This study was authorized by the ethics committee of West China Hospital of Stomatology, Sichuan University (WCHSIRB-D-2017-035). Informed consent was provided by all donors. All saliva samples were collected at the Periodontology Department of the West China Hospital of Stomatology, according to the methods described earlier (Miller et al., 2021). The patients were diagnosed with periodontitis according to the clinical classification and definition of periodontitis (Tonetti et al., 2018). Donors with systemic diseases, including diabetes, a history of smoking, a recent history (3 months) of antibiotic use or the periodontitis treatment were excluded. Eight adult volunteers with periodontitis (four women, four men, age 44–72 years) were selected; and 12 healthy donors of matched gender and age with periodontitis patients were included as the control group. All saliva samples were mixed into samples of equal volume (5 ml) as the representatives of each group, verified by 16S rRNA gene sequencing.

This study was authorized by the ethics committee of West China Hospital of Stomatology, Sichuan University (WCHSIRB-D-2017-069). A total of 12 germ-free BALB/c mice (female mice, 5 weeks old) provided by the Department of Laboratory Animal Science, College of Basic Medical Sciences, Army Medical University (Chongqing, China) were used. Streptozotocin (STZ, Sigma-Aldrich United States) at a dose of 50 mg/kg was injected intraperitoneally into germ-free mice (n = 8) for the five consecutive days as previously described (Chen et al., 2015). One week after the first injection, the fast plasma glucose level was detected to determine the success of model of germ-free mice with STZ-induced T1D before colonization of saliva samples. To ensure the sterile conditions without contamination, the feces of all mice were collected and cultured in the BHI media per week. After germ-free mice with STZ-induced T1D models were successfully established, all the mice were randomly divided into two groups, including the mice inoculated with mixed saliva samples from periodontitis patients (SP, n = 4), of which, one mouse died was excluded, and the mice inoculated with mixed saliva samples from healthy subjects (SH, n = 4). They were inoculated with a 200 μl mixture of fresh saliva by swabs without anesthesia, and the teeth of mice were swabbed for 1 min per mouse (Li et al., 2019). All operations were performed under the sterile conditions throughout the experiment. Untreated germ-free mice (GF, n = 4) were considered as the blank control. Each group of mice was housed in separate gnotobiotic isolators for 2 weeks, before being euthanized.

All suspensions of systemic organs, intestinal contents, and blood of equal quality or volume were collected and cultured on the plates of Brain Heart Infusion medium supplemented with hemin and vitamin K, under strictly anaerobic conditions (80% N₂, 10% H₂, 10% CO₂) at 37°C for 3–7 days. Colony-forming units (CFUs) were counted for each plate. Bacterial DNAs were extracted from the representative colonies (the numbers of CFUs were above 10) according to the colony morphology and Gram staining, or all the colonies (the numbers of CFUs were below 10). Universal primers 27F and 1492R were used for PCR amplification and sent to Biochemical Bioengineering (Shanghai, China) for Sanger sequencing. Sequences were identified by a blast alignment in the NCBI database > 99.5% similarity (Sayers et al., 2010).

Bacterial DNAs of the small intestinal contents were extracted using the QIAamp Fast DNA Stool Mini Kit (Cat No. 51604, TIANGEN Biotech, China) and the TIANamp Swab DNA Kit (Cat No. DP322, TIANGEN Biotech, China) was applied for saliva. F338 and R806 primers were used to amplify the bacterial 16S rRNA genes, which focused on regions of V3-V4 followed by Illumina HiSeq technology sequencing. All the pre-processing of sequences was conducted by an Galaxy-based pipeline in Denglab¹ (Feng et al., 2017). The raw sequencing was quality-filtered by Trimmomatic (V0.33) and merged using FLASH (V1.2.11). Operational taxonomic units (OTUs) table was generated by UPARSE with a clustering threshold of 0.97. In total, 692,105 high-quality sequences were obtained. Silva database was used to annotate the taxonomic information (Quast et al., 2013). Rarefaction curves were analyzed by calculating the species richness of bacterial OTUs. The rarefied OTU table was applied for the most downstream analyses based on the sample with the least sequences (77,627 reads/per sample) and conducted on the website (see text footnote 1). Alpha-diversity indices (Simpson, Observed_richness, and Pielou_evenness) were estimated. Principal component analysis (PCA) was applied to determine the dissimilarity between bacterial communities of the two groups. Three different non-parametric analyses, namely, the multi-response permutation procedure (MRPP), permutational multivariate analysis of variance (PERMANOVA), and analysis of similarities (ANOSIM) were performed. The relative abundances of bacterial taxa at phylum, class, and genus levels were calculated.

The proportions of the following 15 immune cells were analyzed as follows: monocytes (Monos), MFs, mononuclear phagocytes (MNPs), CD11B + dendritic cells (CD11B + DCs), CD11B- dendritic cells (CD11B-DCs), pDCs, group 3 innate lymphoid cells (ILC3s), B cells, γδT cells (gdT), αβT cells (abT), CD4-CD8- T cells (DN), CD8 + T cells, CD4 + T cells, Th17, and Tregs. And the proportions of the following six cytokines were detected: IL17, IFNγ, IL22, and IL10 produced by CD4 + T cells, as well as IL17 and IL22 produced by ILCs. Single-cell suspensions of the small intestine and spleen were prepared as previously described (Benakis et al., 2016). Anti-mouse CD16/32 antibody was used for blocking the Fc domain before staining the surface or intracellular markers. All antibodies were divided into three groups and listed in Supplementary Table 1. To detect the innate immune cells, the first group included antibodies against CD45, CD19, Ly6c, PDCA-1, CD11c, CD11b, F4/80, and CD103. To detect the adaptive immune cells, the second group included antibodies against CD45, CD19, TCRß, TCRgd, CD4, CD8, Foxp3, and Rorγ. To detect the cytokines, the third group included antibodies against CD45, CD4, TCRß, TCRgd, IL17a, IFNγ, IL22, and IL10. Procedures of cell markers staining were following the instructions. BD Celesta was used to detect changes in cells. To ensure equal gating criteria and scoring, the raw data were independently analyzed by two individuals using the Kaluza software (version 2.1.1, Beckman Coulter, United States). Heatmap was drawn through the following steps: First, the fold change cell values in comparison of group GF in each cell type were log2 transformed; second, the values of each row were normalized to [−1, 1]; and finally, a package in R was used to draw the heatmap.

Variance inflation factor (VIF) was applied to determine which cells were highly autocorrelated and cells with the VIF value above 10 were excluded to further analysis, and redundancy analysis (RDA) was adopted to analyze the correlation between geographical distribution of gut bacterial communities and immune cells in small intestine and spleen, using a package “vegan” in R version 3.6.2 (Sadyś et al., 2015). Envfit analysis in package “vegan” was applied to identify the significant immune cells affected by the microbial composition distribution.

Data were calculated as means ± SEM. All statistical analyses were performed using GraphPad Prism 6 software (GraphPad Prism6 Software, Inc., United States) or SPSS (version 24.0, IBM Corporation, United States). Wilcoxon rank-sum test was applied in taxonomy analysis at phylum, class, and genus levels. Two-tailed unpaired Student’s t-test was applied in difference alpha-diversity indices and the immune cells. P < 0.05 was considered significant.

Periodontitis microbiota primarily colonized the small intestines and colons, followed by the lung, blood, and spleen of germ-free mice with STZ-induced T1D. Neither pancreas nor liver could be colonized by periodontitis microbiota (Figure 1A and Supplementary Figure 1H). Sanger sequences were blasted in NCBI website to identify the closely related species, and those with a more than 99.5% similarity are summarized in Table 1. Compared to SH, the diversity of the species was decreased in SP group, and the main bacteria identified were Klebsiella pneumoniae in the extra-oral sites.

Illumina HiSeq technology was applied to determine the bacterial community from seven samples of the small intestinal contents. In total, 692,105 high-quality sequences were obtained with an average length of 470 bp and an average of 98,872 sequences were generated per sample. The Silva database was used to categorize these sequences into 38 classified phyla, 438 genera, and 2,342 OTUs. The rarefaction curves tended to be flat in two groups (Figure 1B). The dissimilarity test showed significant differences by three parameters (MRPP = 0.65, p = 0.03; PERMANOVA = 1.11, p = 0.05; and ANOSIM = 0.87, p = 0.03) between SP and SH. However, the data in alpha diversity indices of the Inv_Simpson (p > 0.05), Pielou_evenness (p > 0.05) and Observed_richness (p > 0.05) indicated a similar bacterial community between two groups (Figure 1C). PCA was conducted to visualize the different diversities of microbiota in two groups (Figure 1D). Proteobacteria (90%) were the dominant phylum in the small intestine of two groups. Compared to SH, the relative abundance of Bacteroidota was increased in the top 10 of phylum level (p < 0.05), and the ratio of Firmicutes to Bacteroidetes was nearly five times lower in SP (Figure 1E). At the relative abundance of top 10 class level, four types were significantly different between SH and SP, including Clostridia, Bacteroidia, Bacilli, and Alphaproteobacteria (p < 0.05) (Figure 1F). Klebsiella were the dominant genera in SH and SP, accounting for nearly 55% and 72%, respectively. Among genera of the relative abundance above 1%, Raoultella (1.06%, p < 0.05) were decreased for seven folds and Lacticaseibacillus (1.10%, p < 0.05) were increased in SP (Figure 1G). As the minor components, several genera with relative abundances less than 1% were significantly different in the two groups, including Bacillus, Agathobacter, and Bacteroides in terms of the relative abundance of top 20 genus level (Figure 1H).

The success of animal models was verified in this study. The fast plasma glucose concentrations of the STZ-treated germ-free mice were 10.55 ± 2.92 mM, while the control group remained normoglycemic, at 6.85 ± 0.42 mM with a significant difference (Supplementary Figure 1A). Namely, the model of germ-free with STZ-induced T1D was constructed successfully. Before bacterial colonization, the diversity of microbial community of the mixed saliva samples was evaluated. 16S rRNA gene sequencing was conducted for the mixed saliva samples from eight periodontitis patients and 12 healthy volunteers (Supplementary Figure 1B). The bacterial distribution was dramatically different in the genus level between periodontitis patients and healthy volunteers. Specifically, the former group was dominated by Porphyromonas, Fusobacterium, and Treponema, which were the red or yellow complex in the periodontitis. At the end point of experiment, the ceca in SH and SP were much smaller than those in group GF (Supplementary Figure 1C), indicating that the oral bacteria successfully translocated into the gut. The pancreatic tissues were collected and then stained with hematoxylin-eosin staining. Compared to the GF group, the pancreatic islets were emptied with vacuolar degeneration, necrosis, and disappearance of pancreatic β cells (indicated by a black arrow in Supplementary Figure 1D). Vascular dilatation and congestion of the intra and para-pancreatic islets were observed (indicated by a blue arrow in Supplementary Figure 1D). The effects of periodontitis microbiota on the fast plasma glucose and fast plasma insulin were also investigated (Supplementary Figures 1E,F). Compared to SH, there were no significant changes in either of the two parameters in SP. Moreover, alveolar bone loss was not induced by transferring periodontitis bacteria to germ-free mice with STZ-induced T1D (Supplementary Figure 1G).

A total of 15 types of immune cells were chosen, including seven innate and eight adaptive immune cells, to investigate how periodontitis microbiota affected local immunity in the small intestine (Supplementary Table 2 and Supplementary Figure 2A). The gating strategy for the two staining panels, namely, innate immune cells (Supplementary Figure 2B) and adaptive immune cells (Supplementary Figure 2C), was demonstrated by representative flow cytometry plots. A total of 195 independent immunophenotypes triggered by oral saliva microbiota in the small intestine were generated. Normalized fold changes relative to the group GF were described in a heatmap (Figure 2A).

As shown in the heatmap, SH and SP shared a similar immune landscape in the small intestine. Compared to SH, there were fewer responsive immune cells of the small intestine in SP, with only four types of cells being different. The proportions of ILC3s (p = 0.078) and CD4⁺ T cells (p = 0.074) were rising in SP, concomitant with a decline of MFs (p = 0.078) and B cells (p = 0.073). The most notable changes were considered as MFs (reported as% of CD45⁺CD19^– cells) and B cells (reported as% of CD45⁺ cells), which were strongly affected by periodontitis microbiota. The frequency of MFs was decreased from 3.9 ± 1.1% to 1.1 ± 0.1%, with nearly three-fold changes. A more than two-fold decrease was shown in the proportion of B cells in SP, from 22.9 ± 4.1% to 10.4 ± 3.2% (Figures 2B,C). However, most of the remaining immune cells did not show significant changes (Supplementary Figures 3A,B).

Cytokines are an important factor for the development and activation of immune cells. Six cytokines were detected to uncover how the function of immune cells changed in the small intestine. CD4 + T cells producing IFNγ, IL17, IL10, and IL22 as well as ILCs producing IL17 and IL22 were included. The gating strategy for cytokines is shown in Supplementary Figure 2B and Supplementary Table 2. Compared to SH, CD4 + T cells were significantly activated by periodontitis microbiota, with the accumulation of IFNγ⁺ CD4 (p < 0.05), IL17⁺ CD4 (p < 0.05), and IL22⁺ CD4 (p < 0.05) T cells in the small intestine (Figures 2D–F). Moreover, the frequency of IL22 (p < 0.05) produced by ILCs was rising in SP. But, the frequencies of IL10⁺ CD4 (p > 0.05) and IL17⁺ ILC (p > 0.05) were not changed significantly in SP (Supplementary Figure 3C).

The same types of immune cells were detected to investigate how periodontitis microbiota affected the systemic immune response in the spleen (Figure 3A). Compared to SH, a total of four innate immune cells and five adaptive ones were significantly altered in SP. In terms of the innate immune response in the spleen, most of the cells appeared to be more responsive than those in the small intestine, demonstrated by modest changes in the proportion of immune cells. The proportions of MFs (p < 0.05), ILC3s (p < 0.05), Monos (p < 0.05), and pDCs (p < 0.05) were increased in SP, of which proportions of MFs and pDCs were strongly changed by more than threefold. The frequency of Monos increased from 0.4 ± 0.1% to 1.8 ± 0.5% combined with a rising proportion of pDCs from 0.6 ± 0.1% to 2.0 ± 0.2% (Figures 3B,C). Regarding adaptive immune response, a marked increase in the frequencies of B (p < 0.05), Th17 (p < 0.05), and DN cells (p < 0.05) was induced by periodontitis microbiota, as well as a decrease in those of αβT (p < 0.05) and γδT (p < 0.05). It occurred dramatically in the proportion of Th17 (reported as% CD4⁺CD8a^–TCRb⁺CD19^–CD45⁺ T cells) and DN cells (reported as% of TCRb⁺CD45⁺CD19^– cells), with a nearly three-fold increase, ranging from 13.2 ± 2.3% to 38.1 ± 8.9% and 1.5 ± 0.2% to 5.5 ± 1.7%, respectively (Figures 3B,C). However, the other cells were not significantly responsive to colonization of periodontitis microbiota (p > 0.05) (Supplementary Figures 4A,B).

To determine whether the immunological changes in local and systemic organs were affected by the gut microbiota community that were colonized from periodontitis microbiota, RDA was performed. A total of 10 variables of the small intestine and seven cell types of the spleen were selected for further analysis in the RDA, of which the values of VIF were below 10 (Tables 2A,B). In both analyses of the small intestine and the spleen, nearly 55% and 30% of the variance at the first and second RDA axes were explained, respectively (Figure 4).

The ANOVA test helped to determine statistically significant axes. The gut microbiota had no significant effect on the innate and adaptive immune cells in the small intestine. However, the pDCs (envfit analysis, r² = 0.8407, p = 0.013) and B cells (envfit analysis, r² = 0.0.6937, p = 0.071) were initially linked to community structure of gut microbiota. Contrary to SH, a negative correlation between B cells and the gut microbiota occurred in SP (p > 0.05). In addition, pDCs were positively related to the periodontitis microbiota that colonized the small intestine (p < 0.05) (Figures 4A,B).

Systemic immune cells in spleen appeared to be more significant variables with respect to the gut microbiota. pDCs (F = 9.47, p = 0.001), CD11B + DCs (F = 5.23, p = 0.011), and MNPs (F = 3.79, p = 0.021) were the most important innate immune cells associated with the gut microbiota. Furthermore, pDCs (envfit analysis, r² = 0.0.8515, p = 0.045) were positively correlated with the periodontitis microbiota that colonized the small intestine, which showed a negative relationship in SH (Figure 4C). In relation to the adaptive immune cells, gdT (F = 3.79, p = 0.021) showed a major significant factor response to the diversity of gut bacterial structure. In addition, B cells (envfit analysis, r² = 0.7248, p = 0.094) and CD4 + T cells (envfit analysis, r² = 0.7236, p = 0.062) were initially related to community structure of gut microbiota. Contrary to SH, there was a positive relationship between the gut microbiota and B cells in SP (p > 0.05). And CD4 + T cells were negatively correlated with the gut microbiota in SP (p > 0.05) (Figure 4D).

Overall, in this study, we found that periodontitis microbiota could migrate to the distal organs, but not colonize the liver and the pancreas in the state of T1D. Klebsiella were the dominant genera in these systemic organs. Periodontitis microbiota also contributed to Lacticaseibacillus, Bacillus, Agathobacter, and Bacteroides, colonizing the small intestine, in contrary to Raoultella (Figure 5A). Innate and adaptive immune responses may be dysregulated for two or three times by periodontitis microbiota in the local or systemic organs. MFs, DN, IL22 + ILCs, Monos, IL22⁺CD4 T cells, IFNγ⁺ CD4 T cells, and pDCs might be more responsive to periodontal microbiota. Ectopic colonization of periodontitis microbiota mainly drove Th1, Th17, and Th22 cells induction and inflammation in the small intestine. pDCs might be the key immune cells which were involved in the relationship between periodontitis microbiota and T1D onset (Figure 5B).