Abstract

Introduction

‘Gut microbiota’ is a term that collectively refers to the bacteria, archaea, and eukarya colonizing the gastrointestinal tract. Accumulated studies have shown that regulation and alterations of the gut microbiome are associated with a variety of disorders, such as type II diabetes, asthma, obesity, autism spectrum disorders, and thyroid cancer [1–5]. Recent studies have emphasized the role of the gut microbiome in the pathogenesis of autoimmune diseases such as rheumatoid arthritis, systemic lupus erythematosus, Hashimoto’s thyroiditis, and Graves’ disease (GD) [6–9]. GD induces hyperthyroidism in which thyroid hormone receptor stimulating antibodies (TSHR-Ab) mimic the function of thyroid stimulating hormone (TSH) to activate the thyroid. However, the pathogenesis of GD has yet to be fully elucidated, and the mechanisms triggering the autoimmune attack on the thyroid remain unknown.

The gut microbiome are considered to play an essential role in host immune homeostasis and the release of metabolites developing a diseased condition [10]. A change in the gut microbiome impact the permeability of the intestinal mucosal barrier, which may lead to increase in toxins, antigens, bacterial metabolites, and can even facilitated the entry of bacteria into the bloodstream from the intestine, triggering inflammation. This could act as trigger for autoimmunity, potentially leading to the development of cause autoimmune diseases [11]. Therefore, several studies have sought to elucidate the roles played by the gut microbiome in the pathogenesis of GD [12, 13]. Notably, experiment involving fecal microbiota transplantation from GD patients have demonstrated in the severity of GD features in mouse models [14].

The predominant therapy for GD in Europe, Asia, and the USA involve the use of an anti-thyroid drug (ATD) [15]. Nevertheless, there is a lack of reports regarding in the alterations in microbiome after treatment of GD with the ATD. Therefore, we used next-generation-based sequencing to analyze the change in gut microbiome among GD patients after 6 months of ATD treatment. Our findings not only show the gut microbiome profile of GD patients but also provide substantial data on the changes in gut microbiome post-treatment. This sheds lights on the causal relationship in GD and introduces novel insights for treatment of the disease. Additionally, we conducted an analysis to explore the association between the modification of the gut microbiome and thyroid function along with antibodies.

Materials and methods

Results

The clinical characteristics of the study participants, both before and after ATD treatment, are listed in Table 1. The participants, aged between 20 and 68 years, had a mean age of 45.7 ± 12.3 years. The average body mass index was 23.2 ± 3.7 kg/m². Fifteen participants (51.7%) were female and seven of them had experienced menopause. Treatment details revealed that sixteen participants received methimazole (10–30 mg), twelve were treated with carbimazole (10–30 mg), and one was administered propylthiouracil (100mg). Throughout the follow-up period, the doses of these ATD were adjusted based on individual patient’s clinical symptoms and laboratory findings. There was no significant difference in microbiota with a relative abundance greater than 1% between patients treated with methimazole and those treated with carbimazole. The average SRS was 13.8 ± 8.2 points. Among the participants, 51.7% were current or past smokers, with an average pack-years of 18.7 ± 29.4. Baseline level of free T4, TSH, and TSHR-Ab levels were 4.5 ± 2.6 ng/dL, 0.01 ± 0.01 uIU/mL, and 9.5 ± 8.0 IU/L, respectively. After 6 months, these levels changed to 1.3 ± 0.6 ng/dL, 2.5 ± 3.6 uIU/mL, and 6.1 ± 6.8 IU/L, respectively.

Difference in gut microbiota between GD patients and healthy controls

Fig 1 shows the difference in alpha diversity between 29 GD patients and the 230 healthy control (HC) group. In GD patients, the alpha diversity, as measured by Chao1, Shannon, and Simpson indexes, was consistently lower (p-values: 0.0004, 0.016, and 0.06, respectively). Additionally, there was significant dissimilarity in beta diversity between the GD and HC groups (p < 0.001) (Fig 1).

Open in a separate window

Fig 1

(a) Alpha diversity between GD and HC. (b) Beta diversity between GD and HC. GD, Graves’ disease; HC, healthy control; NMDS, non-metric multidimensional scaling.

1) Difference in gut microbiota at phylum level

Fig 2 and S2 Table shows the taxonomic composition in GD and HC groups at the phylum level. Among microbiota with more than 1% in relative abundance, GD patients showed lower levels of Firmicutes (p = 0.193) and significantly higher levels of Bacteroidota (p < 0.001) compared to HC. Additionally, GD patients showed lower abundance of Proteobacteria, Verrucomicrobiota, and Desulfobacterota (all p < 0.001; Figs ​Figs33 and ​and4).4). The relative abundance of Firmicutes/Bacteroidota was lower in GD patients (1.2 ± 1.3) compared to the HC group (2.2 ± 3.1; p = 0.003).

Open in a separate window

Fig 2

Taxonomy composition in GD patients (n = 29) and HC (n = 130) at phylum level.

GD, Graves’ disease; HC, healthy control.

Open in a separate window

Fig 3

Differences in bacterial phyla between GD patients (n = 29) and HC (n = 130).

GD, Graves’ disease; HC, healthy control.

Open in a separate window

Fig 4

Distinguishment in bacterial phyla between GD patients (n = 29) and HC (n = 130) according to LEfSe algorithm.

GD, Graves’ disease; HC, healthy control.

2) Difference in gut microbiota at genus level

Fig 5 and S3 Table shows the taxonomy composition at the genus level in GD and HC groups. GD showed significantly lower levels of Roseburia, Lachnospiraceaea, Sutterella, Escherichia-shigella, Parasuterella, Akkermansia, and Phascolarctobacterium than HC (all p < 0.05; Fig 6). Additionally, using a LEfSe algorithm, we identified 9 genera with significant distinctions between the two groups (LDA score > 3.0; Fig 7).

Open in a separate window

Fig 5

Taxonomy composition in GD patients (n = 29) and HC (n = 130) at genus level.

GD, Graves’ disease; HC, healthy control.

Open in a separate window

Fig 6

Differences in bacterial genus between GD patients (n = 29) and HC (n = 130) at genus level.

GD, Graves’ disease; HC, healthy control.

Open in a separate window

Fig 7

Distinguishment in bacterial genus between GD patients (n = 29) and HC (n = 130) according to LEfSe algorithm.

Graves’ disease; HC, healthy control.

3) ROC curve analysis to diagnose GD

We identified the top nine microbiota, with an LDA scored greater than 3.0, to serve as biomarkers for the diagnosis of GD, including Sutterella, Roseburia, Escherichia-shigella, Lachnospiraceae NK4A136 group, Parasutterella, Anaerostipes, Bilophila, f.Enterobacteriaceae, and Incertae Sedis. An ROC curve was generated to evaluate the diagnostic accuracy based on these 9 species, revealing an AUC of 0.92 (p < 0.001), sensitivity of 96.6%, specificity of 82.2%, positive predictive value of 40.6%, and negative predictive value of 99.5% (Fig 8). The variables importance to the classification model is shown in S1 Fig.

Open in a separate window

Fig 8

Receiver operating characteristic curve generated with 9 species for the diagnosis of GD: Sutterella, Roseburia, Escherichia-shigella, Lachnospiraceae NK4A136 group, Parasutterella, Anaerostipes, Bilophila, f.Enterobacteriaceae, and Incertae Sedis.

GD, Graves’ disease; PPV, positive predictive value; NPV, negative predictive value.

4) Gut microbiota in smokers and non-smokers in GD

Among GD patients, two genera with more than 1% in relative abundance showed significant differences between smoker and non-smokers. Smokers exhibited a significantly higher relative abundance of Faecalibacterium (10.0 ± 3.6 versus 5.6 ± 3.9, p = 0.014), while Dialister displayed a lower abundance in smokers (0.0 ± 0.0 versus 2.2 ± 4.6, p = 0.014). Additionally, the genus Anaerostipes showed higher abundance in smokers among GD patients (0.2 ± 0.4 versus 0.8 ± 1.0, p = 0.16).

Difference in gut microbiota in GD before and after treatment

Fig 9 shows the change in the alpha diversity in 25 GD patients 6 months after ATD treatment. Post-treatment, there is a significant increase in alpha diversity across all Chao1, Shannon, and Simpson indexes (p-value; 0.003, 0.002, 0.008 respectively). However, there is no a significant difference in the dissimilarity of beta diversity at the ASV level after treatment (p = 0.784) (Fig 9).

Open in a separate window

Fig 9

Difference in gut microbiota before and after anti-thyroid drug treatment (n = 25).

(a) Alpha diversity between GD and TxGD. (b) Beta diversity between GD and TxGD. GD, Graves’ disease; TxGD, Graves’ disease patients after 6 months treatment with anti-thyroid drug; NMDS, non-metric multidimensional scaling.

1) Difference in gut microbiota at phylum level

Fig 10 and S4 Table shows the alterations in composition in the phylum level following ATD treatment. Firmicutes increased significantly after treatment (p = 0.015), while Bacteroidota decreased in the post-treatment GD group (p = 0.014). At the phylum level, Proteobacteria and Actinobacteriota show no significant differences (p = 0.324, p = 0.550, respectively) (Fig 11). The relative abundance of Firmicutes/Bacteroidota increased from 1.2 ± 1.3 in pretreatment GD to 1.5 ± 0.9 in post-treatment GD group, although this trend was not statistically significant (p = 0.258).

Open in a separate window

Fig 10

Difference in taxonomy composition in GD and TxGD at phylum level (n = 25).

GD, Graves’ disease; TxGD, Graves’ disease patients after 6 months treatment with anti-thyroid drug.

Open in a separate window

Fig 11

Differences in bacterial phyla between GD and TxGD (n = 25).

GD, Graves’ disease; TxGD, Graves’ disease patients after 6 months treatment with anti-thyroid drug.

2) Difference in gut microbiota at genus level

Fig 12 and S5 Table shows the alterations in composition at the genus level in the microbiota with relative abundance greater than 1%. The relative abundance of each genus before and after treatment is detailed in S5 Table. Subdoligranulum significantly increased after treatment (p = 0.010), while Veillonella and Christensenellaceae R-7 group both significantly decreased after treatment (p = 0.023, p = 0.029, respectively) (Fig 13). We also identified 2 genera that could significantly distinguish the two groups using the LEfSe algorithm (Linear discriminant analysis >3.0) (Fig 14).

Open in a separate window

Fig 12

Difference in taxonomy composition in GD and TxGD at genus level (n = 25).

GD, Graves’ disease; TxGD, Graves’ disease patients after 6 months treatment with anti-thyroid drug.

Open in a separate window

Fig 13

Differences in bacterial genus between GD and TxGD (n = 25).

GD, Graves’ disease; TxGD, Graves’ disease patients after 6 months treatment with anti-thyroid drug.

Open in a separate window

Fig 14

Distinguishment in bacterial genus between GD and TxGD by LEfSe algorithm (n = 25).

GD, Graves’ disease; TxGD, Graves’ disease patients after 6 months treatment with anti-thyroid drug.

Results of correlation analysis

1) Spearman correlation analysis at phylum level

At the phylum level, there was a positive association between Bacteroidota and SRS (correlation coefficient [CC] 0.59, p = 0.004), while Firmicutes exhibited a negative association with SRS (CC -0.73, p = 0.007).

2) Spearman correlation analysis at genus level

In Fig 15, the heatmap illustrates the correlation analysis of laboratory finding with microbiota at genus level before and after treatment. At baseline, several genera were associated with TSHR-Ab: Veillonella (CC 0.49, p = 0.007), Escheria-shigella (CC 0.50, p = 0.005), Ruminococcus gnavus (CC 0.55, p = 0.002), and Butyricicoccus (CC -0.42, p = 0.021). Moreover, unclassified_f_Lachnospiraceae (CC 0.71, p < 0.001), Veillonella (CC 0.40, p = 0.031), Escheria-shigella (CC 0.47, p = 0.011), Ruminococcus gnavus (CC 0.46, p = 0.011), and Anaerostipes (CC 0.61, p < 0.001) were associated with free T4 level. Some genera correlated with smoking pack years including Escheria-shigella (CC 0.45, p = 0.01), and Anaerostipes (CC 0.40, p = 0.03).

Open in a separate window

Fig 15

Heatmap of correlations analysis of gut microbiota.

Symbols (+) refer to p-values < 0.05. (a) Correlations before treatment. (b) Correlations after treatment.

Discussion

The present study investigated the gut microbiome in individuals newly diagnosed GD and observed shifts in microflora before and after treatment. The study unveiled a notable distinction in diversity between GD patients and healthy controls. Remarkably, post-treatment, the perturbed diversity in GD patients was restored.

The gut-associated lymphoid tissue harbors a diverse population of T-cells, including pro-inflammatory helper T (Th) cells and anti-inflammation regulatory T (Treg) cells. The intricate balance between these subsets is crucial for immune homeostasis. In the context of autoimmune diseases, the level of Th17 cells play a significant role [25, 26]. GD patients have been reported to have high levels of circulating Treg cell alongside an increased abundance of Th17 cells [27, 28]. Increased Th17 and impaired Treg responses are implicated in pathogenesis of GD. It has been reported that gut microbiome play a role in maintaining the dynamic balance between Th17 and Treg cells [29]. Numerous gut microorganisms have the capacity to produce short chain fatty acids (SCFAs) such as acetic acid, propionic acid, and butyric acid. These SCFAs have demonstrated regulatory effect on the generation of Th1 and Th17, while activating Treg cell activity [30–32]. By enhancing the integrity of the intestinal barrier function, SCFA may play a crucial role in shaping the development of GD.

Roseburia, known for its production of SCFAs such as acetic acid and butyric acid [33], was significantly decreased in GD patients, which is consistent with the previous research outcomes [34]. Furthermore, we observed an increase in Roseburia after ATD treatment, although this change did not reach statistical significant. As discussed earlier, the production of SCFAs has potential to induce Tregs cells and enhance their function [35]. For instance, in patients with ulcerative colitis, there was a discernible decrease in abundance of Roseburia bacteria, which is associated with concurrent reduction in concentrations of SCFAs [36, 37].

Indeed, the beneficial effect of SCFAs are not universal across all types. Unlike Roseburia, Bacteroides has the capacity to produce various SCFAs other than butyric acid, including succinic acid, propionic acid, and acetic acid. Importantly, unlike butyrate, these SCFAs do not induce mucin synthesis. The consequence of this is reduction in intestinal tight junctions as well as increased permeability of the gut mucosa [38, 39]. In GD patients, there was increase in the abundance of Bacteroides, which exhibited a subsequent decrease after treatment. As suggested, the elevated presence of Bacteroides may contribute to the release of pro-inflammatory factors beyond the intestine, potentially leading to immune dysfunction [13]. Similar Increase in Bacteroides have been documented in other autoimmune conditions, including rheumatoid arthritis and type 1 diabetes [38, 40]. Bacteroides may play an auxiliary role in the pathophysiology of GD. Notably, two studies have reported a significantly higher abundance of Bacteroides in GD patients compared to the healthy controls [13, 41]. However, conflicting findings have been reported in other studies [42, 43]. Therefore, further studies are needed to clarify the specific impact of Bacteroides on the thyroid.

The level of Subdoligranulum was found to be lower in GD patients, exhibiting a significant increase after treatment. Additionally, a notable negative correlation was observed between TSHR-Ab and free T4 level, implying its potential influence on the pathogenesis of GD. Likewise, several diseases—such as IBD—were associated with a lower abundance of butyrate producers like Subdoligranulum [44, 45].

Our data revealed an non-significant change in the level of Lactobacillus and Bifidobacterium, which is consistent with the findings in other studies [7, 13, 41]. Lactate-producing beacteria, Lactobacillus and Bifidobacterium, are typically considered to be beneficial microbiota in various contexts. Probiotics, which generally contain Lactobacillus and Bifidobacterium genera, has demonstrated positive effect, including a reduction in colorectal cancer and therapeutic potential against obesity [10, 46]. Certainly, while Lactobacillus and Bifidobacterium are generally considered beneficial, there are exceptions in specific diseases. For instance, increases in the abundance of Lactobacillus have been reported in autoimmune hepatitis [47] and Crohn’s disease [48]. Delcenseri et al. [49] suggested that probiotics may induce inflammation through the production of Th1 cells, leading to a negative influence on autoimmune diseases mediated by Th1 cells. As previously mentioned, our data indicated a decrease in Lactobacillus after ATD treatment, which is consistent with previous works [41, 50]. Furthermore, several studies have shown a positive correlation with Lactobacillus and thyroid antibody of TSHR-Ab and TPO Ab, suggesting its detrimental effect on GD pathogenesis [13, 41, 51].

By contrast, we speculate that Bifidobacterium may exert a probiotic effect on GD that differs from Lactobacillus. Our data showed an insignificant increase in Bifidobacterium after ATD treatment. Bifidobacterium supplementation has been recognized for its ability to inhibit the over-activation of Th17 [52].

At the phylum level, Firmicutes and Bacteroidota emerged as the dominant microbiota in individuals newly diagnosed with GD. The proportion of Firmicutes was lower in the GD group compared to HC, while the proportion of Bacteroidota was higher in the GD group than in the HC.

Notably, the contrasting finding in patients with Hashimoto’s thyroiditis or thyroid cancer, where Firmicutes were more abundant and Bacteroidota were less abundant [9, 53], suggesting a potential difference in pathogenesis of these diseases. The ratio of Firmicutes to Bacteroidetes has been proposed in several previous studies as indicator of the health status [54–56]. As demonstrated in our findings, the reduced ratio of Firmicutes to Bacteroidota was consistent with the results from other studies examining both GD and Graves’ ophthalmopathy (GO) patients [7, 13, 57, 58]. Interestingly, these imbalances were observed to be restored after ATD treatment.

Smoking is recognized for its pro-inflammatory effect and is a well-known as aggravating factor of GD and GO [59]. However, the precise mechanism by which smoking enhance immune pathways like Th2 responses leading to Graves’ disease, remain unclear [60]. In our study, we observed a positive association between Anaerostipes and smoking pack years, indicating that Anaerostipes was more abundant in smokers among GD patients. Consistently, Anaerostipes was linked with tobacco usage in IBD patients [61]. Additionally, our date revealed a significant association between Anaerostipes and TSHR-Ab. Quadbeck et al. [62] determined that smokers had a higher level of TSHR-Ab than nonsmokers. Based on these findings, we speculate that smoking might induce the pathogenic effect of Anaerostipes on autoimmunity, potentially aggravating both GD and GO.

In our study, we observed that alpha diversity was lower in GD patients compared to HC group, and this diversity was recovered after GD treatment. This aligns with findings from several prior studies that have consistently reported reduced richness of gut microbiome in GD patients compared to healthy controls [7, 13, 28, 41, 58, 63]. By contrast, there is one study [57] reporting a slight non-statistically significant increase in microbial diversity and richness compared to HC. However, considering the observed increase in alpha diversity after treatment and overall improvement in the microflora, it is reasonable to speculate that alpha diversity decreases in GD. Similar decreases in alpha diversity have also been observed in IBD [64] and autoimmune hepatitis patients [47].

We found a negative association of the phylum Firmicutes and a positive association of Bacteroidota with the symptoms of GD. Interestingly, Firmicutes and Bacteroidota did not show any correlation with free T4 level or T3 level, hormones typically associated with symptoms like nervousness, sweating, tremor, and hyperactivity. To our knowledge, this is the first research to report an association of GD symptoms with intestinal flora.

Typically, the diagnosis of GD involves the measurement of TSHR-Ab and radionuclide scanning of the thyroid is typically. However, it takes several days to receive the results of these assessments, particularly at primary physician clinics. Moreover, radionuclide thyroid scans can only be performed in a limited number of equipped hospitals and are not suitable for pregnant women. The advancement in microbiome research suggest that the non-invasive sampling of microbiome could offer an alternative for the convenient diagnosis of GD. The diagnostic efficacy of GD using stool samples showed a 96.6% sensitivity and an 82.2% specificity with an AUC value of 0.923. These values closely align with finding of Jiang et al. [13], which showed an AUC value of 0.811 using the same top nine species predicted by LEfSe analysis. However, it is essential to recognize that while this approach holds promise, it may not completely replace established methods. Microbiome profiling, based on 16S rRNA gene sequencing, also requires several days for completion, is influenced by various factors, and may encounter challenges related to data interpretation and standardization. Further development and refinement of microbiome profiling are essential for the potential replacement current assessment tools. Furthermore, it is important to note that while our classifier demonstrates strong discriminatory power between Graves’ disease patients and healthy controls, its true clinical utility lies in its ability to differentiate between Graves’ disease and other symptomatic thyroid disorders. Further validation of this utility is warranted, especially in cohorts exhibiting symptoms of hyperthyroidism.

Our study has several limitations. First, the composition and activities of the gut microbiome could be affected by various factors such as age, diet, race, and drugs [65]. Therefore, our results may be specific to the Korean population. Secondly, the GD patients in our study were not treated with the same medicine; as mentioned, methimazole, carbimazole, and propylthiouracil were used for the treatment of GD. Sun et al. [50] showed that ATD itself caused changes in the gut microbiome structure. Unfortunately, this study did not account for the specific interference of each ATD treatment. Thirdly, we chose to employ 16S rDNA sequencing in our research. This choice is significant given the increasing popularity of shotgun metagenomic sequencing for gut microbiota studies. While shotgun metagenomics sequencing provides a powerful tool for identifying microorganisms at the species or even strain level, it’s essential to recognize that 16S rDNA sequencing remains a widely used method in the field of gut microbiome research, even in studies on Graves’ disease [3, 7, 13, 34, 42, 43, 57, 66]. The utilization of 16S rDNA sequencing enhances comparability and contributes to a more comprehensive understanding of the gut microbiota. Fourth, the open database we utilized for this study did not provide access to the laboratory findings or clinical parameters of each healthy control. Therefore, we were unable to include the thyroid-related indicators for the healthy controls in our research. Lastly, the small sample size is a limitation. However, a notable aspect of our work is the comparison made before and after anti-thyroid drug treatment, which is a relatively rare perspective in the field. Despite the sample size limitation, we believe our research provide valuable insights into microbiome analysis in patients with Graves’ disease.

Conclusion

The pathogenesis of GD has not been fully elucidated. Our research not only pointed out the distinctive features of gut microbiota in GD patients but also demonstrated their alternation and subsequent restoration following ATD treatment. This supports the hypothesis that the gut microbiome may indeed play a role in the pathogenesis of GD.

Supporting information

Funding Statement

This study was supported by Research Fund of Seoul Saint Mary’s Hospital, The Catholic University of Korea. (KC19TESI0523).The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Data Availability

The data of Graves' disease patients generated for this work can be accessed at https://doi.org/10.5281/zenodo.11100304. The data of the healthy controls were sourced from an open public project named ‘Fecal microbiota in the Korean population’ of The European Molecular Biology Laboratory Nucleotide Sequence Database accessed at https://www.ebi.ac.uk/ena/browser/view/PRJEB35019.

References