130 anti-A4-Fla2, anti-CBir1, anti-Fla-X and anti-OmpC antibodies showed overlap with 301, 304, 302 and 166 antibody-bound peptide sequences, respectively (e-values < 0.05, Table S21). Most of the peptides that showed high sequence homology with anti-A4-Fla2, anti-CBir1 and anti-Fla-X sequences were overrepresented in patients with CD, and all three were very closely related to the same antibody-bound peptides (292 [~97%] shared antibody-bound peptides) (Supplementary Figure S2). Among antibody-bound peptides matching with anti-A4-Fla2, anti-CBir-1 and antiFla-X antibodies, 71 (~23%) were among those peptides that were differentially abundant between patients with CD and population controls (Table S4, Figure 3). Notably, antibodybound peptides that demonstrated a high degree of sequence identity and/or alignment length also represented the strongest differentially abundant peptides between patients with CD and population controls (Supplementary Figure S3). In contrast, anti-OmpC antibodies showed only a few unique matches with antibody-bound peptides (166 matches with e-values < 0.05), and these peptides also demonstrated very low prevalences in both patients with IBD and population controls (most < 2%). Concordance between antibody responses and microbiome composition Next, we aimed to investigate the concordance between patients’ serum antibody responses and their fecal microbial composition (Supplementary Figures S4-S5). We leveraged fecal metagenomics data available for a subset of the present cohort and generated within one year of sampling (n = 137). Relative abundances of bacterial taxa were compared between carriers and non-carriers of enriched antibody-bound peptides (Table S22), but no significant associations between bacterial taxa and antibodies were observed (FDR-correction for 448,987 tests). However, when evaluating the top-associated microbial taxa (based on nominal P-values), the observed associations were considerably smaller in effect size and demonstrated a high degree of discrepancy (i.e. many microbial taxa were not matched to their associated antibody-bound peptides). To further evaluate the impact of microbiome composition on antibody responses, we analyzed the extent to which microbial taxa could explain the variation in antibody responses (Table S23). In addition, we performed Principal coordinate analysis (PCoA) on the metagenomics data and defined dysbiosis scores, which represent the median Bray-Curtis dissimilarity distances to a set of reference controls. The first two principal coordinates (PCos) explained a total of 66.6% of the variation in metagenomics data (PCo1: 53.8%; PCo2: 12.8%) and were associated with the presence of IBD (Supplementary Figure S4A). Patients with IBD showed considerably higher dysbiosis scores compared to healthy individuals, whereas there were no notable differences between IBD subtypes (Supplementary Figures S4B–C). Importantly, however, dysbiosis scores were not associated with the first ten PCs describing variation in antibody data (PC1 and PC2 shown in Supplementary Figure 5A). In addition, no differentially abundant antibody-bound peptides were observed when comparing patients with a dysbiotic vs. a non-dysbiotic microbial composition (Table S24). Finally, dysbiosis scores were not associated with antibody diversity (i.e. the number of different enriched antibody-bound peptides per patient, Supplementary Figure S5B). Chapter 4
RkJQdWJsaXNoZXIy MjY0ODMw