Members of Enterobacteriaceae are enriched in the mesenteric adipocytes tissue (MAT) of CD patients compared to the non-CD controls
We performed linear discriminant analysis effect size (LEfSe) analysis (LDA ≥ 4) to identify the differentially abundant taxa in the MAT of CD at multiple levels. We observed that Enterobacteriales, Enterobacteriaceae (a family in Enterobacteriales), and uncultured_f_Enterobacteriaceae were consistently significantly enriched in the mesenteric tissue of patients with CD (Fig. 1a), indicating that they may be closely associated with CD pathogenesis. To evaluate the important role of Enterobacteriaceae in CD, we classified our recruited patients with CD into high Enterobacteriaceae (high E) and low Enterobacteriaceae (low-E) groups according to the median value of the abundance of Enterobacteriaceae in MAT. We observed that the bacterial community in MATs with a high E exhibited distinct microbiotic characteristics compared to those with a lower abundance of Enterobacteriaceae (Fig. 1b). Furthermore, the differentially distributed microbial community was characterized between patients with high and low E, which showed that an increase in Streptococcus, Burkholderia, Pseudomonas, Acinetobacter, Geobaccillus, and Thermus was associated with Enterobacteriaceae accumulation in CD-MAT (Fig. 1c, Figure S1). Additionally, we found that patients with a higher Enterobacteriaceae abundance had a higher bacterial diversity (Fig. 1d), indicating more extensive bacterial translocation in these patients. Moreover, an increased proportion of CD patients who experienced recurrence waas observed in the high-E group (Fig. 1e). Additionally, to explore Enterobacteriaceae’s ability to distinguish CD from non-CD, receiver operator characteristic (ROC) analysis was used to calculate the area under curve (AUC). The results showed an AUC value of 0.6963 for Enterobacteriaceae, while the uncultured_f_Enterobacteriaceae showed an AUC value of 0.7373, indicating that specific members of the family Enterobacteriaceae contributed to the bacterial configuration in CD-MAT (Fig. 1f). Overall, these data demonstrated that mesenteric tissues of patients with CD harbored a significantly differentially distributed microbiota, and enrichment of specific members of Enterobacteriaceae was a determinant characteristic in CD-MAT.
Viable Klebsiella variicola in Enterobacteriaceae is exclusively isolated from the mesenteric tissue of CD patients
To investigate the viable translocated members of Enterobacteriaceae in the mesenteric tissues, we cultured 23 specimens (14 CD and 9 non-CD) under anaerobic and aerobic conditions and picked 229 colonies (174 from CD and 55 from control) with different colony appearances on different culture plates. 16S rRNA gene sequencing revealed that these colonies were classified into 17 families (Fig. 2a), with 13 families identified in the mesenteric tissue of CD and seven in the non-CD control (Fig. 2b). Totally, 32 colonies belonging to Enterobacteriaceae were identified from the MAT of patients with CD, which were categorized into four taxa groups (Escherichia fergusonii, Klebsiella variicola, Shigella flexneri and Klebsiella pneumoniae) (Fig. 2c). As shown in Fig. 2D, Escherichia/Shigella were repeatedly isolated from the mesenteric tissue of CD and the control group, whereas Klebsiella spp. were exclusively isolated from the mesenteric tissue of CD (Fig. 2d). Observation of Klebsiella spp. in CD-MAT was verified by fluorescence in situ hybridization (FISH) using Klebsiella-specific probes under sterile conditions (Fig. 2e).
The colitogenic roles of K. pneumoniae and K. oxytoca in the initiation and perpetuation of intestinal colitis have been revealed [25–27]. To clarify the phylogenetic identity of our recovered Klebsiella spp., we performed whole-genome sequencing. Phylogenetic analysis, based on a set of conserved functional genes, classified this isolated bacterium as K. variicola (Fig. 2f). Functional analysis of the genomic genes in K. variicola revealed 58 genes for cell motility and 135 genes for lipid transport and metabolism (Figure S2a). Additionally, genes involved in protecting against oxidative damage (superoxide dismutase, nitrate reductase, thioredoxin reductase, and peroxiredoxin) and promoting bacterial survival and proliferation in the host tissue (arginase) were also observed in K. variicola’s genome (Figure S2b), suggesting that K. variicola could be available for survival in MAT. To show K. variicola’s ability to invade adipose tissue, we carried out a bacteria-cell co-culture assay in vitro. As expected, K. variicola could successfully invade into 3T3L1 cell (mice preadipocytes cell line) (Fig. 2g). The pro-inflammatory role of K. variicola was evaluated in vitro, and we observed that K. variicola could evoke a substantially stronger pro-inflammatory response under the co-culture with intestinal epithelial cells (IEC-6) and preadipocytes cells (3T3L1), compared to E. fergusonii (our previously isolated commensal bacteria in Enterobacteriaceae) [10] and PBS treated groups (Fig. 2h & i). Taken together, these results confirm the existence of a specific viable translocated Enterobacteriaceae member, K. variicola, in the mesenteric tissue. Notably, K. variicola can invade preadipocytes and act as an inflammation inducer in epithelial and preadipocyte cells.
K. variicola exacerbates colitis in mice
To further evaluate the pathogenic and pro-inflammatory ability of K. variicola in vivo (Fig. 3a). Mice colonized with E. fergusonii or PBS were used as controls. Expectedly, K. variicola-treated mice had exacerbated DSS-induced colitis, as indicated by a sharp weight loss (Fig. 3b), heightened disease activity index (Fig. 3c), and shortened colon (Fig. 3d & e). Colonic inflammation was verified by histological assessment, as shown by increased mucosal erosion, crypt destruction, and inflammatory cell infiltration in K. variicola-treated group in comparison with mice treated with E. fergusonii and PBS (Fig. 3f & g). In accordance with the histological alteration, we observed a significant increase in colonic mRNA expression of TNF-α, IL-1β, and IL-6 in mice colonized with K. variicola (Fig. 3h). Meanwhile, translocation of K. variicola in mouse mesenteric tissue was examined, and K. variicola was positively detected in 5/6 of the mice, with a range of 218–1484 CFU/mg tissue (Fig. 3i).
Unlike wild-type C57BL/6 mice, the interleukin-10 deficient (IL-10−/−) mice spontaneously developed chronic enterocolitis similar with chronic IBD in humans, when exposed to normal commensal bacteria [28]. Thus, to better understand the inflammation-promoting role of K. variicola, an IL-10-deficient mouse model was used in this study (Fig. 3j). Consistently, a significant colitis-promoting role of K. variicola was identified, as evidenced by a significant shortening of colon length (Fig. 3k & l) and a heightened disease activity index (Fig. 3m). Meanwhile, histological features of inflammation characterized by mucosal thickening and aggravation of colonic inflammation were also observed in K. variicola treated mice, but not in those treated with commensal bacteria E. fergusonii or PBS (Fig. 3n & o). Collectively, our results showed that mesenteric-derived K. variicola exacerbated colitis in mice.
K. variicola impairs the intestinal barrier function
Intestinal epithelial barrier dysfunction has been demonstrated to be involved in the initial pathogenesis of CD [29]. Given that K. variicola could translocate from the lumen into MAT, we intended to determine whether K. variicola plays a role in the disruption of the epithelial barrier. For this, we gavaged K. variicola (109 CFU/mouse) daily into SPF mice treated with a 5-day antibiotic cocktail for seven consecutive days. Mice were given drinking water instead of 3% DSS to exclude damage from medicinal chemicals (Fig. 4a). The integrity of the intestinal epithelial barrier was investigated by examining ZO-1, Occludin, Claudin1, and Synaptopodin (SYNPO) expression levels. As shown in Fig. 4b, the mRNA expression of intestinal ZO-1 was significantly suppressed in K. variicola-treated mice compared to those treated with E. fergusonii or PBS, but not for the expression of Occludin and Claudin1 and SYNPO. Immunofluorescence staining confirmed the downregulation of ZO-1 in mouse colonic tissues (Fig. 4c&d). Electron microscopy (EM) confirmed the impairment of tight junctions in K. variicola-treated mice (Fig. 4e). Consistent results were observed in IL-10−/−mice. ZO-1 expression was significantly reduced under the treatment of K. variicola, as demonstrated by RT-qPCR and immunofluorescence staining (Fig. 4f-h). Taken together, our results showed that K. variicola disrupted gut barrier integrity by inhibiting ZO-1 expression.
Active T6SS in K.variicola influences the expression of ZO-1
Bacterial pathogens use various mechanisms to invade mammalian hosts, and one important strategy of gram-negative bacteria is the release of virulence effectors through the secretion system to reach the target [30]. T6SS has been demonstrated to be involved in disrupting the actin cytoskeleton [31]. Moreover, stabilization of the tight junction solute barrier by ZO-1 via coupling to the perijunctional cytoskeleton has been reported [32]. Therefore, we reasoned that the disruption of the intestinal barrier by K. variicola may be mediated by T6SS. To test this, we first blasted the bacterial genome, which identified the existence of core components of T6SS in the genome of K. variicola (Fig. 5a). Additionally, a contact-dependent neighbor-killing assay was designed to assess T6SS activity in K. variicola. The results showed that co-incubation of E. coli LF-82 with K. variicola led to a remarkable reduction in the viability of E. coli LF-82, indicating that active T6SS but not pseudogenes existed in K. variicola (Fig. 5b & c).
To investigate the active role of T6SS in inducing intestinal permeability in K. variicola, we used a CRISPR interference (CRISPRi) system to knockdown ClpV, a core gene in T6SS (Fig. 5d). ClpV is an ATPase associated with various cellular activities that functionally provide energy for sheath disassembly in T6SS. We found that ClpV inhibition led to a mild growth rate deficiency, an attenuated extent of contact inhibition, and a reduced inhibition of ZO-1 expression K. variicola-dcas-ClpV in comparison with K. variicola-vector (Fig. 5e-h). Besides, the in vitro pro-inflammatory effect of K. variicola was also alleviated in K. variicola-dcas-ClpV-treated IEC-6 cells but not in 3T3L1 cells (Fig. 5i-j), indicating that T6SS of K. variicola mainly targets the gut rather than MAT. This is consistent with our previous observation that K. variicola can impair the gut barrier function. Collectively, these results show that active T6SS existed in K. variicola, which mediated the pro-inflammatory and barrier disruption role of K. variicola.
Dysfunction Of T6ss Alleviates Colitis In Mice
To address whether the dysfunction of T6SS in K. variicola influences its colitis-promoting role, we carried out an in vivo mouse experiment (Fig. 6a). A consistent down-regulated expression of ClpV in K. variicola was confirmed in mice (Figure S3a). Dysfunction of T6SS did not affect the colonization of K. variicola in mice (Figure S3b), but attenuated the colitis-promoting effect of K. variicola. Mice colonized with K. variicola-dcas-ClpV showed milder colitis, as indicated by a lower weight loss rate (Fig. 6b), less intensive DAI score (Fig. 6c), reduced colon shortening (Fig. 6d & e). Histological assessment of colonic inflammation was illustrated by decreased mucosal erosion, crypt destruction, and inflammatory cell infiltration in mice treated with K. variicola-dcas-ClpV compared to control group mice (Fig. 6f & g). Additionally, a significantly decreased expression of intestinal inflammatory cytokines (TNF-α, IL-1β, and IL-6) (Fig. 6h) and K. variicola load in mice-MAT was observed in mice treated with K. variicola-dcas-ClpV, compared to K. variicola-vector treated controls (Fig. 6i). Moreover, in line with the observations from the DSS-induced mouse model, reduced colon inflammation was observed in K. variicola-dcas-ClpV infected IL-10−/− mice compared to K. variicola-vector controls (Fig. 6j-n). Meanwhile, dysfunction of T6SS in K. variicola also alleviated its inhibitory effect on ZO-1 expression (Fig. 6o & q). Taken together, these observations demonstrated that T6SS in K. variicola plays a vital role in disrupting intestinal permeability and triggering colitis in mice.
K. variicola and ClpV are enriched in the gut microbiome of patients with CD
The results above have confirmed that K. variicola is a pathogenic bacterium existed in CD-MAT. Since microbiota in MAT has been supposed to be translocated from gut lumen [9], we next explored whether K. variicola and ClpV were enriched in the gut microbiome of patients with CD and analyzed the fecal metagenomic sequencing data in a publicly available database (PRJEB15371) [33]. Here, we showed that a significant enrichment of Enterobacteriaceae, Klebsiella and especially K. variicola existed in the fecal samples of patients with CD (Fig. 7a-c), which confirmed that K. variicola was closely related to the occurrence of CD, and provide a reasonable explanation for the enrichment of K. variicola in CD-MAT. Taxonomy distribution of experimentally validated T6SSs revealed that Gammaproterobacteria accounted for the largest proportion (up to 76%) (SecReT6, T6SS database: https://bioinfo-mml.sjtu.edu.cn/SecReT6/index.php). This class of bacteria includes several medically important bacteria which have been identified to have an active role in the pathogenesis of IBD, and members of the Enterobacteriaceae family are predominant [34]. Consistent with these data, we also observed a significantly increased abundance of Gammaproterobacteria-ClpV in patients with CD (Fig. 7d). These data suggested that high colonization of K. variicola in CD-MAT was associated with the bacterial translocation from gut lumen and that T6SS is closely related to the pathogenesis of CD.