Alteration of microbiota antibody-mediated immune selection contributes to dysbiosis in inflammatory bowel diseases

Abstract

Human secretory immunoglobulins (SIg) A1 and SIgA2 guide mucosal responses toward tolerance or inflammation, notably through reverse-transcytosis, the apical-to-basal transport of IgA2 immune complexes via M cells of gut Peyer's patches. As such, the maintenance of a diverse gut microbiota requires broad affinity IgA and glycan-glycan interaction. Here, we asked whether IgA1 and IgA2-microbiota interactions might be involved in dysbiosis induction during inflammatory bowel diseases. Using stool HPLC-purified IgA, we show that reverse-transcytosis is abrogated in ulcerative colitis (UC) while it is extended to IgA1 in Crohn's disease (CD). 16S RNA sequencing of IgA-bound microbiota in CD and UC showed distinct IgA1- and IgA2-associated microbiota; the IgA1⁺ fraction of CD microbiota was notably enriched in beneficial commensals. These features were associated with increased IgA anti-glycan reactivity in CD and an opposite loss of reactivity in UC. Our results highlight previously unknown pathogenic properties of IgA in IBD that could support dysbiosis.

Keywords: IBD; SIgA; glycosylation; immunity; microbiota.

Figure 1. Purified fecal IgA1 and IgA2 from IBD patients have altered functionality related to Dectin‐1 binding

1. A–D

   ELISA assay of fecal IgA1 (A), IgA2 (B), total IgA (C), and secretory component (D) levels from non‐IBD (n = 7), CD (n = 18), and UC (n = 12) patients' stool.

2. E, F

   Data were further separated according to disease activity for IgA1 (E) and IgA2 (F).

3. G, H

   In vitro assay of purified IgA1 (G) and IgA2 (H) reverse‐transcytosis abilities on an inverted model of FAE from Caco2 and Raji cells co‐culture.

4. I, J

   (I) ELISA assay of IgA1‐Dectin‐1 binding, at a rate of one receptor per 10 IgA; (J) ELISA assay of IgA2‐Dectin‐1 binding, at a rate of one receptor per 10 IgA.

5. K–N

   Percent of IgA1 (K, L) and IgA2 (M, N) reverse‐transcytosis for weak (K, M) and strong (L, N) Dectin‐1 binding.

Data information:  Data were analyzed using Kruskall–Wallis multiple comparisons with Dunn's correction, when possible, or a Mann–Whitney test. P‐values are as follows: (A) **P = 0.0098; (B) *P = 0.0406; (C) Non‐IBD vs. UC *P = 0.0479, CD vs. UC *P = 0.0166; (F) *P = 0.0167; (G) *P = 0.0441; (H) *P = 0.423; (J) Non‐IBD vs. UC *P = 0.0199, CD vs. UC *P = 0.0243; (K) *P = 0.0221; (L) *P = 0.0486. (G–N) For some patients, antibody purification did not yield a high enough concentration, so samples had to be excluded. N are thus as follows: IgA1: Non‐IBD: n = 7; CD: n = 8; UC: n = 4; IgA2: Non‐IBD: n = 7; CD: n = 8; UC: n = 5. All patient samples (biological replicates) have been tested in technical duplicates meaning n × 2.

Figure 2. CD IgA1 N and O‐glycosylations do not recapitulate conventional IgA glycosylation patterns

1. A

   Full MS spectrum of glycopeptides released from CD IgA1 trypsin digest. Main N340‐glycoforms of the IgA1 glycopeptide [332–353] (LAGKPTHVNVSVVMAEVDGTCY) are annotated using CFG nomenclature. Appendix Table S4 summarizes a list of N‐glycopeptides identified. # is the peptide [43–76] of immunoglobulin kappa constant chain (UniProtKB P01834). *Isobaric structures not differentiated by MS/MS experiments (not exhaustive N‐glycans illustrations). (HYTNPSQDVTVPCPVPSTPPTPSPSTPPTPSPSCCHPR).

2. B

   O‐glycoforms identified for CD.

3. C

   O‐glycoforms identified for UC.

Data information: *corresponds to IgA1 peptide [264–273] (WLQGSQELPR). ^#shows contamination by other multiply charge species covering the glycoforms signals. Details related to O‐glycoforms are given in Appendix Table S7. CD: n = 3; UC: n = 1 (biological replicates). All patient samples have been tested in technical duplicates.

Figure 3. CD and UC IgA have distinct anti‐glycan reactivity profiles

1. A, B

   Heatmaps displaying standardized log2‐transformed signal intensities per glycan for IgA1 (A) and IgA2 (B) in non‐IBD, CD and UC groups, plotted by hierarchical clustering of Euclidean distance.

2. C–F

   Volcano plots of log2‐transformed fold changes for IgA1 anti‐glycan reactivity in CD (C) and UC (D) groups, and IgA2 anti‐glycan reactivity for CD (E) and UC (F) compared to non‐IBD IgA1 and IgA2. For CD, P‐values of differentially targeted glycan motifs are listed in Appendix Table S8 and those of UC in Appendix Tables S9 and S10. Significance threshold was placed at P‐value < 0.05 (indicated by the dotted line in C–F); For IgA1, Non‐IBD: n = 4; CD: n = 4; UC: n = 4. For IgA2, n = 6; CD: n = 6; UC: n = 6. IgA1 and IgA2 were taken matched for the same patient. All patient samples (biological replicates) have been tested in technical duplicates. A list of the glycans from each cluster is provided as Appendix Tables S11 and S12.

Figure 4. CD IgA1 have biased selection toward the microbiota

1. A–C

   Flow cytometry analysis of total IgA1‐bound (A) and IgA2‐bound (B) fecal microbiota. (C) Flow cytometry analysis of stool IgA1⁺ IgA2⁻ bacteria (gray dots), IgA1⁻ IgA2⁺ bacteria (black dots) and IgA1⁺ IgA2⁺ bacteria (empty dots). (A–C) Data were analyzed with a Kruskall–Wallis multiple comparison test. In (C) *P = 0.0226. Non‐IBD: n = 7, CD: n = 18, and UC: n = 12.

2. D

   Shannon and Chao1 diversity indices of IgA1‐bound microbiota in non‐IBD, CD and UC.

3. E

   PCA plot based on the Jaccard distance between samples.

4. F

   Phylum‐level composition of IgA1‐bound microbiota in non‐IBD, CD and UC.

5. G

   Boxplots of OTUs for which the abundance was significantly different between non‐IBD and CD.

6. H

   Shannon and Chao1 diversity indices of IgA2‐bound microbiota in non‐IBD, CD and UC.

7. I

   PCA plot based on the Jaccard distance between samples.

8. J

   Phylum‐level composition of IgA2‐bound microbiota in non‐IBD, CD and UC.

9. K

   Boxplots of OTUs whose abundance was significantly different between CD and UC, and between non‐IBD and UC, respectively. OTUs present in less than 25% of samples or with read count lower than 50 were filtered out. Differential abundance was tested using negative binomial model implemented in DESeq2 and p‐values corrected with False Discovery Rate (FDR) procedure. Non‐IBD: n = 4; CD: n = 4; UC: n = 4 (two patients excluded for abundance in both IgA1 and IgA2 analyses). All patient samples have been tested in technical duplicates.

Figure 5. IBD IgA are less neutralizing than non‐IBD IgAs against pathogenic bacteria

1. A–E

   Optical density (600 nm) variations during in vitro growth assay of Salmonella enterica Typhimurium (SL1344) co‐cultured with stool‐purified IgA1 (A) or IgA2 (B) from non‐IBD (n = 4), CD (n = 6) and UC (n = 3). Comparison between IgA1 and IgA2 coculture in non‐IBD (C), CD (D), and UC (E).

2. F–I

   Optical density (600 nm) during in vitro growth assay of Escherichia coli (25,922 strain from ATCC) cocultured with stool‐purified IgA1 (F) or IgA2 (G) from non‐IBD (n = 2) and CD (n = 2). All patient samples (biological replicates) have been tested in technical duplicates. Comparison between IgA1 and IgA2 co‐culture in non‐IBD (H), and CD (I). These IgA samples were from the same patients throughout each experiment.

Data information: Two‐way ANOVA with Holm–Sidak correction after D'Agostino–Pearson nomality test, P‐values are as follows: (A) UC curve: 2 h – P = 0.0024, 3 h – P = 0.003; CD curve: 2 and 3 h P < 0.0001 (B) UC curve: 4 h – P = 0.0186, 5 h – P = 0.0393, 6 h – P = 0.039, 8 h – P < 0.0001; CD curve: 4 h – P = 0.0001, 5 h – P = 0.0028, 6 h – P = 0.0024, 8 h – P < 0.0001. (C) 3 h – P = 0.0422, 4 h – P = 0.0422, 6 h – P = 0.0012, 8 h – P < 0.0001. (F) Gray curve: 4 h – P = 0.0004, 5 h – P = 0.0014; Black curve: 4 and 5 h – P < 0.0001. (G) De‐glycosylated CD vs. De‐glycosylated non‐IBD: 4 h – P = 0.0003, 5 h – P < 0.0001. (H) Native vs. de‐glycosylated: 4 and 5 h – P < 0.0001; De‐glycosylated IgA1 vs. De‐glycosylated IgA2: 5 h – P = 0.0012. (H) Native vs. de‐glycosylated: 4 h – P = 0.001 and 5 h – P = 0.006; De‐glycosylated IgA1 vs. De‐glycosylated IgA2: 5 h – P < 0.0001.

Figure 6. CD IgA2 display biased selectivity in active CD patients

1. A

   Combined FISH‐IF staining for each probe.

2. B

   Total fluorescence detected per probe and per biopsy. Erec482 – *P = 0.00233; Ato291 – *P = 0.0268; Bac303 – *P = 0.0196.

3. C

   Percent of co‐occurred IgA2 and FISH‐bound particles. Erec482 – *P = 0.0073.

Data information: Two‐way ANOVA with Holm–Sidak correction after D'Agostino–Pearson nomality test. Non‐IBD: n = 3; Active CD: n = 6; Inactive CD: n = 4. All patient samples (biological replicates) have been tested in technical duplicates.

Figure 7. Putative model for altered subclass‐dependent selection of the microbiota by IgA in IBD

While in the healthy gut, IgA1 have mostly neutralizing and/or bacteriostatic properties, IgA2 undergoes reverse‐transcytosis via Dectin‐1 and Siglec‐5 to deliver antigens to the immune cells of the PP. This mechanism is dependent on glycosylations, notably sialylations for Siglec‐5 recognition. This allows for efficient elimination of pathogens and opportunists. IgA‐mediated entrapment of commensal bacteria into the mucus is also glycan‐dependent, wherein bacteria, IgA and mucin glycosylation are required. In CD (middle panel), luminal binding of commensal by‐ and RT of IgA1, would promote higher antigen load to the lamina propria and establishment of responses against both commensals and opportunists, which would ultimately favor opportunist's growth. Mechanisms of IgA1 retro‐transport remain to be determined. In addition, de‐sialylation of IgA1 limits its effector function, and would affect adequate commensal selection. Tissular IgA2, in turn, is less efficient at neutralizing opportunists, with increased binding to commensals during active disease, which would favor opportunists again. In UC (right panel), absence of RT for IgA2 and overall loss of anti‐glycan reactivity would lead to impaired tolerogenic responses to commensals but also impaired responses against pathogens and opportunists, resulting in opportunist's resurgence within the microbiota.