Uncovering the gut microbiota’s role in temporomandibular joint disorders: A bidirectional Mendelian randomization study

Abstract

Previous studies have revealed the association between gut microbiota (GM) and temporomandibular joint inflammation, which leads to temporomandibular joint disorders (TMD). However, the causality of the associations remains unknown. This Mendelian randomization (MR) study aims to clarify the causal relationships between GM and TMD. We employed a two-sample MR approach to analyze GM data from the MiBioGen consortium, including 18,340 participants and 211 taxa, and TMD data from the FinnGen consortium R10 release, including 6314 cases and 222,498 controls. Various MR methods, including inverse-variance weighted (IVW), MR-Egger, and weighted median analyses, alongside comprehensive sensitivity analyses, were used to assess causality. After comprehensive sensitivity analyses, the study found causal links between 6 certain GM at the genus level and TMD, including Eubacterium fissicatena group (IVW’s odds ratio [OR] = 1.18; 95% confidence interval [CI]: 1.02–1.37; P = .027), Catenibacterium (IVW’s OR = 1.33; 95% CI: 1.11–1.59; P = .002), Oxalobacter (IVW’s OR = 0.85; 95% CI: 0.75–0.97; P = .013), Ruminococcaceae NK4A214 (IVW’s OR = 0.80; 95% CI: 0.66–0.99; P = .036), and Senegalimassilia (IVW’s OR = 0.75; 95% CI: 0.59–0.96; P = .024). Following replication verification analysis, Catenibacterium (FDR = .04) at the genus level were positively correlated to the risk of TMD. The reverse MR analysis revealed that TMD had no significant effect on the 6 GMs. The findings of this MR study support a strong and negative causal association between genus Catenibacterium and TMD, highlighting potential targets for the prevention and treatment of TMD.

1. Introduction

Temporomandibular joint disorders (TMD), characterized by orofacial pain, joint sounds, restricted mouth opening, refer to a heterogeneous group involving the temporomandibular joints (TMJ) and associated surrounding structures.^[1,2] As the chief complaint of patients with TMD, centralization of the pain may result from inflammatory and/or genetic alterations in the nervous system.^[3] The causal relationship between TMJ inflammatory episodes and TMJ pain has been previously illustrated.^[4,5] Evidence from epidemiological studies suggests the prevalence of TMD up to 10% to 15% within general population, whereas females are more generally affected than males by TMD.^[2,6,7] Among medical students in Zunyi Medical University in China, the prevalence of TMD is 31.7%.^[8] Through the World Health Organization Quality of Life Bref, TMD subjects were found worse health and quality of life due to pain/disability, including arthralgia, muscle disorders, and arthritis.^[9] Therefore, figuring out the pathogenesis of TMD has gradually become a significant public health goal.^[10]

Currently, the pathogenesis of TMD is not yet clear. The research on the association between TMD and gut microbiota (GM) has always been of concern. The GM perturbation has been reported as a critical factor for the development of TMJ inflammation.^[11] Generally, the disorders of the GM cause various diseases, including metabolic, neurologic, cardiovascular, and cerebrovascular diseases.^[12] The GM plays a critical role in various types of chronic pain, including inflammatory pain, which makes it a potentially significant factor of TMD.^[13] There is an experimental study that conveyed decreases gut bacteria Bacteroidetes and Lachnospiraceae and the relevant produced short-chain fatty acids fatty acids (SCFAs) in 3 mice with induced TMJ inflammatory pain, which were compared with the saline control group.^[11] The fecal microbiota transplantation successfully recovered decreased gut bacteria Bacteroidetes and Lachnospiraceae.^[11] Therefore, the GM perturbation is crucial for TMJ inflammation and recovering the targeted GM to normal levels can be used to treat TMD. However, there are no studies directly showing the changes in the abundance of GM based on observational studies or randomized controlled trials.

It is difficult to evaluate the causal relationship between the GM and TMD. The narrow model building range and relatively small sample size may lead to biased results. Randomized controlled trials have strict limitations and therefore are impractical due to the GM diversity and composition. Mendelian randomization (MR) has become a novel method by which genetic variation function as instrumental variables (IVs) to infer the causal associations and direction.^[14] Two-sample MR (TSMR) connects exposure with disease outcomes based on single nucleotide polymorphism (SNP)-exposure and SNP-outcome associations in 2 independent genome-wide association studies (GWASs).^[15] Compared with observational studies, MR analysis can avoid reverse causation and residual confounding to some extent.^[16] We conducted a TSMR analysis to research on the causal associations between the GM and TMD by utilizing GWAS summary datasets from the MiBioGen and FinnGen consortium, and clarified specific related bacterial taxa as potential targets for prevention and treatment. In addition, this research results may be beneficial to the understanding of the complex causes of TMD.

2. Materials and methods

2.1. Study design

TSMR utilizes SNPs as IVs to estimate causal effects in 2 independent populations.^[17] In view of the publicly available databases we used, informed consent or ethical approval was not required. The complete flowchart of our bidirectional TSMR is illustrated in Figure 1. Our MR analysis is based on 3 major assumptions for purpose of reliable results. Firstly, there is a strong association between IVs and the GM (exposure).^[18] Secondly, IVs must be independent of confounders that influence the GM and TMD (outcome). Thirdly, IVs only affect TMD through the GM in absence of alternative mechanisms.^[19,20]

Figure 1.

The whole flowchart of MR analysis in our research. GM = gut microbiota; IV = instrumental variable; IVW = inverse-variance; LD = linkage disequilibrium; MR = Mendelian randomization; MR-PRESSO = MR-pleiotropy residual sum and outlier; SNP = single nucleotide polymorphism; TMD = temporomandibular joint disorders; WM = weighted median.

2.2. Data sources

The studies involving all datasets were publicly available and were approved by respective institutional review boards. No new ethical approval and informed consent was required. In our study, the IVs selected were extracted from the MiBioGen consortium, involving 18,340 participants and 211 taxa in 24 cohorts dominated by European descent mainly from Europe, North American, East Asia, and Middle-Eastern (https://mibiogen.gcc.rug.nl/). Twenty cohorts included European (16 cohorts, N = 13,266), American Hispanic/Latin (1 cohort N = 1097), East Asian (1 cohort, N = 811), Middle-Eastern (1 cohort, N = 481), African American (1 cohort, N = 114), and multi-ancestry (4 cohort, N = 2571).^[21] We opted for the data on the GM at the genus level, focusing on the smallest classification. The average abundance of the GM was >1% in 131 taxa at the genus level. After excluding 12 unidentified genera, we included the 119 known GM genera in the exposure. We obtained data for TMD from the FinnGen consortium R10 release in discovery stage (https://r10.finngen.fi/), which includes 6314 cases and 222,498 controls.^[22] All TMD subjects included in the FinnGen data are of European ancestry. The cases of TMD are classified according to the revised International Classification of Disease (ICD-10) codes K07.6 (https://risteys.finngen.fi/endpoints/TEMPOROMANDIB).

2.3. Genetic IVs

Significant SNPs related to the GM at the genus level were screened from the MiBioGen consortium as IVs. Firstly, SNPs selected should be related to the GM at the genome-wide significance level (P < 1e-5) in Table S1, Supplemental Digital Content, https://links.lww.com/MD/P113. Dentofacial anomalies, occlusal interference, cold stimulation, and other TMD risk factors were examined to prevent pleiotropic effects on MR results.^[23,24] Secondly, the genotype samples in the 1000 European genomes project were referred for linkage disequilibrium analysis. The process was performed with r² < 0.001 and a window size = 10,000 kb in order that SNPs selected were not in linkage disequilibrium and the independence of each SNP from different bacterial taxa at the genus level was ensured. Thirdly, to ensure the MR assumptions, we processed all selected SNPs in the PhenoScanner V2 website to eliminate potential confounders. Fourthly, for the SNPs we selected, the F statistics were ensured >10 to exclude weak IVs.^[25] The calculation formula is $\mathbf{F}={\mathbf{R}}^2\frac{\mathbf{N}-\mathbf{k}-1}{\mathbf{k}(1-{\mathbf{R}}^2)}$, where R² represents the proportion of variation explained by SNPs we selected and N reflects the sample size. Fifthly, palindromic and incompatible SNPs were eliminated by means of harmonizing outcome and exposure data.

2.4. MR analyses

In our study, we estimated the causal associations with the inverse-variance weighted (IVW), MR-Egger, weighted median (WM), weighted mode, and simple mode method. IVW was considered as the primary analysis method, where each IV was estimated respectively. The IVW method uses the ratio method to calculate the causal effect of each IVs, and summarized all estimated values.^[26] The MR-Egger method was proposed by Bowden. This method assumes the existence of horizontal pleiotropy of IVs and estimates the size of its pleiotropy through the intercept term.^[27] When P < .05, it means that there exists horizontal pleiotropy. The WM method could avoid the interference of outliers on the results by means of calculating a WM estimator.^[28] Benjamini–Hochberg FDR correction was used in the replication verification analysis, whose P-value was set at .05. The P < .05 was regarded as statistical significance, but P > .05 was regarded as suggestive associations.^[29]

2.5. Sensitivity analyses

Our sensitivity analysis was performed with MR pleiotropy residual sum and outlier (MR-PRESSO), MR-Egger intercept test, and Cochran Q test.^[30] Furthermore, leave-one-out analysis was used in the sensitivity analysis. MR-Egger intercept test and MR-PRESSO were performed to judge horizontal multiplicity. Horizontal multiplicity exists when the intercept term was statistically significant. MR-PRESSO was conducted to indicate and correct horizontal multiplicity if the ratio of multiplicity IVs was no more than 50%. Cochran Q was aiming to confirm heterogeneity of MR-Egger and IVW methods, while leave-one-out analysis was conducted with individual SNP removed one by one to identify strongly influential SNPs. All the TSMR analyses were run with R Software (www.R-project.org), by which we used the “TwoSampleMR” package and the “MR-PRESSO” package. In the Cochran Q test and MR-Egger intercept test, statistically significant difference was considered at P < .05. In the leave-one-out analysis, P < .05 represented significant difference. MR-PRESSO test P < .05 means horizontal multiplicity.

3. Results

3.1. Characteristics of selected SNPs

There are the study population, sample size, the ICD-10 diagnostic criteria and gender composition in Table 1. In general, 53 genome-wide significant SNPs for 119 known GM genera were identified. All SNPs we select show strong association, whose F statistic>10 (Table S2, Supplemental Digital Content, https://links.lww.com/MD/P113).

Table 1.

Description of TMD and GM GWAS samples used in our research.

Phenotypes	Type	Consortium	Sample size	ICD-10	Populations	Year
TMD	Outcome	FinnGen R10	6314 cases and 222,498 controls	K07.6	European, American Hispanic/Latin, East Asian, Middle-Eastern, African American	2024
Human GM	Exposure	MiBioGen	18,340 individuals	-	Finnish individuals	2021

GM = gut microbiota, GWAS = genome-wide association study, TMD = temporomandibular joint disorders.

3.2. Causal effects of the GM and TMD

Six GM were identified correlation with TMD (Table 2, Fig. 2; Table S3, Supplemental Digital Content, https://links.lww.com/MD/P113). Our IVW results revealed 3 GM including Eubacterium fissicatena group (IVW odds ratio [OR] = 1.18; 95% confidence interval [CI]: 1.02–1.37; P = .027), Catenibacterium (IVW OR = 1.33; 95% CI: 1.11–1.59; P = .002), and Coprobacter (IVW OR = 1.23; 95% CI: 1.06–1.43; P = .007) at the genus level were positively correlated to the risk of TMD. And the 3 GM including Oxalobacter (IVW OR = 0.85; 95% CI: 0.75–0.97; P = .013), Ruminococcaceae NK4A214 (IVW OR = 0.80; 95% CI: 0.66–0.99; P = .036), and Senegalimassilia (IVW OR = 0.75; 95% CI: 0.59–0.96; P = .024) at the genus level were negatively correlated to the risk of TMD. In the WM results, Catenibacterium at the genus level were positively correlated to the risk of TMD (WM OR = 1.32, 95% CI: 1.05–1.67, P = .018). In the MR-Egger results, Oxalobacter at the genus level were positively correlated to the risk of TMD (WM OR = 0.49, 95% CI: 0.28–0.87, P = .036). The reverse analysis in our MR research revealed that TMD had no effect on the 6 GMs (Table S4, Supplemental Digital Content, https://links.lww.com/MD/P113).

Table 2.

Significant IVW results in the causal estimations of GM on TMD in our MR study.

Bacterial taxa (exposure)	Outcome	No. SNP	MR method	OR (95% CI)	P-value	FDR	Beta	SE
Catenibacterium	TMD	4	IVW	1.33 (1.11–1.59)	.002	0.04	0.28	0.09
Coprobacter	TMD	11	IVW	1.23 (1.06–1.43)	.007	0.07	0.21	0.08
Eubacterium fissicatena group	TMD	9	IVW	1.18 (1.02–1.37)	.027	0.09	0.17	0.08
Oxalobacter	TMD	11	IVW	0.85 (0.75–0.97)	.013	0.12	-0.16	0.06
Ruminococcaceae NK4A214	TMD	13	IVW	0.80 (0.66–0.99)	.036	0.11	-0.22	0.10
Senegalimassilia	TMD	5	IVW	0.75 (0.59–0.96)	.024	0.12	-0.28	0.13

CI = confidence interval, GM = gut microbiota, IVW = inverse-variance, MR = Mendelian randomization, OR = odds ratio, SNP = single nucleotide polymorphism, TMD = temporomandibular joint disorders.

Figure 2.

Scatter plots of 5 MR tests in 6 causal associations from 6 GM to TMD. The causal estimation was represented by the slope of the line. GM = gut microbiota; MR = Mendelian randomization; SNP = single nucleotide polymorphism; TMD = temporomandibular joint disorders.

There was no pleiotropy detected in the causal estimation between TMD and the specific related bacterial taxa following MR-PRESSO test and MR-Egger intercept analysis after excluding outliers. In Cochran Q test, no heterogeneity was detected. The results of the leave-one-out analysis illustrated there was no potential SNP driving causality (Fig. 3; Table S5, Supplemental Digital Content, https://links.lww.com/MD/P113).

Figure 3.

Leave-one-out analysis of the causal associations 6 GM at the genus level on TMD. Red lines corresponded to estimations in the IVW test. GM = gut microbiota; IVW = inverse-variance; MR = Mendelian randomization; TMD = temporomandibular joint disorders.

4. Discussion

In the study, we conducted a bidirectional TSMR to evaluate the potentially causal associations between the GM and TMD. The study identified genetic liability to some specific related bacterial taxa at the genus level causally associated with TMD.

TMD usually presents with orofacial and TMJ inflammatory pain.^[5,31] TMD pain can easily cause altered neuronal processing in the central nervous system over time, therefore leading to central sensitization and neuronal hyperexcitability, but also affecting perception of TMD pain.^[32] In the etiology of temporomandibular disorders, the bio-psycho-social model of pain is widely recognized. There is evidence that the GM interacts with central nervous system through the gut–brain axis.^[33] Therefore, we speculated that the GM is involved in TMD-related neuropathic pain. Previous studies showed decreases gut bacteria Bacteroidetes and Lachnospiraceae and the relevant produced SCFAs in mice with induced TMJ inflammatory pain. Subsequently, the microglia in trigeminal nociceptive system were overactivated, which promoted the occurrence and development of TMD.^[11] Recovering gut microbiome has become a novel approach for relieving and treating TMD. Our study aimed to provide more GM targets for treatment of TMD.

Our results indicated that Coprobacter, E fissicatena group, and Catenibacterium at the genus level were positively correlated to the risk of TMD. Chun Yang et al showed that the level of Ruminococcaceae and Fluviicola were regulated in mice with chronic inflammatory pain induced in the left hind paw.^[34] The results are partly consistent with the findings in our research. Fluviicola, as well as Coprobacter in our results, were classified within the phylum Bacteroidetes, which produced SFCAs as signaling molecules to provide nerve cells with energy, including acetic and propionic acids.^[35] SFCAs supposedly affected psychology and emotion through metabolism.^[36] Previous studies showed decreases gut bacteria Bacteroidetes and the relevant produced SFCAs in mice with induced TMJ inflammatory pain, suggesting the mentioned molecules involved in TMJ inflammatory pain and TMD.^[11] An increase in the abundance of Catenibacterium has been observed in patients with autism spectrum disorders, and Catenibacterium was therefore considered as a potential biomarker.^[37] In the study, changes in Catenibacterium abundance could also cause dysregulation of the gut-immune-brain axis, influencing the pathogenesis of TMD. Another study indicated the Western diet could increase Catenibacterium abundance in the GM.^[38] Catenibacterium, Ruminococcaceae NK4A214, and Senegalimassilia have been reported to be associated with psychological distress, including anxiety, sleep disorders, and depression.^[39–41] This suggested that in the process of GM-mediated TMD, not only pain, but also anxiety and depression should be considered, which was consistent with the function of SCFAs and the bio-psycho-social model of pain in TMD etiology. Moreover, the overactivated microglia in trigeminal nociceptive system were also thought to be involved in the development of TMD.^[11] Microglia not only maintained central nervous system homeostasis and diseases, but also participated in endogenous immune responses. The alterations in GM could impact the activity of microglia in the brain,^[36] hinting at a mechanism for influences the GM on TMD through the gut–brain axis. Ruminococcaceae NK4A214, a member of Ruminococcaceae Family, helped degrade fibers.^[42] Cellobiose and cellulose were fermented with the help of Ruminococcaceae in the rumen to produce cellulase, consequently degrading the fiber. The intestinal permeability was negatively associated with the level of Ruminococcaceae, indicating that the decreased Ruminococcaceae likely cause intestinal inflammation.^[43] Intestinal inflammation made it possible for some inflammatory factors such as tumor necrosis factor-α and other harmful factors to enter the blood, promoting the alterations in peripheral and central inflammatory factors and central inflammatory responses.^[44] Besides, oxidative stress was thought to be involved in TMD through intra-articular damage.^[45] The imbalance of GM would affect the antioxidant function in vivo and lead to oxidative stress,^[46] which was reasonably speculated to be another potential mechanism of GM affecting TMD.

Previous studies revealed E fissicatena group was correlated to the risk of psoriasis and obesity.^[47,48] A lower abundance of butyrate-producing bacteria E fissicatena group was colonized in the GM of periodontitis group by comparison with the control group.^[49] In our results, E fissicatena group at the genus level was a possible risk factor for TMD, though^[16] another study found that there is no causal association between periodontitis and TMD.^[23] Oxalobacter played a significant role in oxalate-degrading to treat calcium oxalate kidney stone disease.^[50] Additionally, Oxalobacter was detected to be positively correlated to cardiovascular disease, inflammatory bowel disease and Graves’ disease.^[51–53]

SCFAs, as metabolic byproducts of the gut microbiota, play a pivotal role in modulating host immune responses and inflammation-related diseases.^[54] Particularly, butyrate, a type of SCFA, has been demonstrated to reduce the production of inflammatory mediators by enhancing intestinal barrier function and modulating the activity of immune cells, thereby affecting the local and systemic inflammatory status of TMD.^[55] Moreover, SCFAs exert their effects through G protein-coupled receptor GPR109A, which is a receptor for bacterial fermentation product butyrate and acts as a tumor suppressor in the colon. Recent research has uncovered that the gut microbiota communicates with the central nervous system through the gut–brain axis, influencing pain perception and neuroinflammation. This discovery underscores the role of the gut–brain axis in modulating inflammatory pain, providing us with a new perspective on how the gut microbiota may affect TMD. Specific gut bacteria may alter the levels of neurotransmitters through the gut–brain axis, thereby impacting pain signals associated with TMD and central sensitization. Furthermore, certain groups of gut bacteria, such as Bifidobacterium and Lactobacillus, may influence the pathogenesis of TMD by activating immune cells in the gut, like regulatory T cells, reducing excessive immune responses.^[56] The role of regulatory T cells in controlling TH17 responses has been established, which may be related to the immune modulation of TMD. In general, these bacteria may affect the development of TMD by affecting SFCAs, microglia, inflammatory factors, and oxidative stress. Secondly, the metabolism of tryptophan by the gut microbiota is also associated with the development of TMD. Tryptophan, an essential amino acid, can be converted by the gut microbiota into various molecules, such as indoles and their derivatives, which are thought to be related to the pathogenesis of T2DM.^[57] Indoles stimulate the secretion of GLP-1, leading to insulin release and a reduction in blood glucose levels, while IPA exhibits anti-inflammatory and antimicrobial properties by acting on the aryl hydrocarbon receptor (AhR). Furthermore, trimethylamine N-oxide (TMAO), produced by the gut microbiota, is related to the pathogenesis of T2DM and its associated complications.^[58] TMAO may promote the development of T2DM by promoting insulin resistance, impairing glucose tolerance, and inducing inflammation. TMAO enhances atherosclerosis by upregulating the scavenger receptors CD36 and SR-A1 in macrophages, and intensifying the accumulation of cholesterol in macrophages and the formation of foam cells. In summary, the gut microbiota may influence the development of TMD through its metabolic products and signaling pathways. These pathways include the production of SCFAs, tryptophan metabolism, and the effects of TMAO, which may affect the pathogenesis of TMD by regulating inflammatory responses and immune cell activity. Unfortunately, no definitive research evidence or mechanisms has been reported on the influence specific bacterial taxa on TMD pathogenesis. These findings provide new insights into how the gut microbiota can affect TMD and may offer potential targets for future therapeutic strategies.

One drawback of the MR design is that it can only be applied to risk factors that have the right genetic variation. Genetic variants typically have a small effect on most risk factors (i.e., they explain a small portion of the variation), which can lead to lower statistical power of MR analyses and a risk of false-negative results.^[59,60] This paper uses multiple genetic variants associated with risk factors as IVs to increase the proportion of variance interpretation, thereby improving statistical power. In our study, we concentrated on the causal relationship between the GM and TMD, eliminating interference of confounding factors and performing reverse causality analysis. The data on the GM we selected were derived from the largest GM GWAS, ensuring the strength of IVs. Comprehensive sensitivity analyses were used to detect and rule out pleiotropy, heterogeneity and sensitivity. We acknowledge that although the MR approach helps to control for some unmeasured confounders, it does not completely eliminate the influence of all confounders. In particular, dietary and lifestyle may indirectly affect the risk of TMD by affecting GM composition and function. In addition, poor oral hygiene leads to oral microbial imbalance, which could influence GM and TMD simultaneously. The lack of detailed raw information in the GWAS data prevented the further subgroup analyses to explain the confounding factors. We highlight the importance of future studies that need to more directly measure and control for these confounders to more accurately assess the relationship between GM and TMD.^[61,62] The other several study limitations must be pointed out. One of the most important limitations is that all the GWAS data in our study merely comes from European population, but not American, African and Asian populations. Therefore, our findings could not be confirmed in non-European populations. GM compositions are known to vary across ethnicities and geographical locations, so it is necessary to conduct future studies in more diverse populations. Secondly, our analysis on specific related bacterial taxa was at the genus level, affecting the accuracy of causality. Completed analysis at the other levels is needed in the future. Thirdly, the IVs we selected were associated with the relative abundance of the specific related bacterial taxa of the GM, rather than the functional metabolic pathways of the GM. Finally, we focus on individual specific related bacterial taxa, but largely ignore balance and disbalance of the GM.

5. Conclusion

The findings of this MR study support a strong and negative causal association between Genus Catenibacterium and TMD. The research also puts forward the hypothesis that E fissicatena group and Coprobacter might be potentially risk factors for TMD, while Oxalobacter, Ruminococcaceae NK4A214, and Senegalimassilia played a protective role in TMD. In summary, the study enriches the understanding of the complex causes of TMD, and provides ideas for targeted treatment of TMD through microbiota.

Acknowledgments

The authors thank the participants of all cohorts included in these studies. Special thanks to the residents of the Huadong Hospital affiliated Fudan University who supported the fieldwork of this study. The authors acknowledge the reviewers of this manuscript.

Author contributions

Conceptualization: Tianyi Ni, Zhao Han.

Data curation: Tianyi Ni, Ziyu Shen, Hekai Shi.

Formal analysis: Tianyi Ni, Ziyu Shen, Hekai Shi.

Funding acquisition: Zhao Han.

Investigation: Tianyi Ni, Ziyu Shen, Hekai Shi.

Methodology: Tianyi Ni, Ziyu Shen.

Project administration: Zhao Han.

Resources: Tianyi Ni.

Software: Tianyi Ni.

Supervision: Zhao Han.

Visualization: Ziyu Shen.

Writing – original draft: Tianyi Ni, Ziyu Shen, Hekai Shi.

Writing – review & editing: Tianyi Ni, Zhao Han.