Celiac disease (CD) is characterized by an immune-mediated enteropathy that affects the small intestine (Green and Cellier, 2007; Lebwohl et al., 2018). In predisposed individuals, gluten protein ingestion induces villous atrophy in the small bowel mucosa with lymphocyte infiltration. A systematic review noted that CD has a global prevalence of 1.4% and varies from 1.3 to 1.8% in different continents (Singh et al., 2018). Individuals carry the risk alleles [human leukocyte antigen (HLA)-DQ2, HLA-DQ8, and HLA-DQ7] and gluten ingestion triggering is necessary but insufficient for CD development (Fasano and Catassi, 2001). The gut microbiota and specific metabolites are considered cofactors in CD pathophysiology (Caminero et al., 2019). Therefore, investigating the interaction between host genetics and gut microbiota or metabolites is likely important in CD pathogenesis.

Numerous recent studies have indicated that intestinal flora changes are tightly correlated with autoimmune diseases (Chen et al., 2017; De Luca and Shoenfeld, 2019). As a key regulator of the gastrointestinal tract, gut bacteria influence the synthesis of many nutrients via insoluble fiber digestion, vitamin production, and bile acid metabolism (Olivares et al., 2011; Caminero et al., 2014; Valdes et al., 2018). The gut microbiota regulate gluten protein digestion, which can affect antigen development. Furthermore, human intestinal microbes produce a wide variety of metabolites, which act in the bloodstream and exert systemic effects on humans. Alterations in gut microbiota or metabolites can lead to biological changes in diseases, indicating that they might be causes and treatment targets of these diseases (Vacca et al., 2022; Zoghi et al., 2022).

Although links between gut microbiota and risk of CD development have been established based on observational studies, the causal nature of these links remains poorly understood. Some recent studies investigated the causal relationship between gut microbiota and autoimmune disease with a two-sample Mendelian randomization (MR) analysis (Xu et al., 2021; Zhang et al., 2021). MR analysis is an instrumental variable (IV) approach aimed at inferring a causal relationship between an exposure and an outcome from observational studies (Lee and Lim, 2019). Two-sample MR analysis is a major extension and improvement of the MR method. Genetic variants are associated with exposure as IVs to quantify causal relationships between the exposures and outcomes.

Here, we used a two-sample MR approach to verify the causal relationships among gut microbiota, metabolites, and CD. We identified several genetic variants linked to the bacterial composition that may drive CD pathogenesis. Our findings could form the basis for developing new research lines for diagnosing and treating CD.

The intestinal flora is considered a potential CD risk modulatory factor by interacting with metabolites and/or the host immune system (Valitutti et al., 2019). Microbiome and metabolomics studies over the past decades have greatly enhanced our understanding of the CD pathogenic mechanism (Caminero and Verdu, 2019; Leonard et al., 2020). Most documentation on the roles of the microbiome and metabolome in disease was from case–control studies aimed at determining the alterations that can be associated with specific diseases. Such studies can indicate associations but not causal relationships. Researchers are increasingly using MR to infer credible causal relationships between risk factors and disease outcomes. Here, we revealed suggestive associations between Lentisphaerae, Bifidobacterium, Coprobacter, and Subdoligranulum with CD. The metabolites 1-oleoylglycerophosphoethanolamine, 1-palmitoylglycerophosphoethanolamine, 1,6-anhydroglucose, PAGln, tryptophan betaine, 10-undecenoate, and tyrosine were also related to CD risk.

Bifidobacterium is the major microbe that primarily colonizes the human gut and is believed to influence the development of multiple autoimmune diseases. Olivares et al. reported that breastfed infants at genetic risk of CD had a decreased abundance of Bifidobacterium species in their feces (Olivares et al., 2015), suggesting its protective effect, while opposite results were reported for adult patients (Nistal et al., 2012). Recent studies demonstrated that Bifidobacterium-based probiotic interventions effectively delayed CD progression by reducing tumor necrosis factor-α production (Klemenak et al., 2015). Contrastingly, Smecuol et al. (2013) reported opposite observations. In our study, an increased relative abundance of Bifidobacterium was causally linked to the risk of CD, indicating its detrimental effect on CD. Several factors may have contributed to the inconsistent results as described above, on the one hand, considering that different bifidobacterium strains have different effects on the disease, for example, a particular bacterial group CMS-H004 may aggravate intestinal damage, while B. breve BR03 and B632 may alleviate intestinal damage. We hypothesized that bifidobacterium abundance would affect outcomes. Perhaps low abundance of probiotics can improve disease outcomes, but when abundance is high, it can aggravate disease damage. On the other hand, an increase in the relative abundance of Bifidobacteria will necessarily decrease the relative abundance of other probiotics, thereby weakening the beneficial effects of other probiotics. A more specific level, such as species, is crucial to provide more precise clinical guidance and direct mechanism research.

The phylum Lentisphaerae was proposed in 2004 after the isolation of two marine strains (Cho et al., 2004) and is a member of the Planctomycetes–Verrucomicrobia–Chlamydiae superphylum. Decreased abundance of Lentisphaerae was observed in patients with post-traumatic stress disorder (Hemmings et al., 2017), multiple sclerosis (Castillo-Álvarez et al., 2021), and autoimmune hepatitis (Kodio et al., 2019; Lou et al., 2020), while an increased abundance of this strain was observed in Blastocystis-colonized children (Kodio et al., 2019) and rosacea patients (Chen et al., 2021). In this study, a greater abundance of Lentisphaerae (OR = 0.798, 95% CI: 0.648–0.982, p = 0.034) was significantly related to lower CD risk. However, similar results have not been reported between the abundance of Lentisphaerae and risk of CD. Therefore, such results should be treated with caution. Subdoligranulum is a butyrate producer that is essential for gut health (Chassard et al., 2014), where it affects patients with necrotizing enterocolitis by influencing butyrate production (Lin et al., 2021). Subdoligranulum was significantly decreased in patients with chronic spontaneous urticaria and had potential diagnostic value for such patients (Liu et al., 2021). Subdoligranulum was also decreased in participants with food allergy (Abdel-Gadir et al., 2019). Abundances of Subdoligranulum species were markedly decreased in patients with inflammatory bowel disease (IBD) and extensively associated with concentrations of IBD-linked metabolites, which may lead to new possibilities for intervention (Sokol et al., 2008). In the present study, Subdoligranulum had a protective effect against CD. Nevertheless, the detailed mechanism remains to be illustrated in further studies. The reverse MR analysis revealed that CD was causally associated with Methanobacteria. The Methanogen archaea include Methanosphaera stadtmanae, Methanobrevibacter smithii, and M. oralis microbiota, which form a syntrophic relationship with other microbiota (Dridi et al., 2009). M. smithii prevalence in the intestine and gastrointestinal disorders are apparently linked (Aminov, 2013; Ghavami et al., 2018). However, no study stating its role in celiac disease has yet been reported.

Microbe-derived metabolites are crucial in host–microbe interactions. In the present study, 10-undecenoate and tyrosine were protective factors against CD. Previous MR analysis determined that 10-undecenoate shared an important positive causal relationship with rheumatoid arthritis (Yuan et al., 2022). However, no clinical or experimental studies supported the claim that CD affects blood 10-undecenoate levels. In a sense, this research provides a theoretical basis for follow-up research. Tyrosine is an essential amino acid that participates in host–microbiota crosstalk. Evidence from the other studies identified tyrosine levels were decreased after weight loss, indicating that it is a body mass index-associated amino acid (Geidenstam et al., 2017). Furthermore, tyrosine was associated with insulin resistance (Hellmuth et al., 2016) and the level of tyrosine was identified as a strong predictor of type 2 diabetes mellitus incidence (Tillin et al., 2015). It has been showed previously that gut bacteria metabolize tyrosine to p-cresol phenol sulfate and tyramine, which decreases intestinal epithelial cell viability and intestinal integrity (McCall et al., 2009). Interestingly, however, significantly lower tyrosine levels was observed in potential CD patients compared with CD patients (Upadhyay et al., 2022). Hence, Future validation is needed in cohort with more samples. PAGln was described as a gut microbiota metabolite fermented from dietary phenylalanine (Poesen et al., 2016). The microbial porA gene converts dietary phenylalanine into phenylacetic acid, with subsequent host generation of PAGln (Nemet et al., 2020). An elevated PAGln level may put a person at greater risk for acute myocardial infarction and type 2 diabetes (Poesen et al., 2016). Our results provided clinical evidence that the increased blood PAGln levels were also causally associated with CD occurrence.

3-IS is an important intercellular signal in microbial communities that is an indole metabolism byproduct produced by tryptophanase-expressing bacteria (Lee and Lee, 2010). Low urinary 3-IS levels were associated with higher transplant-related mortality and a worse overall survival rate in bone marrow-transplanted patients (Weber et al., 2015). Another study reported that during renal failure, high 3-IS levels were detected in patients with chronic kidney disease and cardiovascular disease (Vanholder et al., 2014). Moreover, it was previously found that 3-IS promoted the differentiation of anti-inflammatory and immunosuppressive dendritic cells (Ghimire et al., 2018). We assumed that the increased 3-IS levels might aggravate the inflammatory responses in patients with CD. Therefore, more in-depth investigations are necessary to determine the mechanisms by which metabolites promote CD occurrence.

García-Santisteban et al. (2020) and Xu et al. (2021) performed MR analyses to explore the causal relationships between gut microbiota and CD. Our study differed from theirs in two aspects. First, the identification of a causal relationship between metabolism and CD in the present study enabled a more comprehensive and integrated analysis of CD etiology. Second, compared to the study by García-Santisteban et al. (2020) which used gut microbiota data with a relatively small sample size (N = 1,126 twin pairs), we used microbiota data with a much larger sample size (N = 18,340) and more stringent IV screening criteria. Nevertheless, our study has several limitations. First, we failed to identify overlapping participants in the exposure and outcome of the GWAS summary data. Second, the bacterial taxa were not at a more specialized level, such as species or strain level. Third, the majority of participants in the GWAS summary data used in this study had European ancestry, but a small percentage of gut microbiota data was from other ethnicities, which might have resulted in the deviation of statistical results. Finally, owing to the limited variance explained by the SNPs or the limited sample size of the GWAS results, some of our MR analyses did not have sufficient power to detect small effects. Therefore, the results of the present study provide suggestive evidence of association. Given the limitations of the GWAS data, prospective studies are necessary to explore the underlying mechanisms in order to develop effective and feasible treatment strategies.

In conclusion, our findings supported the hypothesis that alterations in the abundance of gut microbiota and concentrations of metabolites are causally linked to the risk of CD. Differences in diet, habits, and lifestyle may contribute to individual susceptibility. Moreover, we identified the specific bacterial taxa and metabolites engaged in CD development. Nevertheless, the mechanisms behind these findings warrant further investigation.

The original contributions presented in the study are included in the article/Supplementary material, further inquiries can be directed to the corresponding author.

TL designed the research and drafted the manuscript. TL, YF, CW, and TS collected and analyzed the data. AA and MZ performed the literature search. FG supervised the study. All authors were involved in writing the manuscript and contributed to the article and approved the submitted version.

This study was supported by the Natural Science Foundation of Xinjiang Uygur Autonomous Region (Nos. 2021D01C200 and 20200103).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.