) for healthy controls and 32 years (IQR 10–45; range 5–55) for patients with MMD. For fecal samples, the median patient age was 32 years (IQR 25–39; range 4–58) for healthy controls and 34 years (IQR 17–46; range 5–55) for patients with MMD. Demographic data, including sex, body mass index, hypertension, dyslipidemia, diabetes, smoking, and alcohol consumption, were also similar between groups with no significant differences.

Table 1 The characteristics of the study cohort in saliva samples and fecal samples.
Full size table

Differences in oral microbiome between patients with MMD and healthy people

For alpha diversity, the Chao1 index and observed species demonstrated significant differences between groups (Fig. 1a). No differences in beta diversity were observed between the groups (Fig. 1b).

Figure 1
figure 1

Comparison of oral microbiome between patients with MMD and healthy people. (a) Box and beeswarm plots showing alpha diversity. The Chao1 index and observed species were significantly different between the MMD and control groups. (b) Principal coordinate analysis plots of beta diversity based on weighted and unweighted UniFrac distances. There was no difference between the MMD and control groups in weighted UniFrac distances (p = 0.779) or unweighted UniFrac distances (p = 0.181). (c) Cladogram showing discriminative taxa identified with default parameters in LEfSe (p  2.0). Campylobacter was enriched in patients with MMD at both the phylum and genus levels. The red bar shows the LDA score for one phylum, two classes, two orders, four families, and six genera that were more abundant in patients with MMD. (d) Heatmap of the discriminative genera identified using LEfSe. These taxa were more abundant in patients with MMD. (e) Volcano plot of the values calculated using ANCOM-BC. Campylobacter had a significantly lower adjusted p-value (< 0.05) with a higher fold change.

Full size image

In the linear discriminant analysis (LDA) effect size (LEfSE), one phylum (Campylobacterota), two classes (Campylobacteria, Coriobacteriia), two orders (Campylobacterales, Coriobacteriales), four families (Saccharimonadaceae, Fusobacteriaceae, Campylobacteraceae, Atopobiaceae), and six genera (TM7x, Fusobacterium, Campylobacter, Mogibacterium, Atopobium, Dialister) showed differentially abundant microbes between groups (Fig. 1c,d). Notably, Campylobacter was more abundant at the phylum and genus levels in patients with MMD. In the analysis of the compositions of microbiomes with bias correction (ANCOM-BC), ten genera containing Campylobacter were differentially abundant between the groups (Fig. 1e, Table 2).

Table 2 Differentially abundant genera in the oral microbiome between patients with MMD and healthy individuals.
Full size table

Comparison of gut microbiome between patients with MMD and healthy people

Differences between the groups were not observed in alpha or beta diversity (Fig. 2a,b).

Figure 2
figure 2

Comparison of gut microbiome between patients with MMD and healthy people. (a) Box and beeswarm plots showing alpha diversity. No significant differences were observed between the groups. (b) Principal coordinate analysis plots for beta diversity. There was no difference between the MMD and control groups in the weighted UniFrac distances (p = 0.456) and unweighted UniFrac distances (p = 0.399). (c) Cladogram showing the discriminative taxa identified using the default parameters in LEfSe. Streptococcus and Lactobacillales were enriched in patients with MMD. The red bar shows the LDA score for one class, five orders, eight families, and nine genera that were more abundant in patients with MMD. (d) Heatmap of the discriminative genera identified using LEfSe. These taxa were more abundant in patients with MMD. (e) Volcano plot of the values calculated using ANCOM-BC. None of the genera showed a low adjusted p-value (p < 0.05).

Full size image

In LEfSE, one class (Bacilli), five orders (Lactobacilales, Micrococcales, Actinomycetales, Enterobacterales, Staphylococcales), eight families (Streptococcaceae, Micrococcaceae, Gemellaceae, Actinomycetaceae, Peptostreptococcales Tissierellales, Enterobacteriaceae, Butyricicoccaceae, Lactobacillaceae), and nine genera (Streptococcus, Rothia, Gemella, Actinomyces, Lactobacillus, Escherichia_Shigella, Lachnospiraceae_UCG_004, Butyricicoccus, Slackia) showed differentially abundant microbes between groups (Fig. 2c,d). However, no differentially abundant microbes were identified in ANCOM-BC (Fig. 2e).

Discussion

We investigated whether patients with MMD have specific oral and gut microbiomes compared to healthy individuals based on the assumption that these microbiomes are involved in the immune response in MMD. The 16S rRNA data showed that patients with MMD had different oral microbiomes, especially Campylobacter, than healthy controls. In addition, differences in the gut microbiome between patients with MMD and healthy individuals were smaller than those in the oral microbiome.

For the first time, this study revealed differences in alpha diversity (Chao1 index, observed species) and differentially abundant microbes in the oral microbiome. As the Chao1 index and observed species quantified the richness of species, patients with MMD had a greater diversity of oral bacteria. Furthermore, both analytical methods for detecting differentially abundant microbes consistently demonstrated that Campylobacter was significantly more abundant in patients with MMD. Oral bacteria can transiently access the bloodstream during daily oral hygiene practices17, and certain oral bacteria (Porphyromonas gingivalis) have been reported to infiltrate the brain18. In addition, although it is not a vascular cell line, Campylobacter modulates the immune response by upregulating TLR4 and MD-2 in HT-29 cells (a cell line with epithelial morphology)19. Therefore, the relationship between these mechanisms of Campylobacter (transient blood-borne infection and infiltration into vascular tissue) and the immune system in MMD should be investigated. Furthermore, Campylobacter is an intracellular bacterium containing lipopolysaccharides, and it was recently reported that RNF213 is involved in the ubiquitination of intracellular bacterial lipopolysaccharides20. In this regard, Campylobacter spp. can be considered a promising environmental factor for MMD. We suspect that the Campylobacter spp. may activate an immune response of the intracranial artery through transient blood-borne infection, contributing to the pathophysiology of MMD.

Regarding the gut microbiome, our results showed that differences between patients with MMD and healthy individuals were smaller than those in the oral microbiome. To date, only one study has investigated the gut microbiome of patients with MMD16. Our LEfSe results were similar to those of a previous report showing that Streptococcus and Lactobacillales were enriched in patients with MMD, suggesting the robustness of our study. Although a previous study showed that Ruminococcus is enriched in patients with MMD, our analysis did not identify this differentially abundant microbe. This difference may be because we enrolled only preoperative patients with MMD, matched control participants to patients with MMD in terms of demographic data, and used different sample collection methods.

The present study has some limitations. First, the study was conducted only in Japan. Therefore, it is essential to note regional and racial differences when comparing the current results to those of other studies. Second, the RNF213 variant was not investigated in this study. Although RNF213 is an important genetic factor for MMD, we considered unnecessary blood sampling to be more invasive than the study merited because our study cohort enrolled young children. Third, the sample size was relatively small. Validation of our study with a larger cohort is necessary for generalization.

Conclusions

The oral microbiome profiles of MMD patients and healthy individuals differed significantly. Campylobacter spp. may be a key environmental factor in the development of MMD.

Materials and methods

A flow diagram of the study is shown in Fig. 3.

Figure 3
figure 3

Flow diagram of the study.

Full size image

Ethical approval and consent to participant

This study was approved by the Institutional Review Board of the Nagoya University Graduate School of Medicine (approval number 2022-0084; approval date, June 3, 2022). Written informed consent was obtained from all the participants or their legal guardians. All methods of this study were performed in accordance with relevant guidelines and regulations.

Study cohort

We prospectively enrolled 16 patients with MMD who did not undergo revascularization surgery between June 2022 and March 2023 at the Nagoya University Hospital, Japan Community Health Care Organization Chukyo Hospital, or Toyota Kousei Hospital. The diagnosis of MMD was based on guidelines proposed by a research committee approved by the Ministry of Health, Labor, and Welfare of the Japanese government21. In brief, the diagnostic criteria were as follows: (1) stenosis or occlusion of the terminal portion of the intracranial internal carotid artery, (2) moyamoya vessels at the brain base, and (3) exclusion of diseases with similar angiographic characteristics (e.g., autoimmune disease, meningitis, brain tumors, Down syndrome, neurofibromatosis type 1, or cerebrovascular lesions after head irradiation). These patients did not use antibiotics within 1 month before sample collection because antibiotics affected the microbiome22. Of all the samples obtained from the 16 patients with MMD, one fecal sample was excluded due to inadequate collection. Finally, this study included 16 saliva and 15 fecal samples from patients with MMD.

For the healthy control group, we recruited healthy individuals who were willing to participate in our study at Nagoya University between June 2022 and March 2023. The healthy individuals had never undergone head surgery and had no history of transient ischemic attacks, ischemic stroke, or intracranial hemorrhage. Similar to the patient group, healthy individuals did not use antibiotics within one month before sample collection. Among the 17 samples collected, we selected 15 that were matched for age, sex, and BMI with the patient group.

Sample collection methods and microbial DNA extraction experiments

OMNIgene ORAL and OMNIgene GUT (DNA Genotek, Ottawa, Canada) were used to collect the saliva and fecal samples. OMNIgene kits demonstrated good microbiome stability at room temperature within 4 days of sample collection23,24. Participants took these samples themselves, and we collected them within 4 days. Immediately after collection, microbial DNA was extracted from each sample using the QIAamp Power Fecal Pro DNA Kit (QIAGEN, Hilden, Germany), following the manufacturer’s instructions.

Library preparation and 16S rRNA sequencing

Library preparation was performed according to the 16S Metagenomic Sequencing Library Preparation protocol (Illumina, San Diego, CA, USA). The V3–V4 region of the 16S rRNA microbial gene was amplified using the 16S Amplicon PCR forward primer (TCGTCGGCAGCGTCAGATGTGTATAAGAGACAGCCTACGGGNGGCWGCAG) and the 16S Amplicon PCR reverse primer (GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAGGACTACHVGGGTATCTAATCC) using the KAPA HiFi HotStart ReadyMix (Kapa Biosystems, Wilmington, MA)25. Illumine sequencing adapters and dual-index barcodes was added to the 16S Amplicon with the Nextera XT Index kit (Illumina). Paired-end 300 base pairs sequencing was performed on the Illumina MiSeq platform using the MiSeq v3 reagent kit (Illumina). Library preparation and sequencing were performed by Macrogen Japan Corp.

Microbiome bioinformatics

Data processing of 16S rRNA sequencing

16S rRNA sequencing results were analyzed using Quantitative Insights into Microbial Ecology 2 (QIIME2) v2023.226. Raw sequence data were imported and demultiplexed. The generated reads were denoised into amplicon sequence variants (ASVs) using the Divisive Amplicon Denoising Algorithm 227. Forward and reverse reads were trimmed, and the median quality score dropped below 20 in the quality score plot28. Taxonomy was assigned to the ASVs using the Scikit-learn naïve Bayes machine learning classifier against the Silva 138 99% 16S rRNA database26,29.

Evaluation of microbiome diversity

Alpha diversity was assessed by calculating the Chao1, observed species, Shannon, and Simpson indices. For statistical comparisons between patients with MMD and healthy individuals, we used the Mann–Whitney U test. Beta diversity was assessed by calculating the weighted and unweighted UniFrac distances. We performed a principal coordinate analysis (PCoA) and permutational multivariate analysis of variance (PERMANOVA). For these analyses, statistical significance was set at p < 0.05.

Analysis comparing microbes between MMD and control groups: LEfSE and ANCOM-BC

To identify differentially abundant microbes between groups, we performed two analytical methods up to the genus level: the LEfSE method and the ANCOM-BC30,31. The matching results between analytical methods increases the credibility of biological interpretation32. LEfSe compares the percentages of microbial composition using a non-parametric statistical test and calculates the LDA score to estimate the effect size of each differentially abundant microbe. ANCOM-BC estimated the unknown sampling fractions and corrected the bias induced by differences among samples before the statistical test. This method compared the read counts of the microbial composition using a parametric statistical test. p values were adjusted using the Benjamini and Hochberg method (adjusted p-value), and genera with an adjusted p-value < 0.05 were considered significant differentially abundant microbes33.

Statistical analysis

To compare the participants’ background characteristics, numerical data were compared using the Mann–Whitney U-test, and categorical data were compared using Pearson’s Chi-squared or Fisher’s exact test. Statistical significance was set at p < 0.05. Statistical analyses were performed using the R version 4.2.2 (https://www.r-project.org/).