Characteristics of the study population

In this study, 76 patients with TAK and 56 healthy controls (HCs) were enrolled (Table 1). The median age at sampling was 51 years (interquartile range: 31–68 years) and 67 patients (88.2%) were women in the TAK group. No significant differences in age, sex, or body mass index were found between the TAK and HC groups. Among 19 (25.0%) patients who were newly diagnosed as TAK, 17 patients had not started immunosuppressive treatment at the time of study enrolment, and the median duration of disease was 12 years (interquartile range: 5–30 years; Table 2). According to the Numano scale [34], 42.1% of patients with TAK had type IIb and 34.2% had type V. Active disease (defined as National Institute of Health [NIH] criteria ≥ 2, see Supplementary Methods) was observed in 14 (18.4%) patients [35]. The number of patients who had previously undergone cardiovascular surgery and endovascular treatment was 15 (19.7%) and 9 (11.8%), respectively. Of these, two patients had undergone both cardiovascular surgery and thoracic endovascular aortic repair for aortic aneurysm (Supplementary Table 1). More than half of the patients received GCs (68.4%) and antiplatelets (57.9%), and a proton pump inhibitor (PPI) was administered in 55 (72.4%) patients to prevent gastric ulcers caused by these medications. An anti-IL-6 receptor antibody, tocilizumab, was administered to 30 (39.5%) patients for treatment of TAK at the time of faecal sampling. Trimethoprim-sulfamethoxazole was given in 21 (27.6%) patients for preventing pneumocystis pneumonia.

Table 1 Baseline characteristics of the patients with TAK and HCsTable 2 Baseline characteristics of the patients with TAKCharacteristics of gut microbiome in patients with TAK

We compared the gut microbiome between all patients with TAK and HCs using 16S ribosomal RNA sequencing. Alpha diversity evaluated by the Shannon index, Faith’s phylogenetic diversity, and observed OTUs were similar in the TAK and HC groups (Fig. 1A). Beta diversity evaluated by principal coordinate analysis using the weighted UniFrac distance was different between the groups, which indicated that the gut microbial components differed between the groups (P < 0.05; Fig. 1B). We calculated the microbial dysbiosis index and found that the TAK group showed significant gut dysbiosis compared with HC group (P < 0.0001; Fig. 1C) [36]. A partial least-squares discriminant analysis (PLS-DA) on the gut microbiome showed different clustering in the observed taxonomy between the two groups (Fig. 1D). Although the relative frequencies at the phylum level were similar between the TAK and HC groups, the TAK group showed a higher abundance of Streptococcaceae and a lower abundance of Bifidobacteriaceae at the family level (Supplementary Fig. S1A, B). At the genus level, the abundance of Streptococcus, Lactobacillus, Haemophilus, Campylobacter, Gemella, and Actinomyces, most of which are oral bacteria, was significantly higher in the TAK group than in the HC group. However, the abundance of Lachnospiraceae CAG-56 and Bifidobacterium, which are lactic acid-producing bacteria, was significantly lower in the TAK group than in the HC group (Fig. 1E–G, Supplementary Fig. S1C, D). These changes were unrelated to the use of biological agents among patients receiving immunosuppressive therapy (Fig. 1H, I). Additionally, because most of the patients with TAK included in this study were inactive, we focused on patients with active TAK (NIH score ≥ 2) and compared their intestinal microbiota with those of HCs. Patients with active TAK showed a similar gut dysbiosis as all patients (Supplementary Figs. S2, S3).

Fig. 1

Gut microbial diversity and taxonomy in patients with Takayasu arteritis (TAK) and healthy controls (HCs). A Alpha diversity. B Principal component analysis (PCoA) at the genus level using the weighted UniFrac distance based on permutational multivariate analysis of variance. C Microbial dysbiosis index analysed by the Mann–Whitney U test. D Partial least-squares discriminant analysis. E Volcano plot of the relative abundance of the gut microbiota at the genus level. The red dots represent increased bacteria and the blue dots represent decreased bacteria in patients with TAK compared with HCs. F and G, Bar plots of relatively increased bacteria (F) and decreased bacteria (G) in patients with TAK. In the bar plots, data are shown as the mean ± standard error of the mean. Each dot represents an individual (n = 76 for the TAK group and n = 56 for the HC group). H and I, Bar plots of bacteria shown in (F) and (G), respectively in patients with TAK with or without biological agents (Bio). Each dot represents an individual (n = 33 for Bio( +) and n = 24 for Bio(–)). ns, not significant. *P < 0.05; **P < 0.01; ****P < 0.0001

The effect of drugs on gut microbiome in patients with TAK

The patients in this study were taking various drugs as in actual clinical practice (Table 2 and Supplementary Table S2). Therefore, we investigated whether these drugs were associated with the relative abundance of the genus Streptococcus, which was the most significantly increased genus in the patients compared with HCs (Fig. 1E). We divided patients with TAK into two groups: untreated groups (patients with no history of taking any kinds of immunosuppressive agents) and treated groups. Considering that PPI affects gut microbiota composition [37,38,39], we also excluded patients taking PPIs from untreated group. First, we compared the gut microbiota among untreated patients (n = 11), treated patients (n = 59), and HCs (n = 56). We calculated the microbial dysbiosis index and found that the index was significantly increased even in untreated patients compared with HCs (Supplementary Fig. S4C). The treated group had higher abundance of oral bacteria such as Streptococcus, Actinomyces, Lactobacillus, Gemella, Haemophilus, and Campylobacter and lower abundance of Lachnospiraceae CAG-56 compared with untreated group and HCs in genus level (Supplementary Fig. S4). These results were almost similar to those observed in the comparison between all patients with TAK and HCs (Fig. 1F, G, Supplementary Fig. S1C, D). Since these changes might be affected by the treatment of TAK, we next calculated a correlation coefficient matrix of the relative abundance of the genus Streptococcus, laboratory test values, oral medication, and host characteristics (Fig. 2A). The relative abundance of Streptococcus in the gut microbiome showed a strong correlation with PPI administration (polyserial correlation coefficient: 0.81). PPIs were taken by 55 patients, and 43 of them received GCs. Other 12 patients were on PPIs without GCs, and 10 of them were treated with antiplatelet or bisphosphonates, both of which are risk factors for gastric ulcer. The remaining two patients were taking PPIs due to gastroesophageal reflux disease rather than prevention of gastric ulcer. This relationship between the relative abundance of genus Streptococcus and PPI administration was similarly observed in patients with active TAK (Supplementary Fig. S5A). A comprehensive analysis showed that an increase in oral bacteria, such as Streptococcus, Actinomyces, and Lactobacillus, was strongly correlated with PPI administration (Fig. 2B). PLS-DA showed a clear separation between patients who were taking PPIs (n = 55) and HCs (n = 56), and the patients without PPI administration (n = 21) were plotted between these two groups (Supplementary Fig. S5B). Component 1 of the PLS-DA consisted of oral bacteria, such as Streptococcus, Lactobacillus, and Haemophilus (Supplementary Fig. S5C). The relative abundance of Streptococcus, Gemella, Rothia, and Campylobacter was significantly higher in patients with TAK taking PPIs than in those not taking PPIs (Supplementary Fig. S5D). These results suggest that although gut dysbiosis observed in patients with TAK, including an increase in oral bacteria, is affected by PPIs, a similar dysbiosis appears to occur in patients with TAK without PPIs.

Fig. 2

Gut microbiota taxonomy and the risk of vascular events in patients with TAK taking PPIs. A Correlation coefficient of clinical parameters and the relative abundance of the genus Streptococcus. B Correlation coefficient between the gut microbiota at the genus level and PPI use by Spearman’s two-sided rank correlation test (r > 0.2). C and D Volcano plot of the relative abundance of gut bacteria in patients with TAK taking PPIs with or without events at the genus level (C) and species level (D). The red dots represent increased bacteria and the blue dots represent decreased bacteria in patients with TAK with events compared with patients with TAK without events. E and F Relative abundance of Campylobacter and Fusobacterium at the genus level (E), Campylobacter gracilis and Fusobacterium mortiferum at the species level (F) in patients with or without aortic aneurysm-related events. The Mann–Whitney U test was used for analysis. In the bar plots, data are shown as the mean ± standard deviation. Each dot represents an individual. ns, not significant; *P < 0.05

The relationship between gut microbiome and aneurysm in TAK

As the disease status of TAK progresses, patients sometimes require surgery or endovascular intervention for the treatment of vascular complications such as aneurysms. Traditional serum biomarkers such as CRP can be negative in patients treated with biologics such as tocilizumab, and frequent imaging studies are problematic owing to radiation exposure. Therefore, we searched for candidate biomarkers in the gut microbiome to predict these events. We compared the gut microbiome of patients with TAK with and without a prior history of aortic aneurysm-related events, namely cardiovascular surgeries or endovascular treatments for aortic aneurysmal dilatation and progression of aortic aneurysms (see Supplementary Methods). Because the gut microbiota underwent significant changes with PPI administration, this analysis was limited to patients with TAK taking PPIs (Table 3 and Supplementary Table S3). PLS-DA clearly separated the patients with aortic aneurysm-related events (n = 14) from those without (n = 41) at the genus level (Supplementary Fig. S6A, C) and at the species level (Supplementary Fig. S6B, D). A volcano plot at the genus level showed that the relative abundance of the genera Campylobacter and Fusobacterium was higher in patients with aortic aneurysm-related events than in those without (Fig. 2C). The abundance of some bacteria, such as Eisenbergiella and Coprococcus, was lower in patients with aortic aneurysm-related events than in those without (Fig. 2C, Supplementary Fig. S6E). A species-level analysis showed that the increase in the genus Campylobacter was mainly due to Campylobacter gracilis, which is often found in the oral cavity (Fig. 2D) [40]. Similarly, a species-level analysis showed that the increase in the genus Fusobacterium was mainly due to Fusobacterium mortiferum, which is also an oral commensal bacteria (Fig. 2D). The genus Campylobacter and C. gracilis were detected in 22 (40.0%) and 12 (21.8%) patients taking PPIs, respectively. The relative abundance of these bacteria was significantly higher in patients with aortic aneurysm-related events than in those without (Fig. 2E, F). In an analysis of all 76 TAK cases, the relative abundance of Campylobacter was associated with PPI administration (Supplementary Fig. S7). The detection rates for the genus Fusobacterium (19, 34.5%) and F. mortiferum (6, 10.9%) were lower than those for the genus Campylobacter and C. gracilis (Fig. 2E, F). Notably, the factors which may affect formation/progression of aortic dilatation such as age, body mass index, duration of disease, HLA-B52, inflammation markers, lipid status, and smoking habits were not significantly different between the patients with aortic aneurysm-related events (n = 14) and those without (n = 41) (Table 3). These results suggest that an increase in the abundance of Campylobacter and Fusobacterium in the gut microbiota is associated with aortic dilatation, and that these bacteria might be predictive biomarkers of progression of aortic aneurysms.

Table 3 Baseline characteristics of patients with TAK taking PPIs with/without surgical and endovascular eventsPrediction of aneurysmal complications based on the profiles of gut dysbiosis

We then investigated whether gut dysbiosis can predict the future progression of aortic aneurysms. Because most of the patients were taking PPIs, we limited this prospective analysis to those taking PPIs. The medium follow-up duration was 16 months (interquartile range, 6.25–19 months). Kaplan–Meier analyses showed that patients who had Campylobacter detected in the gut microbiota had a significantly higher incidence of aortic aneurysm-related events, namely surgeries and endovascular treatments due to aortic dilatation, than patients who did not (hazard ratio 14.65, 95% confidence interval 2.383–90.08, P < 0.005; Fig. 3A). A similar finding was observed when the patients were stratified by the detection of C. gracilis (hazard ratio 13.33, 95% confidence interval 1.453–122.3, P < 0.05; Fig. 3B). In addition, patients who had Fusobacterium detected in the gut microbiota had a significantly higher incidence of aortic aneurysm-related events than patients who did not (hazard ratio 8.171, 95% confidence interval 1.014–65.84, P < 0.05; Fig. 3C). However, there was no difference in aortic aneurysm-related event-free survival according to the relative abundance of Streptococcus (Fig. 3D). Conventional serum markers, such as CRP and the erythrocyte sedimentation rate, also failed to predict these events (Fig. 3E).

Fig. 3

Prospective data of the relationship between gut microbiota taxonomy and patients with aortic aneurysm-related events. A–C Aortic aneurysm-related event-free survival in the patients with or without Campylobacter (A), Campylobacter gracilis (B), and Fusobacterium (C). D Aortic aneurysm-related event-free survival in the patients with Streptococcus abundance higher (High) or lower (Low) than the median. E Aortic aneurysm-related event-free survival in the patients who were positive or negative for conventional serological inflammation markers (CRP ≥ 1.0 mg/dL or an erythrocyte sedimentation rate ≥ 30 mm/hour). P values were calculated by the log-rank test, and hazard ratios were calculated by the Mantel–Haenszel test

Prediction of aneurysmal complications based on PCR analysis of gut microbiome

We also performed polymerase chain reaction (PCR) to detect C. gracilis in the gut microbiota and PCR was positive in 21 patients (Fig. 4A, Supplementary Fig. S8A). Kaplan–Meier analysis showed that patients who were positive for C. gracilis by PCR had a significantly higher incidence of aortic aneurysm-related events than those who were negative (hazard ratio 6.534, 95% confidence interval 1.057–40.39, P < 0.05; Fig. 4B). We examined whether the severity of aortic aneurysms differed between the patients who were positive for C. gracilis by PCR and those who were negative. There was no difference in the proportion of patients treated with tocilizumab with or without C. gracilis positivity (Supplementary Fig. S8B). We evaluated computed tomography or magnetic resonance imaging of 15 patients who underwent interventions owing to aortic dilatation before or after stool sample collection. We divided the aorta into four regions and measured the maximum short diameter of each region and calculated the mean diameter (see Supplementary Methods). Based on Case 16, which had the largest mean diameter among patients who did not have an event, nine cases exceeded this value (Fig. 4C). Among these nine patients, seven were positive for C. gracilis. The three-dimensional reconstructed aortic images of these patients revealed that patients who were positive for C. gracilis by PCR tended to have more severe aortic aneurysms (Fig. 4D, Supplementary Table S4). Smoking, hypertension, male sex, and persistent inflammation are known risk factors for the development of aortic aneurysms in patients with TAK [41]. Case 1 (C. gracilis-positive) had a thoracoabdominal aortic aneurysm with the largest mean shortest diameter, despite receiving adequate immunosuppressive therapy and having none of these risk factors. These results suggest that intestinal C. gracilis may be a novel tool for predicting aneurysmal formation and progression in TAK.

Fig. 4

Severity of aortic aneurysms and C. gracilis in the gut in patients with TAK. A PCR electrophoretic analysis. The upper images show the amplified products of the universal 16S primers (first PCR), and the lower images show those of C. gracilis specific 16S primers (second PCR). B Aortic aneurysm-related event-free survival in the patients who were positive or negative for C. gracilis by PCR. C Mean of the maximum short diameters in four regions of the aorta. D Three-dimensional computed tomography (3D-CT) images before interventions and positivity of C. gracilis by PCR in each patient with TAK who had interventions owing to aortic dilatation. Case 16 is shown as a reference for patients with TAK without interventions. Images in Case 6 are chest CT and abdominal CT. Images of 3D-CT were missing in Cases 12 and 15