Gut Mucosal Microbiome Is Perturbed in Rheumatoid Arthritis Mice and Partly Restored after TDAG8 Deficiency or Suppression by Salicylanilide Derivative

Rheumatoid arthritis (RA), an autoimmune disease, is characterized by chronic joint inflammation and pain. We previously found that the deletion of T-cell death-associated gene 8 (TDAG8) significantly reduces disease severity and pain in RA mice. Whether it is by modulating gut microbiota remains unclear. In this study, 64 intestinal samples of feces, cecal content, and cecal mucus from the complete Freund’s adjuvant-induced arthritis mouse models were compared. The α- and β-diversity indices of the microbiome were significantly lower in RA mice. Cecal mucus showed a higher ratio of Firmicutes to Bacteroidetes in RA than healthy mice, suggesting the ratio could serve as an RA indicator. Four core genera, Eubacterium_Ventriosum, Alloprevotella, Rikenella, and Treponema, were reduced in content in both feces and mucus RA samples, and could serve microbial markers representing RA progression. TDAG8 deficiency decreased the abundance of proinflammation-related Eubacterium_Xylanophilum, Clostridia, Ruminococcus, Paraprevotella, and Rikenellaceae, which reduced local mucosal inflammation to relieve RA disease severity and pain. The pharmacological block of the TDAG8 function by a salicylanilide derivative partly restored the RA microbiome to a healthy composition. These findings provide a further understanding of specific bacteria interactions with host gut mucus in the RA model. The modulation by TDAG8 on particular bacteria can facilitate microbiota-based therapy.


Introduction
Rheumatoid arthritis (RA) is a common, autoimmune, inflammatory, and chronic disease that affects nearly 1% of the adult population worldwide [1,2]. RA is characterized by its severely progressive disability, systemic complications, early death, and health expenditure terms. The pathogenesis of RA is complex and leads to the destruction of both cartilaginous and bony elements of the joint. The dysregulated inflammatory processes in the synovium of the joint are often accompanied by ongoing pain and increased pain during movement. The etiology of RA is ambiguous; the initiation of RA seems to result from both genetic and environmental causes [1,2].
Various risk factors have been indicated as potential causes for RA, and microorganisms have recently been of interest as a risk factor. The overrepresentation of some microorganisms in the intestines could be related to RA morbidity. In fact, fluctuations in bacterial content might lead to altered levels of metabolites that promote joint inflammation [3][4][5]. N is the number of operational taxonomical units. CT, RA, CC and CM indicate wild-type, rheumatoid arthritis, cecal content and cecal mucosa, respectively. Mice (8-12 weeks old) were injected with 5 µL of 100% CFA (5 µg) in the right ankle joint (ipsilateral joint) four times at 1-week intervals, followed by mechanical or thermal behavioral tests. a Weeks since experiment start. b Number of sequences.
The complete and resampled datasets were used to calculate the Bray-Curtis dissimilarity, and then the Mantel test was used to compare both datasets. The results showed a significant correlation (correlation coefficient 0.9802, p = 0.001) and indicated a difference in the number of sequences per sample, causing no effects in the analysis. In addition, the α-diversity revealed no major differences for both matrices (Table 1). Analyses of microbial communities revealed differences in richness (observed OTUs) between RA and healthy mice ( Figure 1A). To standardize the microbiome measures, a minimum of 31,794 sequences was used per sample. Rarefaction curves that reached the plateau phase indicated that the sequencing depth was sufficient for an analysis ( Figure 1A). Faith's Phylogenetic Diversity was statistically significant in the microbiome from RA mice through healthy controls (p < 0.05) for both feces and cecal content; however, the cecal mucus content did not differ between healthy controls and RA mice ( Figure 1B). The significant variations in the microbial diversity at different sites from the same mouse group were analyzed. The bacterial diversity in the cecal mucus resulted in significant differences (p < 0.05) in the microbiome of feces or cecal content in RA mice. Diversity did not differ between feces and the cecal content for mouse groups, either the control or RA model.

Core Microbiome
To test the presence of an identifiable common core bacterial community defined as the shared members among the microbiome and common genera, we used a Venn diagram. We identified 538 and 714 OTUs in feces and cecal mucus, respectively, from healthy mice, and 597 and 688 OTUs in feces and cecal mucus from RA mice (Figure 2A). In total, 398 OTUs were shared between healthy and RA groups, occupying 54% of all To investigate both community evenness and richness, the Shannon diversity index revealed a similar decreasing trend in phylogenetic differences in different groups, but only feces revealed a significant difference (p = 0.04, Kruskal-Wallis) between control and RA mice. The microbial community was markedly less diverse in RA than control mice. Therefore, the development of RA might be related to a decline in the α-diversity of the microbiome. The data from the feces samples agreed with data from the cecal content samples. Therefore, the cecal content samples were ignored and a further analysis focused on the data from the fecal and cecal mucus samples. Healthy controls and RA mice significantly differed in bacterial community in feces (PERMANOVA, p < 0.05) by using the Bray-Curtis distance-based microbiome structure analysis, separating along principal coordinate dimension 1 (PCoA1) and explaining approximately 24.4% of the total variations in data. They also differed in bacterial community in cecal mucus (PERMANOVA, p < 0.05) ( Figure 1C). PERMDISP indicated that dispersion did not contribute to significance (Table S1). The principal coordinate analysis of the weighted UniFrac distance showed that the RA treatment compared to healthy mice resulted in significant differences in βdiversity in both feces and mucus (PERMANOVA, p < 0.05), separating along principal coordinate dimension 1 (PCoA1) and explaining approximately 44.5% of the total variations in data. They also differed in bacterial community in cecal mucus (PERMANOVA, p < 0.05) ( Figure 1D). Similar results were obtained from the microbiome structure analysis when using the unweighted UniFrac and Jaccard distance ( Figure S1). Considering these results, we found evidence for RA-associated differences in both α-and β-diversity in bacterial community between the fecal and cecal mucus samples.

Core Microbiome
To test the presence of an identifiable common core bacterial community defined as the shared members among the microbiome and common genera, we used a Venn diagram. We identified 538 and 714 OTUs in feces and cecal mucus, respectively, from healthy mice, and 597 and 688 OTUs in feces and cecal mucus from RA mice (Figure 2A). In total, 398 OTUs were shared between healthy and RA groups, occupying 54% of all OTUs (737 OTUs) in feces, whereas 473 OTUs were shared between the healthy control and RA mice, representing 50.9% of all OTUs (929 OTUs) in cecal mucus. At the genus level, we identified 63 and 64 genera in feces and cecal mucus from healthy mice, and 54 and 69 genera in feces and cecal mucus from RA mice ( Figure 2B). In total, 45 and 55 genera were shared between healthy and RA groups, representing 62.5% and 70.5% of all genera in feces and cecal mucus, respectively.
All things considered, five and three common unique genera were found in both feces and cecal mucus samples, respectively, from healthy controls and RA mice. The five common unique genera in both the feces and cecal mucus from healthy controls belonged to Eubacterium_Brachy, Eubacterium_Siraeum, Eubacterium_Ventriosum, Monoglobus, and Treponema. Erysipelotrichaceae, Marvinbryantia, and Paraprevotella were not observed in healthy controls as compared with their abundance in RA mice ( Figure 2B).

Featured Microbial Taxa Using Wilcoxon Rank-Sum Tests
To further examine the variation in the relative abundance of different microbial taxa between the healthy control and RA groups, we used comparative analyses at all taxonomic levels for the mean relative abundance of two groups. The bacterial composition of

Featured Microbial Taxa Using Wilcoxon Rank-Sum Tests
To further examine the variation in the relative abundance of different microbial taxa between the healthy control and RA groups, we used comparative analyses at all taxonomic levels for the mean relative abundance of two groups. The bacterial composition of each group at both the phylum and genus levels was analyzed by Wilcoxon rank-sum tests to identify taxa differing in abundance at phylum levels ( Table 2). A total of 16 bacterial phyla was obtained in all samples; the phyla Firmicutes and Bacteroidetes were the most abundant, representing >80% of the gut microbiome. The present data showed a significantly increased abundance of Spirochaetes and Patescibacteria, and a decreased abundance of Proteobacteria and Desulfobacterota in fecal samples from RA mice ( Table 2). The ratio of Firmicutes/Bacteroidetes (F/B) is considered an important marker of the gut microbiome state. In this study, RA mice had a lower F/B ratio in fecal samples as compared to healthy controls ( Figure 5A). At the genus level, we identified 21 differentially abundant genera in fecal samples, with 14 genera enriched in healthy controls (Eubacterium_Ventriosum, Lachnospiraceae_UCG001, Monoglobus, Oscillospiraceae_NK4A214 group; Bacilli_RF39, Roseburia, Ruminococcus, Alistipes, Alloprevotella, Rikenella, Treponema, Brachyspira, Candidatus_Saccharimonas, and Candidatus_Arthromis) and 7 genera (Turicibacter, Tuzzerella, Muribaculum, Odoribacter, Parabacteroides, Rikenellaceae_RC9, and Parasutterella) enriched in RA ( Figure 5B and Table 3). Within the phylum Firmicutes, the genera Eubac-terium_Ventriosum, Roseburia, and Ruminococcus were more abundant (p ≤ 0.01) in healthy controls than in RA mice. Within the phylum Bacteroidetes, Proteobacteria, Patescibacteria, and Spirochaetes, the genera Rikenella, Candidatus_Saccharimonas, and Treponema were more abundant (p ≤ 0.01) in healthy controls than in RA mice.
For cecal mucus samples, three differentially abundant taxa enriched in healthy controls included Parabasalia, Bacteroidetes, and Spirochaetes, whereas Campilobacterota and Deferribacteres were enriched in RA mice ( Table 2). The statistical analysis revealed significant differences in the relative abundance of Bacteroidetes, with an average of a 27.9% and 33.8% abundance for RA samples and healthy controls, respectively. The F/B ratio was higher in cecal mucus samples in RA mice than healthy controls ( Figure 5A). At the genus level, 20 differentially abundant genera included 12 genera enriched in healthy controls (Negativibacillus, Eubacterium_Brachy, Eubacterium_Siraeum, Eubacterium_Ventriosum, Staphylococcus, Intestinimonas, Alloprevotella, Muribaculaceae, Rikenella, Mitochondria, Pseudomonas, and Treponema) and 8 genera enriched in RA mice (Lachnospiraceae_A2, Anaeroplasma, Blautia, Lachnoclostridium, Marvinbryantia, Butyricicoccaceae_UCG009, Helicobacter, and Mucispirillum). Within the phylum Firmicutes, the bacterial genera Eubacterium_Siraeum, Eubacterium_Ventriosum, and Negativibacillus were more abundant (p ≤ 0.01) in healthy controls than in RA mice. Within the phylum Proteobacteria, the genera Mitochondria and Pseudomonas were more abundant (p ≤ 0.01) in healthy controls than RA mice. However, the Treponema and Muribaculaceae genera were more abundant (p ≤ 0.01) in healthy controls than RA mice ( Figure 5C and Table 3).    Figure 5B and Table 3). Within the phylum Firmicutes, the genera Eubac-terium_Ventriosum, Roseburia, and Ruminococcus were more abundant (p ≤ 0.01) in healthy controls than in RA mice. Within the phylum Bacteroidetes, Proteobacteria, Patescibacteria, and Spirochaetes, the genera Rikenella, Candidatus_Saccharimonas, and Treponema were more abundant (p ≤ 0.01) in healthy controls than in RA mice.  Figure 5B and Table 3). Within the phyl terium_Ventriosum, Roseburia, and Ruminococcus were mo controls than in RA mice. Within the phylum Bacteroide and Spirochaetes, the genera Rikenella, Candidatus_Sacchari abundant (p ≤ 0.01) in healthy controls than in RA mice.  Figure 5B and Table 3). Within the phylum Firmicutes, the genera Eubac-terium_Ventriosum, Roseburia, and Ruminococcus were more abundant (p ≤ 0.01) in healthy controls than in RA mice. Within the phylum Bacteroidetes, Proteobacteria, Patescibacteria, and Spirochaetes, the genera Rikenella, Candidatus_Saccharimonas, and Treponema were more abundant (p ≤ 0.01) in healthy controls than in RA mice.  Figure 5B and Table 3). Within the phyl terium_Ventriosum, Roseburia, and Ruminococcus were mo controls than in RA mice. Within the phylum Bacteroide and Spirochaetes, the genera Rikenella, Candidatus_Sacchari abundant (p ≤ 0.01) in healthy controls than in RA mice.  Figure 5B and Table 3). Within the phylum Firmicutes, the genera Eubac-terium_Ventriosum, Roseburia, and Ruminococcus were more abundant (p ≤ 0.01) in healthy controls than in RA mice. Within the phylum Bacteroidetes, Proteobacteria, Patescibacteria, and Spirochaetes, the genera Rikenella, Candidatus_Saccharimonas, and Treponema were more abundant (p ≤ 0.01) in healthy controls than in RA mice.  Figure 5B and Table 3). Within the phylum Firmicutes, the genera Eubac-terium_Ventriosum, Roseburia, and Ruminococcus were more abundant (p ≤ 0.01) in healthy controls than in RA mice. Within the phylum Bacteroidetes, Proteobacteria, Patescibacteria, and Spirochaetes, the genera Rikenella, Candidatus_Saccharimonas, and Treponema were more abundant (p ≤ 0.01) in healthy controls than in RA mice. The ratio of Firmicutes/Bacteroidetes (F/B) is considered an important marker of the gut microbiome state. In this study, RA mice had a lower F/B ratio in fecal samples as compared to healthy controls ( Figure 5A). At the genus level, we identified 21 differentially abundant genera in fecal samples, with 14 genera enriched in healthy controls (Eubacte-rium_Ventriosum, Lachnospiraceae_UCG001, Monoglobus, Oscillospiraceae_NK4A214 group; Bacilli_RF39, Roseburia, Ruminococcus, Alistipes, Alloprevotella, Rikenella, Treponema, Brachyspira, Candidatus_Saccharimonas, and Candidatus_Arthromis) and 7 genera (Turicibacter, Tuzzerella, Muribaculum, Odoribacter, Parabacteroides, Rikenellaceae_RC9, and Parasutterella) enriched in RA ( Figure 5B and Table 3). Within the phylum Firmicutes, the genera Eubac-terium_Ventriosum, Roseburia, and Ruminococcus were more abundant (p ≤ 0.01) in healthy controls than in RA mice. Within the phylum Bacteroidetes, Proteobacteria, Patescibacteria, and Spirochaetes, the genera Rikenella, Candidatus_Saccharimonas, and Treponema were more abundant (p ≤ 0.01) in healthy controls than in RA mice.  Figure 5B and Table 3). Within the phylum Firmicutes, the genera Eubac-terium_Ventriosum, Roseburia, and Ruminococcus were more abundant (p ≤ 0.01) in healthy controls than in RA mice. Within the phylum Bacteroidetes, Proteobacteria, Patescibacteria, and Spirochaetes, the genera Rikenella, Candidatus_Saccharimonas, and Treponema were more abundant (p ≤ 0.01) in healthy controls than in RA mice. FDR, false discovery rate; AUC, area under the receiver operating characteristic curve; CI, confidence interval; CT, healthy control; RA, rheumatoid arthritis.
The ratio of Firmicutes/Bacteroidetes (F/B) is considered an important marker of the gut microbiome state. In this study, RA mice had a lower F/B ratio in fecal samples as compared to healthy controls ( Figure 5A). At the genus level, we identified 21 differentially abundant genera in fecal samples, with 14 genera enriched in healthy controls (Eubacte-rium_Ventriosum, Lachnospiraceae_UCG001, Monoglobus, Oscillospiraceae_NK4A214 group; Bacilli_RF39, Roseburia, Ruminococcus, Alistipes, Alloprevotella, Rikenella, Treponema, Brachyspira, Candidatus_Saccharimonas, and Candidatus_Arthromis) and 7 genera (Turicibacter, Tuzzerella, Muribaculum, Odoribacter, Parabacteroides, Rikenellaceae_RC9, and Parasutterella) enriched in RA ( Figure 5B and Table 3). Within the phylum Firmicutes, the genera Eubac-terium_Ventriosum, Roseburia, and Ruminococcus were more abundant (p ≤ 0.01) in healthy controls than in RA mice. Within the phylum Bacteroidetes, Proteobacteria, Patescibacteria, and Spirochaetes, the genera Rikenella, Candidatus_Saccharimonas, and Treponema were more abundant (p ≤ 0.01) in healthy controls than in RA mice.  The ratio of Firmicutes/Bacteroidetes (F/B) is considered an important marker of the gut microbiome state. In this study, RA mice had a lower F/B ratio in fecal samples as compared to healthy controls ( Figure 5A) Figure 5B and Table 3). Within the phylum Firmicutes, the genera Eubac-terium_Ventriosum, Roseburia, and Ruminococcus were more abundant (p ≤ 0.01) in healthy controls than in RA mice. Within the phylum Bacteroidetes, Proteobacteria, Patescibacteria, and Spirochaetes, the genera Rikenella, Candidatus_Saccharimonas, and Treponema were more abundant (p ≤ 0.01) in healthy controls than in RA mice.   Figure 5B and Table 3). Within the phyl terium_Ventriosum, Roseburia, and Ruminococcus were mo controls than in RA mice. Within the phylum Bacteroide and Spirochaetes, the genera Rikenella, Candidatus_Sacchari abundant (p ≤ 0.01) in healthy controls than in RA mice. nificantly increased abundance of Spirochaetes and Patescibacteria, and a decreased abundance of Proteobacteria and Desulfobacterota in fecal samples from RA mice ( Table 2).  Figure 5B and Table 3). Within the phylum Firmicutes, the genera Eubac-terium_Ventriosum, Roseburia, and Ruminococcus were more abundant (p ≤ 0.01) in healthy controls than in RA mice. Within the phylum Bacteroidetes, Proteobacteria, Patescibacteria, and Spirochaetes, the genera Rikenella, Candidatus_Saccharimonas, and Treponema were more abundant (p ≤ 0.01) in healthy controls than in RA mice.   Figure 5B and Table 3). Within the phyl terium_Ventriosum, Roseburia, and Ruminococcus were mo controls than in RA mice. Within the phylum Bacteroide and Spirochaetes, the genera Rikenella, Candidatus_Sacchari abundant (p ≤ 0.01) in healthy controls than in RA mice.

Phylum Genus
Week

Phylum Genus
Week

Discussion
Alterations in the fecal microbiome between RA and healthy individuals have been reported on since the beginning of this century [3-5,10]. However, details of the microbiome in the colon content and mucus remained unclear for RA patients. In this study, we used samples from feces, cecal content, and cecal mucus in an RA mouse model and healthy controls to investigate the microbial composition. We generated a total of 3,461,121 sequences representing 1110 unique OTUs with a 99% Good's coverage for all samples. Rarefaction curves showed that the sequencing depth was sufficient for further 0.05 Marvinbryantia Figure 6. Composition of the cecal mucosa microbiome in healthy control, RA, TDAG8-deficient, and CCL-2d-treated mice at week 12. (A) Significantly abundant genera in RA and TDAG8-deficient B6 mice. (B) Significantly abundant genera in healthy control, RA, and CCL-2d treated ICR mice. Data are median (horizontal line), interquartile range (box edges) and range (whiskers). * p < 0.05; ** p < 0.01; *** p < 0.001, Wilcoxon rank-sum test. Table 4. Composition of the cecal mucus microbiome at weeks 1-12 and 12.

Phylum Genus
Week

Discussion
Alterations in the fecal microbiome between RA and healthy individuals have been reported on since the beginning of this century [3-5,10]. However, details of the microbiome in the colon content and mucus remained unclear for RA patients. In this study, we used samples from feces, cecal content, and cecal mucus in an RA mouse model and healthy controls to investigate the microbial composition. We generated a total of 3,461,121 sequences representing 1110 unique OTUs with a 99% Good's coverage for all samples. Rarefaction curves showed that the sequencing depth was sufficient for further study because the samples reached the plateau phase ( Figure 1A). From the rarefaction 0.05

Phylum Genus
Week

Discussion
Alterations in the fecal microbiome between RA and healthy individuals have been reported on since the beginning of this century [3][4][5]10]. However, details of the microbiome in the colon content and mucus remained unclear for RA patients. In this study, we used samples from feces, cecal content, and cecal mucus in an RA mouse model and healthy controls to investigate the microbial composition. We generated a total of 3,461,121 sequences representing 1110 unique OTUs with a 99% Good's coverage for all samples. Rarefaction curves showed that the sequencing depth was sufficient for further study because the samples reached the plateau phase ( Figure 1A). From the rarefaction results, a minimum of 31,794 sequences per sample was used for standardizing the microbial estimations. According to the αand β-diversity indices for the microbiome, the fecal and cecal content did not significantly differ in both healthy controls and RA mice. However, the bacterial composition in cecal mucus was significantly different from feces in all conditions. In previous studies, the mucus microbiome was also found different from that in feces from mice, humans, and Rhesus macaque [24][25][26]. The microbial composition in the intestine is partially correlated with that in feces, but the fecal microbiome does not represent the complete picture in the intestine [24][25][26]. In intestinal dysbiosis particularly, the represented mucus microbiome plays an important role because of a close interaction with epithelial cells and the mucus immune system [24][25][26]. In most studies, diversity indices are reduced in terms of phylogenetic diversity, species richness, and evenness in RA mice as compared to healthy individuals [9,10,27,28]. Our results agreed with the published results. The apparent decrease in microbial diversity is an important marker that indicates the association between the etiology of RA and the microbiome.
The investigation of the presence of a common core bacterial community revealed 43% of the genera in all samples. In addition, on comparing RA and healthy control samples in the different intestinal sites, the core bacterial microbiome was stable, which was more than 62% of the genera. The altered gut microbiome acts as an adjuvant criterion for clinical diagnosis to identify patients with autoimmune diseases [29][30][31][32]. In this study, at the phylum level in fecal samples, RA mice showed a decrease in Spirochaetes and Palescibacteria content and an increase in Proteobacteria and Desulfobacterota content as compared to healthy controls. However, at the phylum level in cecal mucus samples, RA mice showed a decrease in Parabasalia, Bacteroidetes, and Spirochaetes content and an increase in Deferribacteres and Campilobacterota content as compared to healthy controls ( Table 2).
The F/B ratio can be used as an important indicator of the gut microbiome state and host health [28,33]. Bacteroidetes found in the gut mainly functions in polysaccharide metabolism and calorie absorption, whereas Firmicutes is important for the production of short-chain fatty acids [34]. Scher et al. found Bacteroidetes absent in patients with new-onset RA as compared to healthy controls [7]. The analysis of the fecal microbiome composition revealed a higher F/B ratio in RA than osteoarthritis patients [28]. The collagen-induced arthritis mouse model used to study the immune-priming phase of arthritis revealed a decrease in Bacteroidetes and an increase in Firmicutes content [33]. In agreement with previous studies [7,28,33], we found a higher F/B ratio in cecal mucus from RA than healthy control mice ( Figure 5A). However, fecal samples did not show a similar trend. Thus, the F/B ratio could be a good indicator for mouse mucosal samples but may not apply to mouse fecal samples.
In both feces and cecal mucus, Eubacterium_Ventriosum, Alloprevotella, Rikenella, and Treponema were significantly less abundant in the RA mouse than healthy control microbiome ( Figure 5B,C). Our results agreed with results from some previous studies [10,35]. Sun et al. investigated samples from 66 Chinese patients with RA and 60 healthy controls by using the bacterial 16S rDNA gene; Alloprevotella and Rikenella were less abundant in the RA than control group. Alloprevotella and Treponema were reported to produce significant amounts of short-chain fatty acids, and their abundance is negatively correlated with metabolic syndrome [35,36]. The Alloprevotella content was found to be positively correlated with inflammation biomarkers and the rheumatoid factor [10]. Severijnen et al.
investigated arthritis-inducing properties of Eubacterium species and revealed a diversity in such properties among different species of the anaerobic genus Eubacterium in inducing joint inflammation [37]. However, the exact effect of Treponema and Eubacterium_Ventriosum on RA is difficult to determine, because the isolation and in vitro cultivation of these strains are challenging. Given that we observed a reduced abundance of Eubacterium_Ventriosum, Alloprevotella, Rikenella, and Treponema in both RA fecal and mucosal samples, these four genera could serve as microbial markers for RA progression.
In our previous study, TDAG8 gene deficiency relieved RA disease severity and chronic pain [15]. A salicylanilide derivative compound, CCL-2d, which inhibits TDAG8 function and expression, also provided similar results as TDAG8 deficiency in mice [15]. To investigate whether the deficiency of the TDAG8 gene affects the composition of the microbiome in the molecular mechanism, the inhibition of TDAG8 expression and function by gene deletion or an inhibitor was performed. The results revealed that mice with TDAG8 gene deficiency showed a restoration of the gut microbial ecosystem by significantly reducing Eubacterium_Xylanophilum, Clostridia, Ruminococcus, Paraprevotella, and Rikenellaceae (Table 4). A reduction in Clostridia was observed in RA patients using Etanercept or Sulfasalazine, drugs used to treat RA [38,39]. In this study, mice receiving TDAG8 deficiency had a decreased number of Clostridia. It suggested that the TDAG8 treatment of RA could be responsible for the reduction in bacterial numbers and could be potentially beneficial to RA. In previous findings, the Ruminococcus content was found to be correlated with intestinal inflammation and a variety of other inflammatory diseases. The inflammatory glucorhamnan polysaccharide was mainly found in Ruminococcus [40,41]. Additionally, RA patients showed an increased content of the genus Rikenellaceae [42]. The results indicated that Clostridia, Ruminococcus, and Rikenellaceae could be proinflammation-related microorganisms promoting RA disease progression. TDAG8 deficiency was demonstrated to reduce the number of satellite glial cells and proinflammatory macrophages that could be the cause of the change in the microbiome. Thus, TDAG8-deficient RA mice showing a reduced disease severity and RA pain could be due to the modulation of gut microorganisms affecting the pathogenesis of RA.
RA patients have chronic inflammation and persistent pain hypersensitivity to mechanical and thermal stimuli [43]. However, it is not easy to establish an animal model which reproduces all RA clinical features. In current RA models, some only have short-term inflammation, some only show unilateral hypersensitivity, some models have persistent mechanical hypersensitivity but short-term thermal hypersensitivity, and some models are not suitable in mice. Our model was adopted and modified from the model established by Gauldie et al. in 2004. Our RA mice displayed long-term inflammation and long-term bilateral pain hypersensitivity to mechanical and thermal stimuli [15,44]. RA clinical features were also found in our RA mice, such as a high concentration of [H + ] in synovial fluid, a continuous serum IL-6 production, and an increased synovial macrophage CD68+ number that marked the disease in the chronic inflammatory state. In addition, we successfully established an RA model in both ICR and B6 mice [15,19,20]. Thus, our RA model could reproduce some of the possible mechanisms at play in RA, rather than OA or other arthritis. Our RA model started from the initiation of autoimmunity and lacked the stage of no symptoms or signs of autoimmunity. It had some limitations in the studies of some risk factors.

Agents
Complete Freund's adjuvant was from Sigma-Aldrich (Darmstadt, Germany). A salicylanilide derivative compound, CCL-2d (3-(4-Chloro-2-fluorophenyl)-7-methoxy-2Hbenzo[e] [1,3]-oxazine-2,4 (3H)-dione), was synthesized as described [45]. All reagents or compounds were first solved in dimethylsulfoxide, then diluted in saline before injection in animal experiments. cutoff value [52]. A summary of all taxonomic information was generated by using the q2feature-classifier classify-sklearn naive Bayes taxonomy classifier against the Silva dataset v138 [53,54]. To standardize results, the equivalent number of sequence reads (based on the lowest number of sequences obtained from a single sample) per sample chosen by rarefaction was used for all subsequent comparisons. To determine the core microbiome, genus abundance > 0.1% was used for analysis. Venn diagrams were constructed by using Venny 2.1. Both matrices for the complete and resampled datasets were calculated and compared by applying the Mantel tests implemented in the R v3.6.3 package Vegan. For beta-diversity analysis, we determined the microbial composition diversity between individuals by using weighted UniFrac, unweighted UniFrac, Jaccard, and Bray-Curtis distance in the q2-diversity plugin [55,56]. The linear Principal Component Analysis (PCA) model was also created by using the q2-diversity plugin. Significant differences in betadiversity were determined with QIIME by PERMANOVA, and PERMDISP was used to check for significant differences in dispersion. For featured taxa selection, we used LEfSe and Calypso [57] to calculate the linear discriminant analysis effect size (LEfSe) and random forest prediction. An LDA score of >3.0 and Kruskal-Wallis α-value of 0.05 were set as thresholds; p < 0.05 was considered statistically significant.

Conclusions
In this study, we compared the microbiome composition in feces and mucus samples of complete Freund's adjuvant-induced arthritis mouse models. Four core bacterial genera, Eubacterium_Ventriosum, Alloprevotella, Rikenella, and Treponema, could be biomarkers of an altered RA microbiome in both fecal and mucosal samples. TDAG8 deficiency decreased the abundance of proinflammation-related Eubacterium_Xylanophilum, Clostridia, Ruminococcus, Paraprevotella, and Rikenellaceae, which reduced local mucosal inflammation to relative RA disease severity and pain. The pharmacological block of TDAG8 function by a salicylanilide derivative partly restored the RA microbiome to a healthy microbiome composition. Understanding the bacterial interaction with the host mucus in the gut and TDAG8 modulation in specific microbiota could facilitate the development of novel microbiota-based therapy.