Periodontitis salivary microbiota exacerbates colitis by CXCL3 derived from gut microbiota-induced macrophages

Abstract

Background

Increasing research has focused on the role of the oral-gut axis in the development of colitis. Saliva contains a large number of oral bacteria that influence gut microbiota and colitis, but the underlying mechanisms remain unclear. In this study, we investigated the role and mechanisms of gut microbiota in salivary microbiota-affected colitis.

Results

We confirmed that periodontitis salivary microbiota (PSM) exacerbated colitis compared to healthy salivary microbiota (HSM). Antibiotics could reverse the effect of PSM in exacerbating colitis, suggesting that the altered gut microbiota was pathogenic. PSM resulted in the enrichment of pathogens, such as Escherichia coli, and lipopolysaccharide in the gut microbiota, and this gut microbiota was shown to be detrimental to colitis by C-X-C motif chemokine ligand 3(CXCL3) in our study. Mechanistically, PSM-derived gut microbiota significantly upregulated CXCL3 in the macrophages, and these Cxcl3 + macrophages contributed to colitis pathology by secreting CXCL3. The macrophages-derived CXCL3 exacerbated colitis via neutrophil chemotaxis and macrophage polarization. CXCL3 induced M2b-like polarization in macrophages, with functions related to immunomodulation and lipid catabolism. These macrophages exacerbated colitis in a gut microbiota-dependent manner. In terms of treatment, administration of Lactobacillus rhamnosus GG, a well-known probiotic, improved gut microbiota and CXCL3, and ameliorated the PSM-exacerbated colitis.

Conclusions

Gut microbiota was a key factor in PSM-exacerbated colitis, which was by activating macrophage to secrete CXCL3. Our study provides new insights into the role of gut microbiota with macrophages and chemokines in colitis, and the mechanism of oral disease affecting the distal organs systemically.

Graphical Abstract

Download video file ^{(47.4MB, mp4)}

Video Abstract

Supplementary Information

The online version contains supplementary material available at 10.1186/s40168-025-02218-3.

Introduction

Periodontitis, a common oral disease with high incidence caused by plaques, is a risk factor for multiple diseases, including inflammatory bowel disease (IBD) [1]. The pathways by which periodontitis affects systemic disease include systemic inflammation, immune responses, and oral bacteria [2]. Numerous studies have shown that the oral-gut axis is an important component of various systemic diseases, especially IBD [3, 4]. Thus, it is important to improve our understanding of the nature of oral-gut axis. Previous studies have shown that oral bacteria can alter the composition of gut microbiota, thereby contributing to host health and disease [5–7]. Kitamoto et al. showed that oral bacteria enter the intestine, activate oral pathobiont-reactive T-helper 17 cells, and promote colitis [8], revealing a potential physical platform for the oral-gut axis. However, the role of the salivary microbiota in systemic disease, which is closer to the clinical situation, has not been fully elucidated.

Saliva contains a large number of oral bacteria, which have different effects on colitis. The pathological role of landmark periodontitis bacteria, such as Porphyromonas gingivalis and Fusobacterium nucleatum, in colitis has been highlighted [2, 9]. Streptococcus salivarius, a commensal salivary bacterium, inhibited colonic inflammation in the colitis mouse [10]. It is well known that the proportion and composition of oral bacteria in the saliva of patients with periodontitis differs significantly from that of healthy individuals [11], indicating that the differences of salivary microbiota between healthy and periodontitis individuals may be pathogenic. Our previous studies have shown that, compared with healthy salivary microbiota (HSM), periodontitis salivary microbiota (PSM) alters the gut microbiota and exacerbates the development of a variety of systemic diseases [12–15]. Thus, we hypothesized that PSM exacerbated colitis through harmfully altering gut microbiota composition, and explored the underlying pathogenic mechanisms of PSM-derived gut microbiota.

The interaction between gut microbiota and immunity can be explained by pathogen-associated molecular patterns (PAMPs) or bacterial metabolites that bind to corresponding receptors and activate a series of immune responses [16]. PSM causes alterations in gut microbiota composition and microbiota-derived metabolites [12, 14], which have multiple outcomes. Therefore, we considered the gut microbiota caused by PSM as a whole and studied the polarization of the immune system induced by PSM-derived gut microbiota in the colitis. Macrophages, the major component of immunity, play a key role in the development of colitis and oral diseases [17, 18]. Unlike the traditionally recognized pro-inflammatory M1 macrophages, our previous study showed that PSM exacerbated colitis and increased the macrophage M2 polarization in dextran sulfate sodium (DSS)-induced colitis mice [12]. C-X-C motif chemokine 3 ligand (CXCL3) is a neutrophil chemokine that primarily promotes carcinogenesis and can be induced by protease-activated receptors (PARs) via oral bacteria; however, its role in colitis remains to be explored [19, 20]. These findings suggest an intrinsic, yet not fully understood, immunity to PSM.

In this study, we first confirmed the key role of the gut microbiota in PSM-exacerbated colitis using antibiotic treatment (ABX) and fecal microbiota transplantation (FMT). We then investigated the mechanisms, focusing on the immune response, by which PSM-derived changes in the gut microbiota exacerbate colitis. In addition, prebiotics and probiotics are considered as potential treatments for improving gut microbiota and gut immunity [21, 22]. Therefore, we explored the possibility of probiotics for the treatment of PSM-related colitis. Our study provides new insights into the oral-gut axis and its relationship with the immune response.

Results

Periodontitis salivary microbiota exacerbates colitis by gut microbiota

To study the effects of salivary microbiota on colitis, we collected salivary microbiota from 12 patients with periodontitis and 10 healthy individuals from the Nanjing Stomatological Hospital. According to the 16S rRNA sequencing, PSM contained 8199 amplicon sequence variations (ASVs), whereas HSM contained 3587 ASVs, with 1741 ASVs found in both PSM and HSM (Fig. 1A). In consistent, the α-diversity of PSM was higher than that of HSM (Fig. 1B). Thus, HSM and PSM exhibited different microbiota compositions according to principal coordinate analysis (PCoA) (Fig. 1C). The compositions of TOP10 gut microbiota at the genus levels are shown in Fig. 1D. Both HSM and PSM contained the same high-abundance genera, such as Neisseria, Prevotella, and Streptococcus, but their proportions were different. In HSM, Streptococcus was the most abundant genus, and in PSM, the sum of other low-abundance genera had the most abundant (Fig. 1E). According to linear discriminant analysis effect size (LEfSe), Treponema and Porphyromonas were represented differential genera in PSM, whereas Streptococcus and Veillonella were that in HSM (Fig. 1F). These studies indicated that the differences between HSM and PSM were the different proportions of high-abundance genera and that PSM had more low-abundance genera. PSM was microbiota represented by low-abundance bacteria, Treponema and Porphyromonas, whereas HSM was represented by Streptococcusas.

Fig. 1.

Periodontitis salivary microbiota exacerbated colitis by gut microbiota. A Venn diagram indicating amplicon sequence variant (ASV) between HSM and PSM. B The difference of α-diversity (Chao1, Faith_pd, and Shannon index) between HSM and PSM (Kruskal–Wallis). C Principal coordinate analysis based on Bray–Curtis of healthy salivary microbiota (HSM) and periodontitis salivary microbiota (PSM) (HSM n = 10; PSM n = 12). D The composition of HSM and PSM at the genus level. E The relative abundance of HSM and PSM at the genus level. F Linear discriminant analysis effect size (LEfSe) analysis was used to identify taxonomic differences between HSM and PSM (LDA > 2.5). G Schematic representation and study design of salivary microbiota-gavage DSS mice with or without antibiotic treatment (Ctrl, control group; DSS, the mice treated with DSS; HDSS, the DSS mice gavaged with healthy salivary microbiota; PDSS, the DSS mice gavaged with periodontitis salivary microbiota; ABX-H, HDSS group treated with antibiotic; ABX-P: PDSS group treated with antibiotic). H–M The changes of the colitis level between groups from 0 to 7 days, including weight loss (H), disease activity index (DAI) score (I), representative physical and statistical diagrams of the length of the colon (J), histological score (K), representative image of hematoxylin and eosin staining (L), and PAS staining (M) (n = 6). p values were determined by two-sided t test (B), two-way ANOVA (H and I; the statistical comparison referred to PDSS vs HDSS group), and one-way ANOVA with Tukey’s multiple comparisons test (J and K), *p < 0.05, **p < 0.01, ***p < 0.001, scale bar = 100 µm

To explore the role of PSM, HSM, and their-derived gut microbiota in colitis, we gavaged the salivary microbiota into mice for 14 days, and used antibiotic treatments to deplete gut microbiota for 7 days, then induced colitis using DSS for 5 days and sterile water for 2 days (Fig. 1G). Compared to HSM, PSM exacerbated colitis, as evidenced by weight loss, increased disease activity index (DAI), and decreased colon length (Fig. 1H-J). Histologically, a significant increase in colon lesions, including barrier destruction (Zo-1 and Occludin; Figure S1A), histological scores, and mucus cells disruption, was found in the PDSS group compared to those of the HDSS group (Fig. 1K–M). Importantly, ABX removed the difference of colitis between HSM- and PSM-treated mice (ABX-H and ABX-P group) (Fig. 1H–M), suggesting that the gut microbiota mediated the PSM affect colitis. Taken together, these results suggest that PSM exacerbated the DSS-induced colitis by gut microbiota.

PSM-derived gut microbiota is causal factor for colitis exacerbation

To study the role and mechanism of salivary microbiota-derived gut microbiota in colitis, fecal microbiota was collected from salivary microbiota-treated mice and gavaged into other mice. Compared to the HSM-derived gut microbiota (HSM-GM)-treated (FMT-H) group, mice in the PSM-derived gut microbiota (PSM-GM)-treated (FMT-P) group showed a significant decrease in body weight and an increased DAI score, which reached the highest point on day 7 (Fig. 2A–E), suggesting that the PSM-derived gut microbiota, but not the HSM-derived gut microbiota, was pathogenic. Supportively, the barrier (Zo-1 and Occludin) and mucus cell destruction was more evident in the FMT-P group than in the FMT-H group, with the FMT-P tissue histological score being significantly higher than the corresponding FMT-H score (Figs. 2F–H and S1B). To exclude the interference of commensal bacteria, we gavaged HSM-GM and PSM-GM into ABX-treated (gut microbiota depleted) mice. Similarly, colitis was more severe in the ABX-treated mice with PSM-GM gavage (ABX-FMT-P) than in those with HSM-GM gavage (ABX-FMT-H), with decreased body weight, increased DAI score, decreased colon length, and more severe histological destruction (Figure S2A–F). The gut microbiota was significantly different between ABX-FMT-P and ABX-FMT-H group, with a higher α-diversity in the ABX-FMT-P group than in the ABX-FMT-H group (Figure S2G and H). These results confirmed that the PSM-GM exacerbated colitis compared to HSM-GM.

Fig. 2.

PSM-derived gut microbiota is causal factor for colitis exacerbation. A Schematic representation and study design of fecal microbiota transplantation. The fecal was collected from salivary microbiota-gavage mice before DSS treatment (FMT, the fecal was from Ctrl group; FMT-H, the fecal was collected from HDSS group; FMT-P, the fecal was collected from PDSS group). B–H The changes of the colitis level between groups from 0 to 7 days, including weight loss (B), disease activity index (DAI) score (C), representative physical and statistical diagrams of the length of the colon (D and E), representative image of hematoxylin and eosin staining (F), histological score (G), and PAS staining (H) (n = 7). I Principal coordinates analysis of Bray–Curtis for metagenome sequence of healthy salivary microbiota-derived gut microbiota (HSM-GM) and periodontitis salivary microbiota-derived gut microbiota (PSM-GM) (n = 5). J Taxonomic differences of gut microbiota of HSM-GM and PSM-GM based by linear discriminant analysis effect size (LEfSe) analysis. K and L The top 20 relative abundance of gut microbiota of HSM-GM and PSM-GM group at the family level (K) and species level (L). M Functional differences of gut microbiota of HSM-GM and PSM-GM group analyzed by LEfSe analysis. p values were determined by two-way ANOVA (B and C; the statistical comparison referred to FMT-P vs FMT-H group) and one-way ANOVA with Tukey’s multiple comparisons test (E and G), **p < 0.01, ***p < 0.001, scale bar = 100 µm

To explore the function of PSM-GM and HSM-GM, we performed a metagenomic assay on the gut microbiota in the colitis mice inoculated with PSM and HSM. A significant difference in microbiota composition was observed between the PSM-GM and HSM-GM (Fig. 2I), with the most obvious differences observed in the Enterobacteriaceae, which was higher in the PSM-GM (Fig. 2J and K), represented by the species Escherichia coli (E. coli; Fig. 2L). Notably, Enterobacteriaceae was also significantly upregulated in PSM (Figure S3A). Similarly, Bacteroidaceae also increased in PSM and PSM-GM (Figs. 2K and S3A), and the relative abundance of oral bacteria, such as Haemophilus, Streptococcaceae, and Fusobacteriaceae, was increased in the PSM-GM (Figure S3B). Kyoto Encyclopedia of Genes and Genomes (KEGG) functional enrichment analysis showed that lipopolysaccharide (LPS) synthesis was a key pathway in the PSM-GM (Fig. 2M). Consistently, we found that the LPS levels in the PSM-GM were significantly higher than HSM-GM by Elisa (Figure S4A). The results of immunofluorescence and PCR showed that PSM-GM resulted in increased expression of TLR4, the receptor for LPS (Figure S5A and B). We gavaged LPS into mice and found that LPS exacerbated colitis in a concentration-dependent manner (Figure S4B–G), suggesting that LPS may be the key substance in the PSM-derived gut microbiota that exacerbates colitis. In addition, several metabolism pathways (e.g., glycan, carbohydrate, and amino acid) (Fig. 2M) and microbiome metabolizing enzymes (e.g., glycoside hydrolases and polysaccharide lyases) were significantly enriched in PSM-GM (Figure S3C). These results suggest that PSM led to gut microbiota disorder and this PSM-derived gut microbiota expand the lesions of DSS.

CXCL3 is the key mediator in PSM-derived gut microbiota exacerbating colitis

To assess the mechanisms by which PSM-GM affects colitis, we performed mRNA sequencing of colonic tissues in the FMT-H and FMT-P groups. The results showed significant group differences at the mRNA level (Fig. 3A). The volcano map of differentially expressed genes (DEGs; FC > 2, p < 0.05) showed that the expression levels of Cxcl3 (FC = 4.17, p = 1.9 × 10⁻⁶), Serpinb2 (FC = 3.45, p = 1.5 × 10⁻⁶), Cxcl13 (FC = 2.81, p = 1 × 10⁻⁷), and Cd79a (FC = 2.13, p = 1.2 × 10⁻⁸) were higher, whereas those of Lgr5 (FC = − 2.70, p = 3.1 × 10⁻⁸), Ugt2b35(FC = − 2.89, p = 3.9 × 10⁻⁷), and Adam18 (FC = − 4.5, p = 2.0 × 10⁻⁵) were decreased in the FMT-P group (Fig. 3B). The GeneOntology (GO) and KEGG functional enrichment analysis of DEGs showed that leukocyte migration, immune receptor activity, and cytokine interaction were the most significant pathways (Fig. 3C and D). Gene set enrichment analysis (GSEA) also showed significant upregulation of inflammatory responses to LPS, chemokine binding, and signaling pathways in the FMT-P group (Fig. 3E). We also examined representative genes of Treg and Th17 in colon and found that the Foxp3/Rorγt ratio was downregulated int the FMT-P group, suggesting PSM-GM disrupted the immune tolerance (Figure S5C-E). These results showed that cytokine and immune responses were the main functional difference between the FMT-H and FMT-P groups, with Cxcl3 and Cxcl13 being the predominant DEGs representing cytokines.

Fig. 3.

CXCL3 was the key mediator in periodontitis salivary microbiota-derived gut microbiota exacerbating colitis. A Differential gene expression heat map based on mRNA sequencing of FMT-H and FMT-P group. B Volcano plot indicated differential genes expression with high fold change (n = 7). C–E Functional enrichment of differential genes expression analysis by GO (C), KEGG (D), and GESA (E) between FMT-H and FMT-P group. F The relative expression of CXCL3 and CXCL13 at healthy induvial and patients with ulcerative colitis. G Immunofluorescence staining of tissues from normal colonic tissue and colonic lesions tissue of ulcerative colitis patients, CXCL3 (green) and CD11b (red). H Schematic representation and study design of fecal microbiota transplantation treated with CXCL3-neutralizing antibodies. The fecal was collected from salivary microbiota-gavage mice before DSS treatment (FMT, the fecal was from Ctrl group; FMT-H, the fecal was collected from HDSS group; FMT-P, the fecal was collected from PDSS group; FMT-P + A, the fecal was collected from PDSS group, and treated with anti-CXCL3 neutralizing antibodies). I–O The changes of the colitis level between groups from 0 to 7 days, including weight loss (I), disease activity index (DAI) score (J), representative physical and statistical diagrams of the length of the colon (K and L), histological score (M), representative image of hematoxylin and eosin staining (N), and PAS staining (O) (FMT, FMT-H, FMT-P + A group, n = 6; FMT-P group, n = 5). p values were determined by two-sided t test (F), two-way ANOVA (I and J; the statistical comparison referred to FMT-P + A vs FMT-P group in J), and one-way ANOVA with Tukey’s multiple comparisons test (L and M), *p < 0.05, **p < 0.01, ***p < 0.001, scale bar = 100 µm

We then examined the expression of CXCL3 and CXCL13 in periodontitis and colitis. For periodontitis, both Cxcl3 and Cxcl13 were upregulated in the colon of different periodontitis-associated mice (FMT group, ligated mice, and PSM gavage mice), and the overexpression degree of Cxcl3 was higher than that of Cxcl13 and statistically significant (Figure S6A and B). Consistently, higher serum CXCL3 levels were observed in the PDSS group than in the HDSS group (Figure S6C). For colitis, the mRNA sequencing data of 43 healthy individuals and 42 patients with ulcerative colitis (UC) from the Group on Earth Observations (GEO) database showed that CXCL3 and CXCL13 were significantly overexpressed in UC patients, and the expression of CXCL3 was more higher and statistically significant in patients with colitis compared to CXCL13 (CXCL3:FC = 1.36, p = 3.0 × 10⁻⁴; CXCL13:FC = 1.20, p = 0.029; Fig. 3F). The immunofluorescence results of colon tissues from UC patients showed that the expression of CXCL3 and macrophages was significantly higher in inflamed tissues than in healthy tissues (Fig. 3G). Similar results were obtained in the colon of DSS-treated mice, and the only highly expressed chemokine-related genes were ligands of CXCR2 (Cxcl1, Cxcl2, Cxcl3, and Cxcl5) in the DSS-treated mice (Figure S6D). Thus, compared with CXCL13, CXCL3 was significantly upregulated in both periodontitis and colitis, suggesting that CXCL3 is a key mediator in the periodontitis exacerbating colitis.

To verify the role of CXCL3 in PSM-GM exacerbated colitis, we injected CXCL3-neutralizing antibodies into the FMT-P group to deplete CXCL3 at day 0 and day 2 during the DSS treatment (Fig. 3H). CXCL3-neutralizing antibodies (FMT-P + A group) significantly improved colitis caused by PSM-GM, characterized as reduced weight loss (Fig. 3I), reduced DAI score (Fig. 3J), and longer colon (Fig. 3K and L). Histologically, a significant decrease in histological scores (Fig. 3M) and the destruction of colon barrier (Fig. 3N) and mucus cells (Fig. 3O) were observed in the FMT-P + A group, compared to the FMT-P group. Consequently, CXCL3 was a key mediator in the PSM-GM exacerbating colitis.

CXCL3 is upregulated in macrophages induced by PSM-derived gut microbiota and exacerbates colitis

Next, we explored the origin cell of CXCL3 in the PSM-GM exacerbates colitis. According to single-cell sequencing data from patients with UC, CXCL3 was mainly expressed in transit-amplifying cells, macrophages, and enterocytes (Fig. 4A and B), and had the highest expression per cell in monocyte-macrophages (Fig. 4C). CXCL3 was highly expressed in inflamed macrophages and positively correlated with inflammation-related genes, such as CXCL2, NFKBIA, and IL1B, suggesting an inflammation-related role of CXCL3 (Fig. 4D). Similar results were observed in patients with Crohn’s disease (Figure S6E–H). Thus, CXCL3 is mainly expressed in macrophages of patients with colitis.

Fig. 4.

CXCL3 was upregulated in macrophage induced by PSM-derived gut microbiota and exacerbated colitis. A The number and type of top 10 CXCL3-positive cells in colon tissue of ulcerative colitis (UC) patients. B The graph of t-distributed stochastic neighbor embedding (t-SNE) subgroups for colon tissue of UC patients. C The top 10 cells of CXCL3 relative high expression in colon tissue of UC patients. D The correlation of CXCL3-related gene obtained based on the Pearson analysis in colon tissue of UC patients. E The expression of CXCL3 in Raw 264.7 and THP-1 after treatment with fecal supernatant. The fecal was collected from salivary microbiota-gavage mice before DSS treatment (n = 3, each test is independent; PBS-GM, the fecal was from PBS-gavage mice; HSM-GM, the fecal was collected from HSM-gavage mice; PSM-GM, the fecal was collected from PSM-gavage mice). F Immunofluorescence staining of colon tissues from FMT group, CXCL3 (green) and F4/80 (red). G The representative image of CXCL3 level analyzed by flow cytometry in cell line Raw 264.7 treat with fecal supernatant. H Schematic representation and study design of tail vein injection of Cxcl3⁺ macrophages (DSS + Mø, tail vein injection of blank virus-treated macrophages in DSS mice; DSS + Mø^cxcl3+, tail vein injection of Cxcl3⁺ macrophages in DSS mice). I–N The changes of the colitis level between groups from 0 to 7 days, including weight loss (I), disease activity index (DAI) score (J), representative physical and statistical diagrams of the length of the colon (K), representative image of hematoxylin and eosin staining (L), histological score (M), and PAS staining (N) (n = 6). p values were determined by one-way ANOVA with Tukey’s multiple comparisons test (E, K, and M), two-way ANOVA (I and J), and two-sided t test (P). *p < 0.05, **p < 0.01, ***p < 0.001, scale bar = 100 µm

We stimulated different cells with the fecal microbiota of PSM-gavaged mice and found that PSM-GM stimulated Cxcl3 overexpression in Raw264.7, THP-1-derived macrophages, and bone marrow-derived macrophages (BMDM) (Figs. 4E and S7A–B), and had no significant effect on Cxcl3 expression in Caco-2(Figure S7C and D), suggesting that Cxcl3 was mainly expressed in macrophages due to PSM-GM. The results of immunofluorescence showed an increase in macrophage numbers (F4/80) and CXCL3 levels in the FMT-P group (Fig. 4F). Moreover, PSM-GM increased the number of Cxcl3⁺ Raw264.7 cells compared with HSM-GM group, and ABX prevented this change, suggesting that this is an effect of microbiota in the PSM-GM (Fig. 4G). These results revealed that PSM-GM upregulate CXCL3 expression in macrophages.

To assess the role of Cxcl3 on macrophages, we generated Cxcl3-overexpression Raw264.7 cells (Cxcl3⁺ macrophage) by viral transfection and injected them into the tail veins of DSS-treated mice (Fig. 4H). Compared to DSS-treated mice, macrophage injection improved weight loss, DAI score, and colon length, whereas Cxcl3⁺ macrophages (DSS + Mø^cxcl3+ group) exacerbated the DAI score, weight, and colon length loss compared to normal macrophage injection (DSS + Mø group) (Fig. 4I–K). Compared with macrophages, Cxcl3⁺ macrophages exacerbated colonic destruction and mucus cell reduction (Fig. 4L–N). Collectively, these results suggest that PSM-driven gut microbiota exacerbate colitis by upregulating CXCL3 in macrophages.

Cxcl3+ macrophages partially exacerbate colitis by CXCL3-neutrophilic chemotactic axis

To explore how Cxcl3⁺ macrophages are involved in the development of colitis, we first sequenced Cxcl3⁺ macrophages (CXCL3 group) and blank virus-treated macrophages (NC group) and identified 111 DEGs (CXCL3 group versus NC group; p < 0.05, log2FC > 1; upregulated 38, downregulated 73). Cxcl3 was most markedly upregulated, followed by Shf and Efnb2, whereas significantly downregulated genes included Nlgn3, Tgfb2, and Cxcr2 (Fig. 5A and B). DEG enrichment analysis revealed that the main functions of Cxcl3⁺ macrophages were neutrophil degranulation, G protein-coupled receptor (GPCR)-related pathways, and chemokine receptor-binding chemokines (Fig. 5C). Cxcl3 was involved in all pathways (Fig. 5D), and the critical genes for Cxcl3⁺ macrophages were Cxcl3, Mapk12, and Rasl10b according to the STRING-based protein interaction network analysis (Fig. 5E). These studies suggest that Cxcl3 is a core gene in Cxcl3⁺ macrophages. Cxcl3 encodes the chemokine CXCL3; thus, we examined CXCL3 levels, and found that CXCL3 levels in the supernatant of the Cxcl3⁺ macrophages were significantly higher than that those in the blank virus-treated macrophages (Fig. 5F). Similarly, serum CXCL3 levels were increased in DSS + Mø^cxcl3+ group than DSS + Mø group (Fig. 5G), suggesting the key role of Cxcl3⁺ macrophages were secreting CXCL3.

Fig. 5.

Cxcl3 + macrophages partially exacerbated colitis by CXCL3-neutrophilic chemotactic axis. A Volcano plot indicated differential gene expression based on mRNA sequencing of blank virus-treated (NC group) and Cxcl3⁺macrophages (CXCL3 group) (CXCL3 vs NC group; n = 3). B The radar graph of top 30 differential genes expression of blank virus-treated and Cxcl3⁺macrophages. C and D Functional enrichment of differential genes expression analyzed by reactome between blank virus-treated and Cxcl3⁺ macrophages. E The interaction network analysis based on String. Red indicates upregulated genes, and blue indicates downregulated genes. The size of the node indicates the number of related genes, the larger the node the more related genes (CXCL3 vs NC group; n = 3). F The expression of CXCL3 in the supernatant of blank virus-treated and Cxcl3⁺macrophages. G The serum level of CXCL3 in the DSS mice of tail vein injected with macrophages. H Immunofluorescence staining of colon tissues, Ly6G (red) and NE (green). I Schematic representation and study design of treated with Ly6G- and CXCL3-neutralizing antibodies (DSS + Mø^cxcl3+ + Anti-Ly6G, tail vein injection of Cxcl3⁺ macrophages, and treated with anti-Ly6G neutralizing antibodies in DSS mice; DSS + Mø^cxcl3+ + Anti-CXCL3, tail vein injection of Cxcl3⁺macrophages, and treated with anti-CXCL3 neutralizing antibodies in DSS mice). J–O The colitis level between groups from 0 to 7 days, including weight loss (J), disease activity index (DAI) score (K), representative physical and statistical diagrams of the length of the colon (L), histological score (M), representative image of hematoxylin and eosin staining (N), and PAS staining (O) (n = 6, *p < 0.05, **p < 0.01, ***p < 0.001). p values were determined by two-sided t test (F and G), two-way ANOVA (I and J), and one-way ANOVA with Tukey’s multiple comparisons test (L and M). *p < 0.05, **p < 0.01, **p < 0.001, scale bar = 100 µm

Due to CXCL3 chemotaxis to neutrophils, we assessed colon neutrophils in the colitis mice using Wright’s staining, which showed that neutrophil was enriched in the PDSS and FMT-P groups (Figure S8A and B). ABX significantly reduced neutrophil (Figure S8C). Moreover, Cxcl3⁺ macrophages increased the expression of Ly6G and neutrophil elastase (NE), suggesting the production of neutrophil extracellular traps (Fig. 5H). To verify the role of neutrophils in Cxcl3⁺ macrophages exacerbating colitis, we treated mice with Ly6G- and CXCL3-neutralizing antibodies to deplete neutrophils and CXCL3 (Fig. 5I). Cxcl3⁺ Macrophages significantly increased the colon neutrophils (Ly6G) and proportion of serum neutrophils, whereas CXCL3-neutralizing antibodies decreased the proportion, suggesting that CXCL3 affects neutrophils (Figure S8D and E). Both Ly6G- and CXCL3-neutralizing antibodies ameliorated Cxcl3⁺macrophage-exacerbated colitis, leading to reduced body weight loss, reduced increase in DAI index, and restoration of colon length, but this improvement was more pronounced with the CXCL3-neutralizing antibody than Ly6G-antibody (Fig. 5J–L). The histological results showed that CXCL3-neutralizing antibodies significantly improved histological scores, inflammatory cell infiltration, and mucus cell in the DSS + Mø ^Cxcl3+ group (Fig. 5M–O). Thus, depletion of neutrophils partially alleviated Cxcl3⁺ macrophages-exacerbated colitis compared to depletion of CXCL3, suggesting Cxcl3⁺ macrophages partially exacerbate colitis by CXCL3-neutrophilic chemotactic axis, but CXCL3 have other roles in the pathogenesis of colitis.

Macrophage-derived CXCL3 exacerbates colitis by macrophages M2b-like polarization in colitis mice

Since macrophages were key cells in PSM exacerbating colitis, to explore the other roles of CXCL3, we investigated the effects of CXCL3 on macrophages. CXCL3 promoted macrophage proliferation, and the addition of CXCL3-neutralizing antibodies prevented this effect (Fig. 6A). Similarly, PSM-GM also promoted macrophage proliferation at 24 h, and CXCL3-neutralizing antibodies reduce this proliferative response, especially at 12 h (Fig. 6B). As previous study has found that PSM induced macrophage M2 polarization in DSS mice [12], we examined the polarization of CXCL3-treated macrophage. The ratio of CD86⁺/CD206⁺ was decreased in the CXCL3-treated macrophages compared to LPS-induced (M1) and Ctrl (M0) macrophages, whereas was close to that in IL4-induced (M2) macrophages (Fig. 6C). Similarly, CXCL3 significantly upregulated the CD206⁺/CDC86⁺ ratio in BMDM (Figure S9A). CXCL3 also upregulated the mRNA expression of Il10 and Arg1 in the macrophage (Figure S9B and C). Thus, CXCL3 affects the function of macrophage, including proliferation and M2 polarization.

Fig. 6.

Macrophage-derived CXCL3 exacerbated colitis by macrophages M2b-like polarization in colitis mice. A The proliferation of RAW264.7 treated with CXCL3 using CCK-8 assay. B The proliferation of RAW264.7 treated with fecal supernatant and CXCL3-neutralizing antibodies using CCK-8 assay (PBS-GM, the fecal was from PBS-gavage mice; HSM-GM, the fecal was collected from HSM-gavage mice; PSM-GM, the fecal was collected from PSM-gavage mice). C Representative FACS plots of expression of CD86 and CD206 in Raw264.7 cells treated with LPS, IL4, and CXCL3. D The heatmap expression of differentially expressed genes between control macrophage (Ctrl group) and CXCL3-treated macrophage (CXCL3 group). E Volcano plot indicated differential genes expression with high fold change (n = 3). F Functional enrichment of differential genes analyzed by KEGG in the CXCL3-treated macrophages. G and H The functions of upregulated (G) and downregulated (H) differential genes analyzed by KEGG in the CXCL3-treated macrophages. I The heatmap expression of M2-related genes in the CXCL3-treated macrophages. J Schematic representation and study design of CXCL3-treated macrophage and ABX treatment (DSS + Mø, tail vein injection of blank virus-treated macrophages in DSS mice; DSS + CXCL3-Mø, tail vein injection of CXCL3-treated macrophages in DSS mice; DSS + Mø + ABX, DSS + Mø group treated with antibiotic in DSS mice; DSS + CXCL3-Mø + ABX, DSS + CXCL3-Mø group treated with antibiotic in DSS mice). K–P The changes of the colitis level treated with CXCL3-treated macrophage and ABX, including weight loss (K), DAI score (L), representative physical and statistical diagrams of the length of the colon (M), histological score (N), representative image of hematoxylin and eosin staining (O), and PAS staining (P) (n = 6). p values were determined by one-way ANOVA with Tukey’s multiple comparisons test (A, B, K, and L) and two-way ANOVA (C, M, and N). *p < 0.05, *p < 0.05, **p < 0.01, ***p < 0.001, scale bar = 100 µm

To explore the exact role of these M2 macrophages, we sequenced the transcriptome of CXCL3-treated macrophages. The genes of CXCL3-treated macrophages (CXCL3 group) and control macrophages (Ctrl group) were significantly different (Fig. 6D). The upregulated representative genes of CXCL3-treated macrophages were Htra4, Cxcl10, Gbp2b, and Gdnf, and the downregulated representative gene was Car6 (Fig. 6E). Function enrichment analysis showed that the DEGs were mainly involved in responses to bacterial and defense responses (Fig. 6F). The functions of upregulated DEGs were related to immune response, including inflammatory response, regulation of programmed death, and cytokine-mediated signaling pathways (Fig. 6G). The functions of downregulated DEGs were related to lipid catabolism and cell recognition (Fig. 6H). To determine the subtype of M2 macrophages, we noted the expression of DEGs related to M2 macrophage. Compared with control macrophages, the expression of M2b-related gene, such Il-1β, Il-10, Il-6, Il1rn, and Socs3, were increased in CXCL3-treated macrophages, whereas the remaining genes related to M2 macrophage subtypes (M2a Ccl17, Ccl22; M2c Tgfb1-3; and M2d Vegfa, Vegfb) were relatively disturbed (Fig. 6I). The results of qPCR showed that CXCL3 upregulated M2b-related genes (Sphk1, Il10, and Light) in the macrophage (Figure S9B and C). These results showed that CXCL3 induces functional and genetic alterations in macrophages, with a tendency towards M2b polarization.

To assess the role of CXCL3-treated macrophages in colitis, we injected control (DSS + Mø group) and CXCL3-treated macrophages (DSS + CXCL3-Mø group) into the tail veins of DSS-treated mice (Fig. 6J). Compared to DSS + Mø group, CXCL3-treated macrophages exacerbated colitis, as evidenced by an increase in weight loss, colon length loss, DAI score, histological score, and mucus cell destruction in DSS-treated mice (Fig. 6K–P). Previous studies have shown that the M2b macrophages were associated with pathogens in disease progression [23]. Thus, to investigate the role of gut microbiota in M2b macrophages-exacerbated colitis, we use antibiotics to deplete gut microbiota. Antibiotic treatment removed the differences in weight loss, DAI, and histology and reversed the decrease in colon length between the DSS + Mø and DSS + CXCL3-Mø groups, suggesting that the exacerbation of colitis by CXCL3-treated macrophages is dependent on the gut microbiota (Fig. 6K–P). The above studies suggest that macrophage-derived CXCL3 lead to M2b-like polarization, and that such macrophages significantly exacerbated colitis by gut microbiota.

Lactobacillus rhamnosus GG ameliorates PSM-exacerbated colitis by improving gut microbiota and CXCL3

The above studies showed that gut microbiota and CXCL3 were key factors in the exacerbation of colitis by PSM, so finally we attempted to explore a treatment for colitis targeting them. Although antibiotics significantly improved colitis (Fig. 1A–H), our study found a high mortality rate in DSS-treated mice following ABX (Figure S10A). Thus, we used the prebiotics and probiotics for treatment, represented by Lactobacillus rhamnosus GG (LGG) and puerarin (PUE), which have been shown to improve gut microbiota and colitis in previous studies [24, 25]. PUE had no significant effect on Cxcl3 expression in the colon and Raw264.7 cells (Figure S10B and C). LGG administration significantly improved colitis (Figure S10D–I), and downregulated the mRNA expression Cxcl3 in the colon (Fig. 7A) and serum CXCL3 levels (Figures S10J). Consistently, the results of immunofluorescence showed that LGG significantly decreased CXCL3 and macrophage levels in colon lesion tissue (DLGG vs DSS group; Fig. 7B). For the gut microbiota, although LGG had no significant change in α-diversity (Figure S10K), LGG altered the composition of the gut microbiota, with a significant increase in the relative abundance of beneficial bacterium Allobaculum and Bifidobacterium (Fig. 7C and D). After treatment of LGG, the relative abundance of oral bacteria significantly was decreased in the gut microbiota, including Porphyromonadaceae, Bacteroidaceae, Enterobacteriaceae, and Streptococcaceae (Figure S3D). Thus, LGG can improve the gut microbiota and CXCL3 expression in DSS mice.

Fig. 7.

Lactobacillus rhamnosus GG ameliorated periodontitis salivary microbiota-exacerbated colitis by improving gut microbiota and CXCL3. A The colonic CXCL3 expression using RT-PCR of DSS-induced colitis mice with (DLGG) or without (DSS) LGG treatment (n = 6). B The representative image of colon immunofluorescence staining from DSS and DLGG group, CXCL3 (green) and F4/80 (red). C Principal coordinates analysis of Bray Curtis between DSS and DLGG group (n = 6). D Taxonomic differences of gut microbiota of DSS and DLGG group by LEfSe analysis. E Schematic representation and study design of LGG treatment in salivary microbiota-gavaged DSS mice (Ctrl, control group; DSS, the mice treated with DSS; HDSS, the DSS mice gavaged with healthy salivary microbiota; PDSS, the DSS mice gavaged with periodontitis salivary microbiota; PDSS + LGG, the DSS mice gavaged with periodontitis salivary microbiota and LGG). F–K The colitis level between groups from 0 to 5 days, including weight loss (F), DAI score (G), statistical diagrams of the length of the colon (H), representative image of hematoxylin and eosin staining (I), histological score (J), and PAS staining (K) (n = 6). p values were determined by two-sided t test (A), two-way ANOVA (F and G; the statistical comparison referred to PDSS vs PDSS + LGG group), and one-way ANOVA with Tukey’s multiple comparisons test (H and J). *p < 0.05, **p < 0.01. DAI, disease activity index; LGG, Lactobacillus rhamnosusGG, scale bar = 100 µm

To verify the effects of LGG on PSM-induced colitis, PSM-gavaged mice were treated with LGG for 2 weeks and DSS for 5 days (PDSS + LGG group; Fig. 7E). Colitis on day 5 was significantly worse in the PDSS group than in the HDSS group, including decreased body weight and higher DAI score, decreased colon length, higher histological score, and lower mucus cell number (Fig. 7F–K). LGG significantly improved colitis in the PDSS group at day 5 (PDSS + LGG vs PDSS group; Fig. 7F–K), where weight loss in the PDSS + LGG group intermediate between the PDSS and HDSS groups (Fig. 7F), and the DAI score, colon length, histological score, and mucus cell content were better than or equal to those in the HDSS group (Fig. 7H–K). These findings showed that LGG ameliorated PSM-induced colitis by improving the gut microbiota and downregulating CXCL3.

Discussion

Periodontitis is a common oral disease that affects many systemic diseases [3], and the underlying mechanisms warrant investigation. The oral cavity is the second largest microbiota reservoir, and alterations in the oral microbiota have been observed in various diseases, such as diabetes mellitus, cancer, autoimmune diseases, and IBD, suggesting that oral bacteria are involved in the development of systemic diseases [26]. Here, our study confirms that the gut microbiota induced by PSM is a key factor in periodontitis-exacerbated colitis, and that this gut microbiota exacerbates colitis by activating Cxcl3⁺ macrophages, which then secrete of CXCL3.

The total number of bacteria in the oral cavity is 10¹¹–10¹², whereas only 10⁷ bacteria are found in the stomach and intestinal tract; thus, it is a worthwhile debate on how oral bacteria affect the gut microbiota and intestinal diseases [16]. Some reports have shown that oral microbiota can be found in the gut, such as F. nucleatum and Campylobacter concisus, suggesting that oral bacteria exacerbate intestinal disease by colonizing the intestine [9, 27, 28]. Our study also showed that the relative abundance of oral bacteria increased in the PSM-GM, and the high-abundance gut microbiota in PSM-GM-treated mice were also increased in PSM, such as Enterobacteriaceae and Bacteroidaceae. These studies showed that PSM can lead to the “oralization” of gut microbiota, regardless of single type of periodontal pathogen or mixed salivary microbiota [5, 29]. The probiotics improves these compositions and colitis, suggesting that PSM exacerbates colitis by altering the composition of the gut microbiota. Besides oral bacteria, PSM has a variety of effects on the gut microbiota components. For example, E. coli, a pathogen that is highly expressed after PSM treatment, was associated with IBD [30]. PSM can also lead to metabolic disorders, such as arachidonic acid metabolism, and worsen colitis [12]. These virulence factors of pathogens, such as LPS and peptidoglycans, increase intestinal permeability and promote the entry of PAMPs into the intestine by interacting with pattern recognition receptors [31, 32]. To facilitate the study of the mechanisms by which PSM-GM affect colitis, we considered the disordered gut microbiota as a whole and explored the immune response induced by PSM-GM. Among these, LPS is the most representative component, which was enriched in the PSM-GM. PSM can activate the TLR4, a pattern recognition receptor for LPS, and induce inflammatory responses [14]. In addition, PSM also lead to metabolite alterations through a variety of microbial enzymes, which may also affect the immune response. Thus, the effective components of PSM-GM in PSM-exacerbated colitis still need to be clarified.

The immune response is an integral part of the development of colitis and is associated with the gut microbiota, especially macrophages [33]. In our study, PSM activated Cxcl3⁺ macrophages and subsequently exacerbated colitis, which was related to the secretion of CXCL3. CXCL3 is a chemokine that induces neutrophil chemotaxis, and modulates cellular behaviors by ERK [34]. Previous studies have shown that periodontal bacteria can activate PAR in oral keratinocytes, and the activation of PARs increases CXCL3 expression through ERK1/2 [19]. Consistent with previous studies, our study showed that CXCL3 exacerbated colitis by M2 macrophage polarization [12]. However, M2 macrophages are traditionally considered to be anti-inflammatory; the role of macrophage M2 polarization in colitis needs to be explored. Pathogens tend to induce macrophage M2 polarization by releasing cytokines to escape macrophage bactericidal action and facilitate infection [35]. Enrichment of pathogen, such as E. coli, in PSM may contribute to exacerbation of colitis. TcpC is a virulence factor of E. coli, which evades macrophage and promoting macrophage M2 polarization [36]. For subtype, our results showed that CXCL3 lead to the alteration of immunoregulatory and lipid metabolic functions, which was M2b macrophage feature [37]. Recent studies have shown that CXCL3 is a characteristic of M2b macrophages [38]. M2b monocytes/macrophages not only have no antimicrobial effect but also increase susceptibility to opportunistic pathogens [23, 39, 40]. DSS directly induces polarization of M2b macrophages, suggesting that an increase in pathogens, such as E. coli, is responsible for DSS-induced colitis [41]. Consistently, our study also indicates that this macrophage exacerbated colitis depending on the gut microbiota. Therefore, we hypothesized that the significant enrichment of E. coli in PSM-GM may be responsible for PSM-exacerbated colitis, which is caused by the interaction of gut microbiota, macrophage and CXCL3. However, it remains to be studied whether the upregulation of E. coli is a direct effect of PSM or secondary to the colitis. Furthermore, macrophage polarization is a dynamic process that is influenced by different cells. For example, periodontitis can lead to disruption of Th17/Treg homeostasis and loss of immune tolerance, which may account for the upregulation of CXCL3 in macrophage by PSM. Therefore, the effect of different immune cells on macrophages and the polarization states of macrophages on colitis still needs to be investigated.

Disruption of the intestinal barrier is an important aspect of PSM that exacerbates colitis. DSS impairs intestinal barrier function, which enabled gut bacteria to contact the cells directly and play a role in exacerbating colitis [42]. Our study found that PSM also causes the downregulation of intestinal Lgr5 expression, impairing the intestinal epithelium and extracellular matrix in mice with colitis, which may be related to the recruitment of neutrophils. Lgr5⁺ intestinal stem cells (ISCs) can differentiate into all types of intestinal epithelial cells and maintain the integrity of the intestinal epithelium [43]. During the repair of damaged intestinal tissue, Lgr5⁺ ISC interact with stromal and immune cells, participating in inflammation, immune homeostasis, and tissue repair. When these processes fail, the lamina propria immune system is activated, triggering IBD [44]. Neutrophils, which can be chemotactic by CXCL3, are the first immune cells to be recruited to the inflammatory site. Neutrophil extracellular traps can capture and kill pathogens by locally delivering high concentrations of antimicrobial substances. However, if this process is uncontrolled, the feedback is amplified, leading to severe tissue damage [45]. Thus, CXCL3, neutrophils, and the intestinal barrier are important mediators in the PSM affecting intestinal disease. However, the underlying mechanisms in the interaction between PSM, immune responses, and intestinal epithelial cells require further investigation. In addition, the differences between DSS and IBD are noteworthy. Although DSS is a classical model for UC patients, it is not an exact replica of human colitis [46]; thus, the mechanism by which PSM affects IBD still needs to be validated in clinical samples.

Periodontitis is prone to recurrence as plaque can reappear within 16 h, and improving periodontitis requires the cooperation of patients and doctors, which makes treatment difficult [47]. The gut microbiota promotes the maturation of intestinal immunity; therefore, treating periodontitis-induced colitis by improving the gut microbiota is the preferred approach [48]. Although antibiotics can deplete gut pathogens, these drugs are also harmful to the gut microbiota and cause drug resistance [49]. Our study showed that although ABX improved colitis, antibiotic-treated mice had a higher mortality rate after DSS treatment. Commensal bacteria contribute to host health by catabolizing the diet and eliminating pathogens such as Klebsiella pneumoniae, which is a major gut pathogen in periodontitis-exacerbated colitis [8]. Commensal bacteria also produce lactic acid, which activates GPCRs on Paneth and stromal cells and promotes the regeneration of intestinal epithelial cells, shaping mucosal immunity and host-bacterial homeostasis [50]. Thus, the essence of PSM-exacerbated colitis is the alteration of commensal bacteria, and the disordered gut microbiota directly influences immune cells due to DSS-induced intestinal barrier disruption. To address this, we chose probiotics as the therapeutic tool. Lactobacillus reuteri can maintain the number of Lgr5⁺ ISC, promote the development of intestinal epithelial cells, and reduce intestinal pro-inflammatory cytokines and serum LPS concentrations [24, 50]. Consistent with previous studies, our study confirmed that probiotic therapy is an effective and safe treatment option for the diseases caused by periodontitis [51]. With the growing recognition of the importance of the gut microbiota, probiotics are an available treatment option for the management of systemic diseases.

In conclusion, our study shows that PSM promotes the development of colitis through the gut microbiota, a process in which macrophage-derived CXCL3 plays a key role. We have confirmed that the gut microbiota is a key factor in periodontitis-affected systemic diseases, promoting disease progression by inducing macrophages and chemokine. Our study provides evidence for the relationship between oral diseases and systemic diseases, and new perspectives on the role of chemokines and macrophage in the development of colitis.

Method

Collection of saliva samples

Saliva samples of 12 patients with periodontitis and 10 healthy individuals were collected at Nanjing Stomatological Hospital, Medical School of Nanjing University. All donors gave informed consent, and approval was granted by the Ethics Committee of Nanjing Stomatological Hospital, Medical School of Nanjing University (2019NL-008(KS)). All examination and diagnoses were made by experienced periodontists. The description of donor’s periodontal status is shown in Appendix Table 1. Before collection, patients were not allowed to rinse mouth to ensure that the collected saliva was naturally secreted. Including and excluding criteria were as follows referenced to the 2017 World Workshop on the Classification of Periodontal and Peri-Implant Diseases and Conditions:

Including criteria

(1) Aged between 25 and 55; (2) 18 teeth retained in the mouth; and (3) two or more of the patient’s affected teeth have three of the following characteristics: (a) probing depth > 6 mm, (b) adhesion loss of 5 mm, (c) alveolar bone was absorbed for more than half of the root length.

Excluding criteria

(1) Received periodontal treatment within the last year; (2) prolonged treatment with antibiotics or other medication within the last 6 months; (3) patients with systemic diseases, such as diabetes, immune deficiency, chronic gastrointestinal diseases, and cerebrovascular diseases; (4) patients suffering from severe inflammation-related oral diseases, such as periodontal abscesses; (5) pregnancy or lactation; (6) smokers (> 5 cigarettes/day).

Processing and treatment with salivary microbiota

The collected saliva was centrifuged at 1000 rpm for 10 min, and the supernatant was collected. The supernatant was then centrifuged at 3300 × g for 10 min at 4 °C and suspended in phosphate-buffered saline (PBS; 1 mL PBS per 5 mL of saliva) to collect salivary microbiota for subsequent experiments. The remaining salivary microbiota was suspended in an equal volume (w/v) of PBS containing 20% glycerol, frozen rapidly in liquid nitrogen, and stored at − 80 °C until use. When used, to ensure that each mouse received the same salivary microbiota, same volume of the frozen salivary microbiota from the same group was mixed, and then centrifuged at 3300 × g for 10 min, and the deposit was suspended in PBS. Before DSS treatment, the Ctrl, DSS, and ABX groups were gavaged with PBS, whereas the HDSS, ABX-H groups, and PDSS, ABX-P groups were gavaged with HSM and PSM, respectively, every other day for 2 weeks (200 µL processed salivary microbiota per SPF mouse).

Animal models

Eight-week-old wild-type (WT) C57BL/6J male mice were purchased from Beijing Vital River Laboratory and raised under specific pathogen-free (SPF) conditions with a 12-h light/dark cycle. Each group contained 6–7 mice. After 1 week of acclimatization, the mice were randomly assigned to groups. The experimental timeline is shown in the schematic diagram, and the experimental details are given in the “Methods” section. The animal experiments were approved by the Animal Ethics Committee of Nanjing Agriculture University, Nanjing, China (No. PZW2021020, PZW2022053).

Cell culture

Raw264.7 cells were cultured in Dulbecco’s Modified Eagle Medium (4.5 g/L glucose), Caco-2 cells were cultured in Minimum Essential Medium α (MEM α) with 1% non-essential amino acids (v/v), and THP-1 cells were cultured in Roswell Park Memorial Institute-1640 medium. THP-1 cells were induced with 100 ng/mL phorbol myristate acetate. The cell culture medium was supplemented with 10% (v/v) fetal bovine serum, 100 U/mL penicillin, and 100 mg/mL streptomycin, but penicillin and streptomycin were not added to the cell culture medium in the fecal transplantation experiments. All cells were kept at 37 degrees in 5% CO₂, 100% humidity.

To isolate bone marrow-derived macrophages (BMDMs), the mice was sterilized with 70% ethanol, flush marrow cavities of femurs with PBS using syringe and dissociate clumps via pipetting. Then, lyse red blood cells with Red Blood Cell Lysis Buffer, 300 × g/5min centrifugation, resuspend cells in complete medium (RPMI-1640 with 10% FBS and antibiotics), and seed (5 × 10⁵/mL per well) in culture plates supplemented with 30 ng/mL M-CSF (CAT#315–02, ThermoFisher). Incubate at 37°C/5% CO2, replenish fresh cytokine-containing medium on day 3, and allow differentiation for 7 days until adherent macrophages exhibit elongated morphology. For downstream assays, PBS washing and then harvest macrophages using trypsin–EDTA.

Bacteria

Lactobacillus rhamnosus GG was purchased from ATCC (ATCC53103). The solid medium for the Lactobacillus rhamnosus GG strain was MRS Broth (52 g) (CAT#HB0384-1/HB0384, Hope biotech) and MRS Agar (62 g) (CAT#HB0384, Hope biotech) dissolved in 1 L of water. The liquid medium used for passaging was formulated as MRS Broth (52 g) dissolved in 1 L of water. The medium was then autoclaved and cooled. The bacteria were grown under anoxic conditions and identified using full-length 16S rRNA gene sequencing.

Dextran sulfate sodium-induced colitis

To establish a colitis model, the mice were administered 3% dextran sodium sulfate (CAT#0216011090, MP Biomedical) ad libitum for 5 days, followed by sterile water for 2 days. The DAI was measured based on the following criteria: weight loss (0, none; 1, 1–5%; 2, 5–10%; 3, 10–20%; 4, > 20%), stool consistency, and hemoccult (0, normal; 2, hemoccult positive; 4, gross blood). Histological scores were measured as follows: severity of inflammation (0, none; 1, low density confined to the mucosa; 2, moderate or higher density in the mucosa or low-to-moderate density in both mucosa and submucosa; 3, high density in the submucosa or extension to the muscularis; 4, high density with frequent transmural extension) or extent of epithelial/crypt damage (0, none; 1, basal 1/3; 2, basal 2/3; 3, crypt loss; 4, crypt and surface epithelial destruction). Each variable was multiplied by a coefficient for the percentage of colon involvement (1, 0–25%; 2, 26–50%; 3, 51–75%; 4, 76–100%), and the total score was the sum of all variables.

Antibiotic treatment

The antibiotic solution was prepared by mixing ampicillin (1 mg/mL), vancomycin (0.5 mg/mL), neomycin (1 mg/mL), and metronidazole (1 mg/mL) in sterile water, and the mice drink the water with antibiotics ad libitum when used. To deplete salivary microbiota-driven gut microbiota, during treatment with the salivary microbiota, the ABX, ABX-H, and ABX-P groups drink the water for the last 1 week prior to DSS treatment in an SPF environment. For removing the effect of gut symbiotic bacteria, some mice drink the water prior to FMT for 2 weeks (ABX-FMT, ABX-FMT-H, and ABX-FMT-P group).

Fecal microbiota transplantation

Fresh feces were collected from mice treated with salivary microbiota (before treatment with DSS) and processed immediately for 2 weeks. The feces were stirred and centrifuged at 1000 rpm for 5 min in 1 mL of sterile PBS, and the supernatant was collected. The suspension (200 µL) was immediately gavaged into each mouse every other day for 2 weeks. Approximately one pellet of feces per mouse was used in this study. The FMT and ABX-FMT group were treated with the feces of the Ctrl group, and the FMT-H, ABX-FMT-H, and FMT-P, ABX-FMT-P groups were treated with the feces of the HDSS and PDSS groups, respectively. For cell experiments, 1% (v/v) of the above bacterial solution was added to the cells and incubated for 12 h before testing.

Protein, probiotic and drug treatments

In animal experiments, 20 µg/mL CXCL3 recombinant protein (CAT#D620677-0100, BBI) was injected intraperitoneally into mice at 100 µL per mouse. Anti-CXCL3 neutralizing antibodies (CAT #AF5568, R&D) were administered intraperitoneally at a concentration of 5 µg/mL in 100 µL per mouse. LPS (CAT#L4516, Sigma-aldrich) was administered intraperitoneally at 1 or 10 µg/mL, 100 µL per mouse every alternate day for 14 days. For PUE treatment, each mouse was injected intraperitoneally with 160 mg/kg PUE solution (CAT#H20033292, Fangming) during DSS treatment every day for 7 days. Lactobacillus rhamnosus GG was resuspended at a concentration of 10 × 10⁹ CFU/mL in sterile PBS and gavaged into the mice. Each mouse was gavaged with 200 µL every other day for 14 days. Anti-Ly6G antibody (CAT#16–9668-82, eBioscience; 1:2000) was selected to deplete neutrophil at a working concentration of 1 mg/mL, and each mouse was gavaged with 100 µL each time.

In the cellular assays, 10 ng/mL CXCL3 recombinant protein was used, and anti-CXCL3 neutralizing antibodies were used at a concentration of 200 ng/mL. LPS was used at a concentration of 1 or 10 ng/mL at 100 µL per well for 24h. For PUE treatment, cells were added at a concentration of 20 µM PUE per well for 24 h. For macrophage polarization, M1 macrophages were induced by 1 µg/mL LPS, M2 macrophages were induced by 20 ng/mL IL-4 (CAT#HY-P701093, MedChemExpress), and CXCL3-treated macrophages were stimulated by 10 ng/mL CXCL3 recombinant protein.

Murine ligature model

To model periodontitis, we placed 5–0 silk sutures on the maxillary second molars of mice for 3 weeks. Anesthesia was administered with isoflurane during ligation. Maxillary block biopsies were fixed in 10% formalin and examined for bone loss using microcomputed tomography.

Flow cytometry

To detect the presence of M1/M2 macrophages, the cells were stained with anti-CD86 (CAT#105007, biolegend) and anti-CD206 antibodies (CAT#17-2061-82, ebioscience) according to the manufacturer’s instructions. To detect CXCL3 expression, cells were incubated on ice in the dark with 5 g/mL brefeldin A 2 h prior to collection. The collected cells were washed three times, and then fixed, and the cell membrane was permeated using Fix&Perm Kit (CAT#FMS-FP0050, Fcmacs Biotech) according to the manufacturer’s instructions. Briefly, the cells were fixed for 15 min, washed 3 times, the cell membrane was permeated, followed by incubation with 1 µL of anti-rabbit GRO gamma monoclonal antibody (CAT#ab220431, Abcam) for 1 h. After washing and resuspension, the cells were exposed to goat anti-rabbit IgG (H&L) Alexa Fluor 647 (CAT#RS3811, ImmunoWay) for 30 min, washed, resuspended, and assayed. Prior to staining, cells were blocked using anti-CD16/32 antibodies (CAT#101302, biolegend). The data were analyzed using FlowJo software.

Blood cell composition analysis

To analyze the blood cell composition of mouse, we collect the blood from animal sampling using the anticoagulant tube, then test with 20 µL blood of each sample using the Automatic Animal Blood Cell Analyzer (Mindray; BC-2800Vet).

Hematoxylin and eosin, periodic acid-Schiff and Wright’s staining

The collected fresh samples were fixed in 4% paraformaldehyde for 48 h, embedded in paraffin wax after dehydration, and sliced into 5-µm sections. Hematoxylin and eosin staining was performed to examine colonic morphology. Periodic acid-Schiff staining (CAT#G1008, Servicebio) was performed according to the manufacturer’s instructions to observe the mucus layer and goblet cells. Wright’s staining (CAT#G1007, Servicebio) was performed to observe the neutrophil.

Immunofluorescence

Five-micrometer sections were obtained as the above section. An EDTA solution (pH 8.0) was used for antigen repair. The sections were incubated with anti-human CXCL3(CAT #PA5-103,136, Invitrogen; 1:2000), CD11b (CAT# GB115693, Servicebio; 1:1000), anti-mouse CXCL3 (CAT#YT2075, ImmunoWay; 1:2000), F4/80 (CAT#GB113373, Servicebio; 1:500), TLR4 (CAT#GB12186, Servicebio; 1:500), ELA (CAT#27,642–1-AP, Proteintech; 1:1000), and Ly6G (CAT#16–9668-82, eBioscience; 1:2000) antibodies overnight at 4°C. The sections were then incubated with a cyanine 3-conjugated goat anti-rabbit IgG (H&L) secondary antibody and HRP-conjugated goat anti-rabbit secondary antibody with FITC-tyramide (tsa) or CY3-Tyramide (tsa). Finally, the sections were stained for 10 min at room temperature in the dark using 4′,6-diamidino-2-phenylindole (DAPI) solution. Images were obtained using a Nikon Eclipse TI-SR fluorescence microscope.

Quantitative real-time polymerase chain reaction of gene expression

Colon tissue was ground on a multi-sample tissue grinder and total RNA was extracted from the colon tissues using an RNA Easy Fast Tissue Kit according to the manufacturer’s instructions (CAT#DP451, TianGen). Cellular mRNA extraction used Super FastPure Cell RNA Isolation Kit (CAT#RC102-01, Vazyme). The concentration and quality of RNA were measured using a Nanodrop One (Thermo Fisher Scientific, USA), and cDNA was reverse transcribed with 500 ng of RNA using a HiScript III RT SuperMix Kit (CAT# R323-01, Vazyme). Quantitative real-time PCR was performed using PowerUp SYBR Green Master Mix (CAT# A25776, Thermo Fisher Scientific) on a ViiA V7Dx qPCR-384 system (Thermo Fisher Scientific). Analysis was performed using the 2^−ΔΔCt method. The primer sequences are shown in the following table:

	Forward	Reverse
Human CXCL3	AACCGAAGTCATAGCCACACTCAAG	CAGTTGGTGCTCCCCTTGTTCAG
Human ACTB	TGACGTGGACATCCGCAA AG	CTGGAAGGTGGACAGCGAGG
Mouse Cxcl3	CACTGGTCCTGCTGCTGCTG	CGTCACCGTCAAGCTCTGGATG
Mouse Actb	GTGGGAATGGGTCAGAAGGA	CTTCTCCATGTCGTCCCAGT
Mouse Cxcl13	TTGTGATCTGGACCAAGATGAA	GACTTTTGCTTTGGACATGTCT
Mouse Rorγt	ACAAATTGAAGTGATCCCTTGC	GGAGTAGGCCACATTACACTG
Mouse Foxp3	TTTCACCTATGCCACCCTTATC	CATGCGAGTAAACCAATGGTAG
Mouse Tlr4	CTGTTCCTCCAGTCGGTCAG	CGTCGCAGGAGGGAAGTTAG
Mouse Light	TTGTGGTGGATGGACAGACG	CAGGAGAAACCAGCCCTGAG
Mouse Sphk1	CATGAGGTGGTGAATGGGCT	CCTGCTCGTACCCAGCATAG
Mouse Il10	TTCTTTCAAACAAAGGACCAGC	GCAACCCAAGTAACCCTTAAAG
Mouse Arg1	CATATCTGCCAAAGACATCGTG	GACATCAAAGCTCAGGTGAATC

ELISA analysis

Mouse serum remained at room temperature for half an hour, was centrifuged at 3500 rpm for 10 min, and the supernatant was collected. This procedure was repeated twice. The fecal supernatant was used as described in the section Fecal microbiota transplantation. Cell supernatant was centrifuged at 1000 rpm, and the resulting supernatant was collected; this process was repeated twice. The obtained samples were assayed for proteins using an ELISA kit (LPS ELISA kit, CAT#CSB-E13066M, Cusabio; CXCL3 ELISA kit, CAT#CSB-EL006249MO, Cusabio) according to the manufacturer’s protocol. If the determined value was less than the minimum detection value, it was set to 0.

Cell proliferation assay

To detect cell proliferation, we used the Cell Counting Kit-8 (CAT#CK04, Dojindo) according to the manufacturer’s instructions. In brief, Raw264.7 cells were seeded into 96-well plates at a density of 3 × 10³ cells per well. After adding the stimuli for the designated time, Cell Counting Kit-8 was added, and the cells were incubated for 3 h. The supernatant was aspirated, and the optical density value was measured at 450 nm.

Lentivirus transfection and treatment

The following sequence was obtained from the NCBI website for transcript NM_203320(CTCGAGCCaccatggcccctcccacctgccggctcctcagtgctgcactggtcctgctgctgctgctggccaccaaccaccaccaggctacaggggctgttgtggccagtgagctgcgctgtcagtgcctgaacaccctaccaagggttgattttgagaccatccagcttgacggtgacgcccaggaccccactgcacccagacagaagtcatagccactctcaaggatggtcaagaagtttgcctcaacccccaaggccccaggcttcagataatcatcaagaagatactgaagagcggcaagtccagctgaGAATTC). Lentiviral vectors (Plv6Ltr-ZsGreen-Puro-CMV-Cxcl3) were purchased from Corues Biotechnology Corporation (Nanjing, China). All lentiviral vectors contained green fluorescent protein (GFP) and a purinomycin-resistance gene (Puro). Raw264.7 cells were seeded into 6-well plates and prepared for transfection when they reached 80% confluence. Transfections were carried out overnight using medium without penicillin/streptomycin and serum, after which the medium was discarded, and medium containing 2 mg/mL puromycin was added for screening. The proportion of GFP-positive cells observed under an inverted fluorescence microscope was greater than 90%, indicating an appropriate infection rate. The optimal multiplicity of infection (MOI) was selected based on the qRT-PCR assay, fluorescence observation, and cell differentiation status, which was 20 MOI. Transfected cells were injected into mice via the tail vein, and the control group was injected with control lentivirus at 10 × 10⁶ cells/mouse.

16S rRNA gene sequencing

Bacterial genomic DNA was extracted from 100 mg of cecum content and 1 mL of saliva using the OMEGA Soil DNA Kit (CAT#M5635-02, Omega Bio-Tek). The V3–V4 region, with primers forward 5′-ACTCCTACGGGAGGCAGCA-3′ and reverse 5′-CGGACTACHVGGGTWTCTAAT-3′, was sequenced to construct the 16S rRNA gene library using the NovaSeq-PE250 platform. These sequences were analyzed using QIIME2 and R packages and mapped back to non-singleton ASVs identified based on Greengenes. The β-diversity was calculated using Bray–Curtis distance metrics and plotted using PCoA. LEfSe was used to identify biomarkers that were statistically different in abundance. All analyses were performed using R software, and the above process was completed at Personal Biotechnology Co., Ltd.

Metagenome DNA extraction, library construction and data analysis

Collecting 100 mg cecum contents from colitis mice gavaged with HSM and PSM, and total microbial genomic DNA was extracted using an OMEGA Mag-Bind Soil DNA Kit according to the manufacturer’s instructions. The quantity and quality of extracted DNAs were measured using a Qubit™ 4 Fluorometer and agarose gel electrophoresis. Metagenome shotgun sequencing libraries were constructed and sequenced on the Illumina NovaSeq platform using the PE150 strategy. Raw sequencing reads were processed to obtain quality-filtered reads for further analysis, which included removing sequencing adapters from sequencing reads using Cutadapt and trimming low-quality reads using a sliding-window algorithm in Fastp. Taxonomical classification of metagenomic sequencing reads was performed using Kraken2. The functionality of nonredundant genes was obtained by using annotated mmseqs2 against the KEGG, EggNOG, and CAZy databases. EggNOG and GO were obtained using an EggNOG mapper. GO analysis was obtained using map2slim (www.metacpan.org). Based on the taxonomic and functional profiles of nonredundant genes, LEfSe was used to detect differentially abundant taxa and their functions across groups. Analysis of β-diversity was performed to investigate the composition and function of microbial communities using Bray–Curtis distance metrics and visualized via PCoA. Metabolizing enzyme was analyzed using CAZymes analysis, and data were presented using LEfSe analysis. The above detection and analysis processes were performed by Personal Biotechnology Co., Ltd.

Transcriptomic profiling

We performed transcriptome analyses to study the genetic changes in the mouse colon, Cxcl3⁺ macrophages, and CXCL3-treated macrophages. The libraries were sequenced using an Illumina NovaSeq 6000 platform. Reads for each gene were generated using high-throughput sequence statistics and fragments per kilobase per million. DEGs were screened using DESeq with the following conditions: p < 0.05, > twofold change. They were analyzed using the R package for KEGG, GO term enrichment analysis, association analysis, and mapping. The sequencing data of DSS-treated mice were also analyzed and mapped using Metasacape. Sperman correlation analysis were using R software, and were mapped using cytoscape. The mRNA sequencing data of 43 healthy individuals and 42 patients with ulcerative colitis was collected from the GEO database (GSE22619, GSE37283, GSE38713, GSE47908).

Single-cell sequencing data download, preprocessing, and cell annotation

Single-cell sequencing data are from Single Cell Portal (broadinstitute.org) website (UC dataset: https://singlecell.broadinstitute.org/single_cell/study/SCP259/intra-and-inter-cellular-rewiring-of-the-human-colon-during-ulcerative-colitis; CD dataset: https://singlecell.broadinstitute.org/single_cell/study (/scp1423/predict-2021-paper-cd)). Data were normalized, scaled, and downscaled (PCoA and UMAP) according to website preprocessing. We set dim = 15 in the “FindNeighbors” function to construct images based on the standard deviation of the principal components and visualized them with the “ElbowPlot” function. Cell types and cell clusters were determined by cell population markers, mainly referring to the annotations on the data source website (Single Cell Portal). runUMAP was used for visualization to plot t-SNE (t-distributed stochastic neighbor embedding) subpopulations. Correlations were assessed using Pearson analysis, and data visualization was performed using R software.

Statistical analysis

Differences between two groups were assessed using the two-sided t-test or Wilcoxon’s test. For more than three groups, statistical analyses were performed using one-way analysis of variance (ANOVA) with Tukey’s multiple comparisons test. The α-diversity was performed with the Kruskal–Wallis test. To determine the effect of time points on different colitis groups, two-way ANOVA was used. Statistical analyses were performed using Prism 9 or R software. Data are shown as the mean ± standard error of the mean.

Authors’ contributions

Literature search and study design: J.Q., Q.G. and F.Y.; Experimentation and data collection: J.Q., Q.T., M.W., R.C., W. L. and Y. H.; Human sample collection: Q.T., Y.S., L.W. and H.S.; Data interpretation and analysis: J.Q., Q.L. N.W., Z.Y. and Y. L.; Figures and writing original draft: J.Q., Q.T., Y.S., R.G, and Z.Y.; Writing review & editing: J.Q., J. L., Q.G. and F.Y.; All authors have reviewed and approved the final version of this manuscript.

Funding

This work was supported by The National Natural Science Foundation of China (No. 82270979, 81970939), Jiangsu Provincial Medical Key Discipline (Laboratory) Cultivation Unit (JSDW202246), Nanjing Clinical Research Center for Oral Diseases (No. 2019060009), and High-Level Hospital Construction Project of Nanjing Stomatological Hospital, Affiliated Hospital of Medical School, Institute of Stomatology, Nanjing University (No. 0224C001, 0224C030).

Data availability

The datasets supporting the conclusions of this article have been deposited in the Sequence Read Archive (SRA) repository (PRJNA978509, PRJNA978338; https://www.ncbi.nlm.nih.gov/) and are publicly available as of the date of publication.

Declarations

Ethics approval and consent to participate

The research was approved by the Ethics Committee of Nanjing Stomatological Hospital, Medical School of Nanjing University (2019NL-008(KS)), and the Animal Ethics Committee of Nanjing Agriculture University, Nanjing, China (No. PZW2021020, PZW2022053). All donors gave informed consent.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.