115 Figure 1 | Methodological workflow of the study. (A) Antigens derived from bacterial genes and species, pathogenic, antibody-coated and probiotic bacteria, virulence factors, phages, allergens, immune epitopes and antigens for experimental controls (covering a total of 344,000 antigen peptides) were integrated into a synthetic DNA oligonucleotide library and displayed on bacteriophages. After addition of human blood samples, reactive antibodies from these samples bind to their corresponding antigens present in the library. After incubation, antibody-bound phages are extracted and separated from the unbound antibodies following immunoprecipitation (IP). Finally, antibody-bound phages are amplified in a high-throughput manner using next-generation sequencing (NGS). (B) Tree map visualization showing the composition of the microbiota antigen phage library. Approximately 36% of the full library consisted of antigens of bacterial genes (derived frommetagenomics sequencing data), followed by phages, allergens and immune epitopes (from the Immune Epitope Database, IEDB), bacterial strains, pathogenic bacteria, virulence factors (from the virulence factor database, VFDB), antibody-coated bacterial species, probiotic bacteria and several biological and technical controls (see Supplementary Methods for details). (C) Analysis of the PhIP-Seq data was performed in steps. First, a cohort description was provided consisting of demographic and clinical characteristics of patients with IBD. Subsequently, an overall characterization of the data was generated by calculating summary statistics of the antibody epitope repertoires and performing principal component analysis (PCA). This step was followed by case–control analyses in which significantly bound peptides were compared between patients with IBD and healthy individuals while controlling for the effect of age and sex. Predictions of the diagnosis of CD or UC using the antibody epitope repertoires were established by evaluating classification accuracy. Antibody epitope repertoires were then analyzed in relation to (IBDspecific) phenotypes, e.g. Montreal disease classification and surgical history. Finally, the concordance between blood antibody epitope repertoires and fecal microbiome data (metagenomics sequencing) was assessed. Abbreviations: GWAS, genome-wide association study; HLA, human leukocyte antigen; IBD, inflammatory bowel disease; IP, immunoprecipitation; LL-DEEP, Lifelines-DEEP cohort; NGS, next-generation sequencing. Overall characterization of serum antibody epitope repertoires in patients with IBD We performed dimensionality reduction analyses (principal component analysis, PCA) to visualize the heterogeneity of the antibody epitope repertoires in patients with IBD (Figure 2). Two distinct clusters were observed based on the first two principal components (PCs) (Figure 2A-C), and these were further confirmed by a k-means clustering analysis (Figure 2C). Antibody epitope diversity (i.e. the number of enriched antibody-bound peptides per patient) clearly explained the variation in both PC1 and PC2 (both P < 0.001). After comparing antibody-bound peptides between both cluster identities, this clustering was found to be mainly driven by antigens derived from cytomegalovirus (CMV). This observation was confirmed when comparing both clusters for matching serological CMV measurements (measured by ELISA for n = 297 patients, R = 0.91, P < 0.001, Supplementary Figure S1A) and was also in line with the clustering observed in a population-based cohort.22 No significant batch effects were observed, as the IDs of plates onto which samples were loaded were not associated with the first two PCs or the subsequent eight PCs, which cover a cumulative explained variance of 21% in the antibody data (Supplementary Figure S1B). The median number of positive antibody-bound peptides per individual was 1,015 The antibody epitope repertoire in IBD
RkJQdWJsaXNoZXIy MjY0ODMw