Although intestinal MAMs have been observed to be involved in the pathophysiology of UC, many of the previous studies focusing on MAM used mucosal biopsy in adults as well as pediatric patients [30, 31, 32]. In this study, we employed CLFs and paired fecal samples to determine specific alterations in MAM according to mucosal inflammation or disease activity. We collected CLFs using colonoscopy via a sterile tube and carried out high-throughput 16S rRNA gene metagenomic analysis. In order to evaluate the microbiota residing proximal to the mucous layer, we aspirated the CLFs immediately after vigorously washing the mucosal surfaces using the waterjet function of the endoscope. This method has advantages for safe sampling from a wide mucosal area along the colon because biopsy at many sites is invasive, especially for children. In addition, the lavage contains a sufficient number of microbial cells, which minimizes the effect of environmental contamination on PCR-based analysis and allowing more efficient isolation of the microbes of interest. To our knowledge, this is the first report to conduct a comparative analysis of microbiota between CLFs and paired samples of feces from pediatric UC patients.
Consistent with other reports [33, 34, 35], substantial inter-individual differences in gut microbiome was observed among the patients (Fig. 1). However, these differences accounted for less variation in the UC than in the non-IBD microbiomes: 56.8% vs 83.8%, respectively. In addition, compared with UC, the microbiome variations in non-IBD correlated more with sampling sites (A, T, S, R or F) and sample types (CLF or feces). Correspondingly, the α-diversity (number of observed OTU and/or Chao1 index) showed significant differences between the CLFs and feces in non-IBD while no such difference was observed in UC (Fig. 2). The microbiome composition (β-diversity) of MAM in UC was homogenous along the colon and similar to that in feces. However, in the cases of non-IBD, the microbiomes in MAM tended to be different from those in feces (Fig. 3). These results suggest that compartmentalized distribution of microbiome along the mucosa-luminal axis is disrupted in UC due to the compromised mucous layer function even in the remission stage. Of note, compared with mucosal inflammation (Matts score), the disease severity (PUCAI) had a stronger association with alteration in MAM. This might indicate that gut physiological dysfunction (diarrhea, bleeding or fast transit time etc.) has a larger impact on the gut microbiome.
Among the blood analysis data, fibrinogen level was associated with nearly half of the variation in the MAM (Fig. 1). Shen et al. has reported that UC patients showed elevated fibrinogen levels, which they therefore proposed as a diagnostic marker for UC [36]. However, we observed a similar association in non-IBD patients, indicating that the fibrinogen is not a UC-specific marker but reflects the nonspecific inflammatory response. Notably, the serum levels of all classes of immunoglobulins (IgA, IgG and IgM) were significantly associated with the variation in the MAM, and the degree of correlation was more evident in UC than in non-IBD. MaAsLin analysis showed that abundances of Enterococcaceae and Lactobacillaceae were strongly associated with these antibody levels, indicating that UC patients developed humoral immune responses to these bacterial groups (Fig. 2). These findings were consistent with the report by Bourgonje et al., in which Streptococcus, Lactobacillus, Lactococcus, Enterococcus, Veillonella and Enterobacteriaceae were enriched among IgG-coated bacteria in adult UC [37]. In UC, levels of the autoantibodies pANCA and cANCA correlated with the relative abundance of Staphylococcus and Enterobacteriaceae (Fig. 2). This association was probably due to the strong inflammatory response evoked by these bacteria. Luo et al. reported that systemic translocation of Staphylococcus was associated with germinal center B cell activation and production of autoantibodies to nuclear antigen (ANA) and double strand DNA [38]. They also showed a higher plasma LPS level in HIV-positive patients with high ANA than with low ANA levels. It was also reported that E. coli-derived caseinolytic protease B induces autoantibody formation [39]. Overall, the levels of serum inflammatory markers and levels of immunoglobulin were more correlated with the MAM than fecal microbiota (Fig. 1), suggesting that the alteration in MAM reflects the host humoral immune response to gut microbiota.
Several studies have reported that the gut microbial diversity and composition differ between feces and mucosal biopsy in healthy adults [17, 18] as well as in adult UC patients [19]. However, to our knowledge, no reports have been published that compared the microbiota in CLFs and the paired feces from pediatric UC patients. Inconsistent with these reports in adult cases, our results showed that MAMs are homogenously distributed along the colon, and no difference in α- and β-diversities among the sites was observed in the pediatric UC as well as the non-IBD patients. Notably, the microbial composition between CLFs and feces was similar, particularly in the pediatric UC patients. These findings (homogenous distribution of intestinal MAM) might account for the higher prevalence of severe pancolitis in pediatric UC than in adult cases [40, 41]. Age-dependent alteration in MAM along the intestine should be investigated to address the precise role of the MAM on the pathophysiology in UC.
Unexpectedly, CLFs in non-IBD pediatric patients showed higher microbial diversity compared with feces, and the bacterial families usually residing in oropharynx or upper intestines such as Enterobacteriaceae, Pasteurellaceae, Fusobacteriaceae and Neisseriaceae were significantly more abundant in the MAM. On the other hand, Bifidobacteriaceae and Lachnospiraceae, representative luminal members of gut microbiota, were more abundant in feces (Figs. 3 and 4). Microbiota of oral origin are known to increase in the intestines of IBD patients [42]. Schirmer et al. reported that Veillonella dispar, Veillonella parvula, Aggregatibacter segnis, Haemophilus parainfluenzae, Campylobacter sp., Lachnospiraceae, Megasphaera sp. increased in the intestine of treatment-naïve pediatric UC patients with the disease course [43]. Somineni et al. reported that an increase in oral bacteria before treatment was associated with aggravation of UC [44]. Moreover, Atarashi et al. reported that oral microbiota were significantly more abundant in the fecal samples from UC patients compared with healthy controls. In addition, they showed that Klebsiella sp. isolated from the salivary microbiota are strong inducers of Th1 cells when they colonize in the gut [45].
Originally, normal intestinal microbiota resist colonization of foreign microorganisms, which suppresses the colonization and proliferation of pathogenic bacteria. It has been reported that the intestinal tracts of IBD patients are more likely to be colonized by oral microbiota, resulting from attenuated colonization resistance [46, 47]. In addition, an increase in oral bacteria migrating to the intestinal tract has been observed in those patients. In IBD patients, genetic factors, treatments and intestinal inflammation may increase the abundance of swallowed oral bacteria that reach the lower intestine [48, 49, 50]. However, our data showed that the relative abundances of these bacteria in CLF from non-IBD were higher than those from feces (Fig. 5), indicating that they normally reside proximity to the mucous layer due to their adhesive properties and/or relative tolerance to reactive oxygen species. On the other hand, in MAM from the pediatric UC patients, we detected only enrichment in Enterobacteriaceae. These results indicate that the compromised mucous layer in UC had lost the ability to separate MAM and luminal microbiota, and also that Enterobacteriaceae is more capable of surviving in inflamed mucosa.
In agreement with other reports [51], mucosal inflammation or disease activity correlated with the alteration in MAM (Fig. 6). As Matts score (mucosal inflammation) or PUCAI (disease activity) increased, the Bray-Curtis dissimilarity in MAM among the sites (A, T, S and R) became larger, suggesting that the UC disease severity promotes the intra-individual regional variations in MAM by selective pressures (e.g. oxidative damage or clearance by immune system). We detected the differentially enriched bacterial genera at each disease condition determined by Matts score or PUCAI (Figs. 7 and 8). In terms of disease activity, the relative abundances of representative luminal members such as Bacteroides, Ruminococcus, Bifidobacterium and Blautia decreased with disease severity, while those of skin and environmental bacteria (Phyllobacterium, Sphingomonas, Ralstonia, Planococcaceae and Staphylococcus) and oral and upper intestinal tract origin (Gemellaceae_UKG, Porphyromonas and Prevotella) increased. A similar trend was observed regarding the Matts score. Among the genera belonging to Enterococcaceae (Enterococcus and unassigned genus), some increased and others decreased. This difference is probably due to variations in proinflammatory potential among Enterococcus species, as described previously [52]. A larger number of differentially enriched genera (> 4-fold change) was detected in the analysis based on disease activity compared with Matts score (Additional file 6 and Fig. 8), again suggesting that gut physiological dysfunction, compared with local mucosal inflammation, has a greater impact on MAM in pediatric UC.
Phascolarctobacterium was identified as the genus that was most differentially increased (moderate vs inactive) or second most differentially increased (mild vs inactive and severe vs inactive) in MAM compared with inactive disease (Additional file 6). This bacterial group is known to consume succinate [53, 54, 55], which might explain the increase in the succinate concentration at the mucous layer with mucosal inflammation. Succinate has been reported to act as a danger signal that induces macrophages to produce IL-1β [56]. Among the gut microbes, Bacteroides, Parabacteroides and Prevotella produce succinate as an end product of sugar metabolism. The alteration of succinate metabolism in MAM with disease severity might be involved in mucosal inflammation in pediatric UC.
Finally, we identified potential microbial markers in MAM to discriminate UC from non-IBD pediatric patients (Fig. 9). Lactobacillus and Enterococcus were identified as potential markers for UC; and others, including typical luminal bacteria such as butyrate-producing bacteria (Faecalibacterium and Blautia), were identified as potential markers for non-IBD pediatric patients. Fusobacterium of oral origin has been associated with periodontitis [57–60] and colon adenocarcinoma [61–68]. However, our results might indicate that the resident gut Fusobacterium plays an important role to keep homeostasis on host-microbe interaction.
There are some limitations in this study. First, the small cohort size in this study was insufficient to reach statistical power in some analyses. Second, therapeutics may have affected the MAM, but therapeutic information for individual patients was not included in the analysis. Third, non-IBD controls in this study had abdominal symptoms, and healthy children were not enrolled as a control. Fourth, we could not do quantitative analyses (microbial cell density or short chain fatty acid level) due to the difficulty in adjusting the sampling volume. Finally, we did not perform the longitudinal evaluations in all the patients. Since UC is a condition with repeats in relapse and remission, we are now evaluating the longitudinal progress of MAM in individual patients.