κ-CGN alters colonic microbiota composition
After 90 days of κ-CGN oral gavage at different concentrations, intestinal microbiota community (based on 16S rRNA analysis) in mice had changed, with a significant increase of species richness (operational taxonomic unit [OTU], Chao1 and abundance-based coverage estimator [ACE] indices). Total bacterial load was increased in the high dose group (CGN-H) by approximately 13% compared with the control group (NC) (P < 0.01, Fig. S1a). Shannon and Simpson indices showed no significant differences in microbiota diversity after 90 days of treatment. The changes in the composition of the microbiota led to the changes in the enterotype. Mice in the NC group were clustered into the Prevotella (enterotype 2), an intestinal bacteria type are capable of generating short-chain fatty acids (SCFAs) [17]. Mice in the low and medium dose groups (CGN-L and CGN-M) were clustered into Ruminococcus (enterotype 3), with dominant bacteria effectively bind mucin and hydrolyze them into monosaccharides. Mice in the CGN-H group was clustered into Bacteroides (Enterotype 1), containing bacteria with a large amount of glycosidase capable of degrading polysaccharides (Fig. S1b) [18, 19].
We performed microbiota fecal transplants from NC and κ-CGN-treated mice to germ-free recipients (NCtrans+, CGNtrans+ groups, Fig. 1a). The CGN-L, CGN-M groups and the corresponding fecal transplantation groups had poor separation from the NC mice in Principal coordinates analysis (PCoA) clustering. However, significant separation of microbial composition was found between the CGN-H and NC groups, and the CGN-Htrans+ and NCtrans+ groups. In addition, CGN-H and CGN-Htrans+ mice clustered closely together, indicating high similarity in fecal bacterial composition between the κ-CGN groups and their corresponding transplantation recipients (Fig. 1b, c). Univariate analysis at the phylum level showed that compared with the NC/NCtrans+ mice, CGN-H/ CGN-Htrans+ mice had a significantly increased abundance of Bacteroidetes (54.0% vs. 60.3%; 52.3% vs. 58.7%, P < 0.05), and a significantly reduced abundance of Proteobacteria (4.6% vs. 2.1%; 4.2% vs. 1.5%, P < 0.01); Firmicutes was decreased in CGN-H/ CGN-Htrans+ mice, but with no significant (37.3% vs. 34.1%, 38.1% vs. 36.7%; P > 0.05;Fig. 1c). Rhodospirillaceae was significantly more abundant in CGN-M/CGN-H groups compared with the NC group (CGN-H: 6.4-fold increase, P < 0.05; Table S1). At the genus level, the results of fecal bacteria transplantation were consistent with the changes caused by gavaging κ-CGN. Compared with the NC/NCtrans+ groups, the CGN/CGNtrans+ treatments had significantly increased relative abundance of the bacteria Ruminococcaceae_unclassified and the intestinal mucus-degrading Bacteroides [20] (all increased more than 3-fold; P < 0.05). The relative abundance of the intestinal mucosal resident bacteria Akkermansia, which inhibits the contact of pathogenic bacteria with the host mucosa [16], was significantly reduced (both reduced by greater than 70%, P < 0.05). In addition, the relative abundance of the intestinal mucosal bacterial genera Anaeroplasma and Mucispirillum, which stimulate host-production of IgG and IgA; the cellulose-degrading bacteria and short-chain fatty acid (SCFA)-producing bacterial genera [20–22], [Ruminococcus]_torques_group, Ruminiclostridium_5, Lachnospiraceae, and [Eubacterium]_brachy_group were also significantly reduced (P < 0.05, Fig. 1d, Tables S1 and S2).
κ-CGN aggravates the inflammatory outbreak of pathogenic bacteria by influencing colonic microbiota
κ-CGN did not induce overt colitis in conventional C57BL/6 or germ-free mice, with no significant increase in level of lipocalin-2 (Lcn-2, a sensitive marker of intestinal inflammation in mice) (P > 0.05; Fig. 2a, b), CGN-Htrans+mice did not exhibit changes in fecal Lcn-2 levels, either (P > 0.05). Administration of C. rodentium led to significantly increased levels of Lcn-2 compared to NC group, with a 2.3-fold increase (P < 0.01). Interestingly, in mice treated with CGN-H, followed by C. rodentium (CGNC. rodentium), fecal Lcn-2 was significantly increased (~ 1.8 fold) relative to mice administered C. rodentium alone (P < 0.01).
Similar to Lcn-2, after 90 days of κ-CGN treatment, six pro-inflammatory cytokines were unchanged compared to NC group (P > 0.05, Fig. 2c). C. rodentium treatment significantly increased the expression of TNF-α and interleukin-6 (IL-6), by 2.8 and 2.4-fold, respectively (P < 0.01) relative to the NC group. Importantly, TNF-α and IL-6 were significantly elevated in CGNC. rodentium compared with the C. rodentium-only treatment group by 3.6 and 1.4-fold, respectively (P < 0.01), as well as the chemokine MCP-1, which was increased by 1.6-fold (P < 0.01, Fig. 2c).
H&E staining of the colon tissues showed that the intestinal tissue structure of mice in the κ-CGN, CGN-Htrans+, and CGN-Hfree+ (CGN-H treated germ-free mice) groups was normal, with tightly arranged goblet cells and no inflammation and infiltration in the crypts (Fig. 2d). The CGN/CGNtrans+ treatment groups had no sign of hyperemia, edema or ulceration (Fig. 2d, Fig. S2). C. rodentium induced colon tissue damage in conventional mice, which was further aggravated in CGNC. rodentium mice; both groups had submucosal edema in colon tissues, and a significantly increase of inflammatory cells in the basal layer. The colonic mucosa of CGNC. rodentium mice was clearly hyperemic, with edema and hemorrhagic ulcers (Fig. 2d, Fig. S2), demonstrating a state of heightened mucosal inflammation in CGNC. rodentium mice.
κ-CGN alters microbiota composition, leading to changes in mucus degradation genes and SCFA
Metagenomics analysis was consistent with the results from 16S rRNA analysis, Chao1 and ACE indices increased significantly after 90 days with CGN-H treatment (P < 0.01, Fig. S3a), while Shannon and Simpson indices did not. In addition, the relative abundances of SCFA-producing genera, including Ruminiclostridium_5, Lachnospiraceae and Eubacterium brachy_group [23] were reduced in CGN-H-treated mice. The relative abundance of Bacteroides was increased in CGN-H-treated mice (Fig. S3b, Table S3), and excessive proliferation of Bacteroides may lead to destruction of polysaccharides in the intestinal mucus layer [24]. We therefore sought to determine the abundance of carbohydrate utilization genes in relation to intestinal mucus composition. A total of 28,687 genes encoding carbohydrate active enzymes (CAZymes) were identified in mouse fecal samples, with the largest number of genes belonging to the family of glycoside hydrolases (Fig. 3a). CGN-H treatment led to a significant increase in genes encoding mucosal polysaccharide binding proteins and mucin degrading enzymes. For example, CBM32 and CBM40, which encode mucosal glycan binding proteins, were increased 12.5 and 2.1-fold, respectively, compared with the NC group (P < 0.05); the genes encoding N-acetyl galactosidase, CE9 and CE11 were increased by 9.8 and 36.1-fold, respectively (P < 0.01); genes encoding mucosal polysaccharide and glycosyltransferase, GH106 and GT4, were increased 29.5 and 26.7-fold, respectively (P < 0.01, Fig. 3b). By performing correlation analysis between CAzymes genes changes and microbial abundance (Fig. 3c, Table S4), we found that the abundance of Bacteroides ovatus (Bacteroides genus) containing CBM32, CE11, GH106, GH109, GH84, and GH85 genes increased 25.35-fold (P < 0.01). The abundances of Bacteroides nordii and Bacteroides thetaiotaomicron containing GH33 were 17.89 and 21.35 folds higher, respectively than that of the NC group (P < 0.01). The abundance of Bacteroides intestinihominis containing GT4 gene was increased 3.45-fold (P < 0.05).
As shown in Fig. 3b, the abundance of genes involved in the use of cellulose, starch and mannan were significantly reduced in CGN-H treated mice. For example, compared with the NC group, GH5 and GH151 encoding β-1, 4-xylan hydrolase were reduced 68.3 and 51.6-fold, respectively (P < 0.01); GH36 and GH77 encoding α-glucosidase were reduced 29.3 and 28.4-fold (P < 0.01), respectively; GH30 encoding β-mannosidase was decreased by 23.2-fold (P < 0.01). In Fig. 3c, correlation analysis showed that the abundances of Lachnospiraceae_ bacterium_10 and Ruminococcus torques containing cellulose degrading genes (GH5, GH42, GH15 and GH151) and starch and mannan degrading genes (GH36, GH77 and GH30) were reduced 13.25 and 6.75-fold, respectively after CGN-H treatment (P < 0.01).
These bacterial can produce large amounts of SCFAs [19, 23]. We therefore quantified SCFAs and, as expected, fecal SCFA contents were significantly altered by κ-CGN treatment. OPLS-DA demonstrated clear separation between CGN-M/CGN-H and NC group, but with limited separation between CGN-L and NC groups (Fig. S4). Among several typical SCFAs, butyric acid, isobutyric acid, valeric acid, and isovaleric acid were all reduced in CGN-M and CGN-H groups, especially butyric and valeric acid, which were significantly reduced by 67.4% and 60.6%, respectively in the CGN-H group (P < 0.01). Similar changes in SCFA contents were seen in mice receiving fecal transplants from CGN-H mice. Compared with the NCtrans+ group, the CGN-Htrans+ group had a 65.3% reduction in butyric and a 69.7% reduction in valeric acid (P < 0.01). In germ-free mice, SCFA contents were very low in general, even after κ-CGN transplantation (Fig. 3d), which indicates that the change in SCFA contents in feces of CGN-treated mice was related to changes in microbiota composition.
Correlation analysis between SCFA abundance and intestinal microbiota (Fig. 3e, Table S1) showed that bacteria affected by κ-CGN treatment, including cellulose degrading bacteria, and the SCFA-producing bacterial genera Ruminiclostridium_5, Lachnospiraceae, and [Eubacterium]_brachy_group [23] were positively correlated with changes in isovaleric acid, caproic acid, and acetic acid (P < 0.05). In addition, positive correlations were also found between Ruminiclostridium_5 and isobutyric acid and butyric acid; and between Lachnospiraceae and butyric acid and valeric acid (P < 0.05); whereas a negative correlation was found between Ruminococcaceae and valeric acid and butyric acid (P < 0.05).
Shifts in microbial composition contribute to intestinal barrier dysfunction
Fluorescent in situ hybridization (FISH) examination revealed both κ-CGN treatment of conventional mice and the corresponding fecal bacteria transplantation caused thinning of the intestinal mucus layer, by reducing 61% and 58%, respectively (Fig. 4a, b; P < 0.01), but not correlated with the expression of the major intestinal mucin Muc2, Krüppel-like factor 4 (Klf4) or goblet cell protein (Tiff3). The expression of these three proteins were not altered by κ-CGN treatment (P > 0.05, Fig. 4c). However, after 90 d of κ-CGN treatment in germ-free mice, the thickness of the mucus layer was almost unchanged. No bacteria were observed in the intestinal mucus layer within 25 µm from the epithelial cells of NC group (Fig. 4d, e). In contrast, bacteria penetrated the mucus layer more frequently in mucosal biopsies obtained from κ-CGN-H treated and fecal transplant groups, compared with NC group, with the average distance reduced by greater than 87% in the CGN-H group and 83% in CGN-Htrans+ group (P < 0.01). Such microbiota encroachment correlated with reduced mucus thickness. A classical permeability marker FITC-dextran test revealed a significant increase in mucus penetrability in CGN-M, CGN-H and CGNtrans+ groups with higher FITC-dextran levels than in NC mice (P < 0.05, Fig. 4f). To determine whether the observed defective mucus layer was caused directly by κ-CGN treatment, we next studied mucus properties in germ-free mice. Compared with NCfree+ group (germ-free control mice), no significant difference in penetrability was detected in CGN-Hfree+ mice: mucus thickness, penetrability and expression of Muc2, Klf4, Tiff3 were not affected (Fig. 4a, b, f).
Probiotics attenuate pathogen-induced inflammatory mucosal damage
We supplemented mice with two probiotics Bifidobacterium longum NCC-2705 and Faecalibacterium prausnitzii after 90 d of oral gavage of κ-CGN (CGNPro+, Fig. 5a). PCoA analysis of 16S rRNA sequencing at genus level showed two distinct clusters between NC or CGN-H mice; after 30 d of subsequent probiotic intervention, the bacterial composition of CGN-Hpro+ group was closer to that of the NC group (Fig. 5b). Univariate analysis at phylum level showed decreased relative abundance of Bacteroides in CGN-Hpro+ mice compared with CGN-H mice (45.1% vs. 55.7%, P < 0.01), and increased relative abundances of Firmicutes (50.4% vs. 43.0%, P < 0.01) and Proteobacteria (3.2% vs 1.1%, P < 0.01; Fig. 5c). At genus level, the increases in relative abundance of CGN-H-induced proinflammatory bacterial genera Comamonas [25], Alistipes, Escherichia-Shigella [26] and Clostridium were ameliorated by probiotic intervention as was the CGN-H-induced decrease in the relative abundance of Akkermansia. While the abundance of Bifidobacterium and Faecalibacterium, two gavaged probiotics, increased significantly (P < 0.01, Fig. 5d, Table S5). Anaeroplasma and Mucispirillum bacteria, and the genera Ruminiclostridium_5, Lachnospiraceae, and [Eubacterium]_brachy_group were significantly increased in CGN-Hpro+ vs. CGN-H mice (P < 0.05). In addition, the polysaccharide-degrading bacteria Enterorhabdus were significantly reduced in CGN-Hpro+ vs. CGN-H mice (P < 0.05, Fig. 5d, Table S5).
Probiotic intervention significantly increased butyric acid, isobutyric acid, and valeric acid (P < 0.05, Fig. 5e), which were originally reduced by CGN-H. Spearman’s correlation analysis the different SCFAs between CGN-Hpro+ and CGN-H groups and the intestinal microbiota after probiotic intervention showed that several SCFAs, including butyric acid and valeric acid were significantly and positively correlated with Lachnospiraceae_bacterium_615 and Lachnospiraceae_bacterium_DW8/17/22 and with F. prausnitzii (P < 0.05, Fig. 5f). In addition, Mucispirillum_schaedleri was positively correlated with changes in caproic acid (P < 0.05); and Alistipes_sp._CHKC1003, Alistipes_finegoldii, Clostridium_sp._Culture-27, Enterorhabdus_mucosicola and Desulfovibrio_sp._ABHU2SB were negatively correlated with changes in butyric acid, isobutyric acid, acetic acid, and propionic acid (P < 0.05).
In addition, the phenomenon of thinning of the mucus layer and increased intestinal permeability caused by exposure to CGN-H was reversed after probiotic intervention. The thickness of the mucus layer recovered by 55.6% (P < 0.01, Fig. 5g, h), and intestinal permeability was dampened by probiotics, approaching levels similar to that of the NC group (P < 0.05, Fig. 5i). Colonic tissue and inflammatory factors were not significantly changed by probiotic intervention (Fig. S5).
Then, C. rodentium was administered by oral gavage after 90 d of CGN-H or after 90 d of CGN-H plus 30 d of probiotic intervention (CGNC. rodentium+Pro+). C. rodentium treatment significantly reduced the content of SCFAs in feces of conventional mice, except for caproic acid (P < 0.05, Fig. 6a); compared with the C. rodentium group, SCFA contents were further reduced in the CGNC. rodentium group (P < 0.05, Fig. 6a). Probiotic intervention significantly reversed this reduction in intestinal SCFAs (P < 0.05), with the content of propionic acid and butyric acid higher than that of the C. rodentium group (P < 0.05). Even acetic acid, isobutyric acid, and isovaleric acid were basically restored to normal levels. H&E histochemical analysis produced similar findings: Probiotic intervention significantly reduced intestinal inflammation, eliminated submucosal edema, and significantly reduced the infiltration of inflammatory cells in the basal layer (P < 0.05, Fig. 6b, Fig. S6). Similarly, probiotic intervention significantly improved C. rodentium-induced increase in inflammatory cytokines, which was further promoted by CGN-H. Compared with the CGNC. rodentium group, TNF-α, IL-6 and MCP-1 were reduced by 26%, 20%, and 18%, respectively in the CGNC. rodentium+pro+ group (P < 0.05, Fig. 6c), suggesting the ability of probiotics to dampen the pro-inflammatory effect of C. rodentium.