The difference in gut bacteria between healthy controls and IBD patients

We first analyzed whether IBD patients have different gut microbiota compared with healthy controls. Information on IBD patients (n = 20) and healthy controls (n = 12) is indicated in Table 1 and Supplementary Table 1. By 16S rRNA sequencing, we observed a significantly different diversity of microbiota between healthy controls and IBD patients (Supplementary Fig. 1a-b). In the case of phylum level, the composition of microbiota was different between healthy controls and IBD patients. IBD patients showed an increase in Proteobacteria and Campylobacterota and a decrease in Firmicutes and Verrucomicrobiota, especially Firmicutes, which was significantly reduced in both UC and CD patients. (Fig. 1a-d, Supplementary Fig. 1c-d). In the analysis of the amplicon sequencing variant (ASV) levels, there was no significant difference because the microbiota differed among individuals (Supplementary Fig. 2a). To analyze more details of the difference in microbiota composition between the two groups, we performed Lefse analysis. The increase　of Enterobacteriaceae and Campylobacterota and the decrease of Lachnospiraceae and Verrucomicrobiota were confirmed (Supplementary Fig. 2b). Other articles also reported that these bacteria are enriched in IBD patients [3, 25, 26]. These results indicate that our samples represent a widespread dysbiosis in IBD.

Fig. 1

IBD patients have aberrant IgA to gut bacteria. a–d Relative abundances of a Firmicutes, b Verrucomicrobiota, c Proteobacteria, and d Campylobacterota in healthy controls (HC) (n = 12) and IBD patients (n = 20). e–f The concentration of fecal human endogenous IgA and IgG in HC (n = 12) and IBD patients (n = 20, UC: n = 13, CD: n = 7). g Frequency of human endogenous IgA-bound bacteria in gut bacteria of HC (n = 12) and IBD patients (n = 20). h–i Human endogenous IgA index of h HC (n = 12) and i IBD patients (n = 16). Statistical analysis was performed by (a-g) by Mann–Whitney U test. Data are expressed as mean ± s.d. in (a–g), and median ± range in (h–i)

Disorder of endogenous IgA binding ability to microbiota in IBD patients

Because IgA plays a pivotal role in the regulation of gut microbiota, we hypothesized that dysbiosis in IBD patients is caused by aberrant production of intestinal IgA in either quantity or quality. Therefore, we measured the concentration of fecal endogenous IgA and IgG in IBD patients and healthy controls. The fecal IgA and IgG were significantly increased in IBD patients compared with healthy controls, indicating enough amount of antibody production in IBD patients of this study (Fig. 1e-f). Fecal IgG production, but not IgA, is significantly increased in UC patients (Supplementary Fig. 3a). Next, we examined the relationship between antibody level and the severity score of UC patients (Mayo score) or of CD patients (CD activity index; CDAI). The correlation between antibody levels and severity score could only be observed in fecal IgG levels vs. Mayo score in UC patients, but not other correlations (Supplementary Fig. 3b-c), as reported previously [27, 28].

Since the quality of IgA is important to maintain gut homeostasis [13, 17], we checked whether endogenous IgA in the two groups has different binding ability against gut bacteria. In addition to the increasing quantity of fecal IgA, the proportion of endogenous IgA-bound bacteria was also significantly increased in IBD patients compared to healthy controls (Fig. 1g, Supplementary Fig. 3d). In UC patients, the proportion of IgA-bound bacteria and the severity score have a strong correlation, but not in CD patients (Supplementary Fig. 3e). These results suggest that the binding ability of endogenous IgA is helpful to evaluate the severity of UC patients, but not of CD patients.

Next, we analyzed what kinds of bacteria were trapped by endogenous IgA. We sorted endogenous IgA-bound and -unbound bacteria (Supplementary Fig. 4a). In healthy controls, we could observe a significant difference in β-diversity, but not in α-diversity (Supplementary Fig. 4b-c). Lefse analysis revealed that Enterobacteriaceae was enriched in IgA-bound bacteria (Supplementary Fig. 4d). In IBD patients, however, there was no significant difference in diversity analysis (Supplementary Fig. 4e-f). In Lefse analysis in IBD patients, endogenous IgA bound not only to Enterobacteriaceae but also to Lactobacillaceae (Supplementary Fig. 4 g), indicating aberrant binding ability of intestinal IgA derived from IBD patients.

Because Lefse analysis could not completely reflect individual microbiota, we performed IgA-seq to evaluate IgA-bound (IgA index > 0) and -unbound (IgA index = 0 or < 0) bacteria in individuals more precisely. In healthy controls, their endogenous IgA strongly bound to Enterobacteriaceae in IgA-seq (Fig. 1h). However, in IBD patients, their endogenous IgA bound most strongly to Lactobacillaceae, in addition to Veillonellaceae, Enterobacteriaceae (Fig. 1i, Supplementary Fig. 4 h-i). These results indicate that IBD patient-derived endogenous IgA has a disordered binding ability to microbiota compared with healthy control-derived ones and that this disorder of IgA quality may cause dysbiosis in IBD patients.

Mouse IgA binds to human colitogenic gut bacteria

Because IBD patient-derived endogenous IgA has an aberrant binding ability to gut bacteria, we hypothesized that oral treatment of mouse IgA, which binds selectively to colitogenic bacteria, may ameliorate dysbiosis in IBD patients. To prove our hypothesis, we selected six mouse IgA clones (rW27, W37, rPG151A, rPG160A, PGSI1A, and SPFSI12), which bind to both E. coli and C. difficile (Supplementary Fig. 5a-d).

We addressed the question of whether these mouse IgA bind to human microbiota. As shown in Fig. 2a, all six mouse IgA bound to gut bacteria of both healthy controls and IBD patients. Four clones (rW27, W37, rPG151A, and rPG160A) showed significantly high-binding ability to microbiota from IBD patients compared with healthy controls (Fig. 2a, Supplementary Fig. 6a). We next sorted and analyzed mouse IgA-bound and -unbound bacteria in IBD patients. Diversity analysis was not significantly different (Supplementary Fig. 6b-c). Next, we performed Lefse analysis between mouse IgA-bound and -unbound bacteria. The rW27, W37, and PGSI1A showed a high binding ability to Enterobacteriaceae (Supplementary Fig. 7). In addition, rW27 clearly showed a weak binding ability to Bifidobacteriaceae, indicating that rW27 is the best to distinguish harmful from beneficial bacteria in human gut bacteria (Supplementary Fig. 7). In contrast to rW27, rPG150A, rPG160A, and SPFSI12, showed weak binding to Actinobacteriota but did not bind to harmful bacteria (Supplementary Fig. 7).

Fig. 2

Mouse IgA antibodies bind to human colitogenic bacteria. a Frequency of mouse IgA-bound gut bacteria in HC (n = 12) and IBD patients (n = 20). b Mouse IgA index of IBD patients (rW27 (n = 16), W37, rPG151A, rPG160A, PGSI1A, and SPFSI12 (n = 12)). Statistical analysis was performed by a Mann–Whitney U test. Data are expressed as mean ± s.d. in (a), and median ± range in (b)

To assess more details of the binding ability of each antibody to gut bacteria, we performed IgA-seq (Fig. 2b, Supplementary Fig. 8a-b). IgA-seq revealed that all six mouse IgA showed high IgA index to Enterobacteriaceae and Gemellaceae, but low IgA index to beneficial bacteria, such as Lactobacillaceae and Lachnospiraceae, including short-chain fatty acid (SCFA)-producing bacteria in IBD patients (Fig. 2b). In addition, rW27 also bound to Veillonellaceae and Enterococcaceae, both of which are considered as colitogenic bacteria (Fig. 2b) [29, 30]. These results suggest that, among six IgA clones, rW27 is the best candidate IgA binding to a broad range of colitogenic bacteria including Enterococcaceae [30].

To confirm whether rW27 selectively binds to colitogenic bacteria, we next analyzed rW27-bound and -unbound bacteria in healthy controls, in which colitogenic bacteria are not enriched. Diversity analysis showed no differences in α-diversity, but a significant difference in β-diversity (Supplementary Fig. 9a-b). In Lefse analysis, Enterobacteriaceae was not detected, but Clostridiaceae was recognized as rW27-bound bacteria in healthy controls (Supplementary Fig. 9c). Furthermore, IgA-seq revealed that rW27 antibody showed weak binding (IgA index = 0 or < 0) to healthy control-derived microbiota (Supplementary Fig. 9d).

Taken together, mouse IgA, especially rW27, selectively binds to colitogenic bacteria to which IBD patients-derived endogenous IgA have not strongly bound. These results raise a possibility that oral treatment of mouse IgA ameliorates dysbiosis of IBD patients by complementing the aberrant endogenous IgA secreted in IBD patients.

Mouse IgA inhibits the growth of IgA-bound colitogenic bacteria

Because our previous report [17] indicated that W27 IgA derived from hybridoma has shown growth inhibition activity against E. coli, we analyzed whether rW27 inhibits the growth of strongly rW27-bound bacteria, such as Enterobacteriaceae, Enterococcaceae, Gemellaceae, and Veillonellaceae (Fig. 2b). We examined the abundance of these four kinds of bacteria in IBD patient-derived microbiota and found that a particular patient had a high proportion of unique bacterial family (Supplementary Fig. 10a–d). For growth inhibition assay, we selected E. coli, E. faecium, G. morbillorum, and V. dispar as representative bacterial strains from each family. We first checked whether rW27 binds to these bacteria and confirmed that rW27 binds to them (Fig. 3a-b).

Fig. 3

rW27 inhibits the growth of human colitogenic bacteria. (a-b) Binding analysis of indicated monoclonal IgA to E. coli, E. faecium, G. morbillorum and V. dispar by a ELISA and b FACS. b Grey histogram: unstained bacteria. Red histogram: mouse IgA-bound bacteria. c Growth inhibition of E. coli (n = 6), E. faecium (n = 3), G. morbillorum (n = 3), and V. dispar (n = 6) mediated by rW27. a, b Representative data of repeated experiments. Results of E. coli were same as in Supplementary Fig. 5 a and c. Statistical analysis was performed by c unpaired Student’s t test. Data are expressed as mean ± s.d. in (c)

As reported previously [17], we tested whether rW27 inhibits the growth of them and found that rW27 significantly suppressed the growth of these kinds of bacteria (Fig. 3c). These findings suggest that rW27 may inhibit the growth of these colitogenic bacteria in the intestinal lumen.

Mouse IgA suppress gut inflammation in gnotobiotic mice with IBD patients-derived microbiota

The rW27 showed a high binding ratio to intestinal bacteria of IBD patients (Fig. 2a) and inhibited the growth of rW27-bound bacteria (Fig. 3c). Therefore, we investigated whether oral administration of mouse IgA to gnotobiotic mice with IBD patient-derived microbiota could improve intestinal dysbiosis and influence susceptibility to colitis. Because normal endogenous IgA in wild-type mice affect the gut bacteria, we used AID-deficient mice [9, 10], which do not produce IgA in the gut. We inoculated mice with IBD patient-derived microbiota (UC10) with the highest rW27 binding proportion (44.7%) to clearly evaluate the effect of rW27 (Fig. 2a, Supplementary Fig. 6a). Considering that there are differences in the effect between antibodies due to the different microbiota in each patient, a comparative study was conducted with W37, which also showed a high binding ratio (62.5%) to UC10 microbiota (Fig. 2a, Supplementary Fig. 6a).

After inoculated with UC10 microbiota, mice were orally administered with mouse IgA. During the oral administration period of antibody, mice were induced colitis with DSS, and the sensitivity of colitis was evaluated (Fig. 4a). We could not observe any significant difference in body weight change and feces score among three groups (Supplementary Fig. 11a). Therefore, to evaluate the low levels of inflammation in the colon, we measured the concentration of fecal lipocalin 2 as an inflammation marker before (day 10) and after (day 14) DSS treatments. Interestingly, orally treated mice with both rW27 and W37 showed no increase of fecal lipocalin 2 compared with the untreated group (Fig. 4b). The weight/length ratio of the colon is also lower in the mouse IgA-treated groups than in the untreated group (Fig. 4c). In addition, we inoculated mice with other IBD patient-derived microbiota (CD6) with the highest endogenous IgA binding proportion (88.2%, Fig. 1g, Supplementary Fig. 3d). In these mice, the increase in lipocalin 2 levels after DSS in the rW27 group was suppressed compared with the other groups (Fig. 4d-e, Supplementary Fig. 11b). Compared to mice inoculated with IBD patients-derived microbiota, mice inoculated with healthy control-derived ones had little increase in lipocalin 2 (Fig. 4f-g, Supplementary Fig. 11c). These findings suggest that susceptibility of DSS-induced colitis was reduced by oral treatment of rW27.

Fig. 4

The rW27 oral treatment ameliorates gut inflammation in gnotobiotic mice with IBD patients-derived gut bacteria. a Schedule of fecal transplantation, IgA and DSS treatments. b, c b Concentration of fecal lipocalin 2 at day 10 and day 14, c The colon weight/length ratio in gnotobiotic mice with UC10-derived microbiota (n = 4), d, e d Concentration of fecal lipocalin 2 at day 10 and day 14, e The colon weight/length ratio in gnotobiotic mice with CD6-derived microbiota (n = 3). f, g f Concentration of fecal lipocalin 2 at day 10 and day 14, g The colon weight/length ratio in healthy control-derived microbiota (HC5: n = 3, HC8: n = 4, HC10: n = 4). Statistical analysis was performed by b–g Kruskal–Wallis test with Dunn’s multiple comparison test or one-way ANOVA with Tukey's multiple comparisons test. Data are expressed as mean ± s.d. in (b–g)

Mouse IgA remedy dysbiosis in gnotobiotic mice with IBD patients-derived microbiota

Because oral administration of mouse IgA suppressed intestinal inflammation, we hypothesized that antibody administration affects gut microbiota in gnotobiotic mice. We compared microbiota before (day 7) and after (day 10) IgA administration. While the microbiota of UC10 patient contained Fusobacteriota and Proteobacteria, Firmicutes and Bacteroidota were predominant in gnotobiotic mice, indicating the transplanted microbiota could not be reproduced completely in the mouse intestine (Fig. 5a, Supplementary Fig. 12a-b). Only limited bacteria from donor’s microbiota were reproduced in mice inoculated with CD6- or healthy control-derived gut bacteria (Supplementary Fig. 12c-d). Since Fusobacteriaceae were not originally present in mice intestine, they were usually difficult to colonize. However, Fusobacteriaceae were detected in 7 of 12 animals in the gut on day 10, indicating a partial value for UC10 fecal transplantation experiment (Fig. 5a, Supplementary Fig. 12a). Comparison of α-diversity between day 7 and day 10 showed the significant increase in rW27- and W37-treated group, but not in the untreated group (Fig. 5b). In addition, β-diversity was significantly changed at day 10 compared with day 7 in the rW27-treated group, indicating that mouse IgA affected gut microbiota (Fig. 5c).

Fig. 5

Mouse IgA antibodies modulate gut bacteria of gnotobiotic mice with IBD patient-derived gut bacteria. a Relative abundances of bacterial taxa at phylum level in UC10 patient gut microbiota, PBS-, rW27-, and W37-treated mice at day 7 and day 10 (n = 4). Red arrows indicate Fusobacteriaceae. b Shannon index of PBS-, rW27-, and W37-treated mice at day 7 (left bar) and day 10 (right bar) (n = 4). (c) Unweighted unifrac distance of PBS-, rW27-, and W37-treated mice at day 7 (blue circle) and day 10 (red circle) (n = 4). d–f Relative abundance of indicated bacteria of d PBS-, e rW27-, and f W37-treated mice at day 7 and day 10 (n = 4). Statistical analysis was performed by b, d–f by Mann–Whitney U test and c PERMANOVA comparison. Data are expressed as median ± range in (b) and mean ± s.d. in (d–f)

In the untreated group, Fusobacteriaceae was colonized at day 7 but not detected at day 10 because the abundance of this bacteria was low (Fig. 5d). In Lefse analysis, there was a significant increase in Eggerthellaceae and Muribaculaceae, which were reported as inflammation-inducing bacteria (Fig. 5d, Supplementary Fig. 13a) [31, 32]. In contrast, the rW27-treated group showed that an increase in Oscillospiraceae including SCFA-producing bacteria and a decrease in Enterobacteriaceae and Fusobacteriaceae (Fig. 5e, Supplementary Fig. 13b), indicating that rW27 administration improved dysbiosis. Although W37 reduced fecal lipocalin 2, Fusobacteriaceae was not changed between day 7 and day 10 (Fig. 5f, Supplementary Fig. 13c). However, an increase of Oscillospiraceae and a decrease of Enterobacteriaceae were also observed in the W37-treated group (Fig. 5f, Supplementary Fig. 13c). The different effects of rW27 and W37 on human microbiota suggest that an appropriate selection of antibodies is necessary to remedy dysbiosis in individual patient.