Phage display sequencing reveals that genetic, environmental, and intrinsic factors influence variation of human antibody epitope repertoire

Abstract

Phage-displayed immunoprecipitation sequencing (PhIP-seq) has enabled high-throughput profiling of human antibody repertoires. However, a comprehensive overview of environmental and genetic determinants shaping human adaptive immunity is lacking. In this study, we investigated the effects of genetic, environmental, and intrinsic factors on the variation in human antibody repertoires. We characterized serological antibody repertoires against 344,000 peptides using PhIP-seq libraries from a wide range of microbial and environmental antigens in 1,443 participants from a population cohort. We detected individual-specificity, temporal consistency, and co-housing similarities in antibody repertoires. Genetic analyses showed the involvement of the HLA, IGHV, and FUT2 gene regions in antibody-bound peptide reactivity. Furthermore, we uncovered associations between phenotypic factors (including age, cell counts, sex, smoking behavior, and allergies, among others) and particular antibody-bound peptides. Our results indicate that human antibody epitope repertoires are shaped by both genetics and environmental exposures and highlight specific signatures of distinct phenotypes and genotypes.

Keywords: PhIP-seq; antibody repertoire; environment; genetics; lifestyle.

Figure 1.. PhIP-Seq antibody-bound peptide profiles of 1,443 individuals representative of the Dutch population show temporal stability and family similarity.

A. Cohort characteristics. Lifelines-Deep (LLD) is a population cohort from Northern Netherlands. In this work, we performed PhIP-Seq in 1,443 participants (including 26 trio families), 322 of whom have data from a second time point after 4 years. Other data layers include phenotypes (questionnaires and clinical measurements), genetics (imputed microarrays) and microbiome (bacterial taxonomic quantification). There is a higher proportion of females within the participants (57%). The age distribution is slightly left skewed, with a mean of 44.5 years (female effect on age = −1, p = 0.16). B. Prevalence of antibody-bound peptides in the population. X-axis depicts seroprevalence. Y-axis is the number of antibody-bound peptides with a given seroprevalence. C. Principal component analysis identified two clusters (color represents cluster labels after 2-medoids clustering). D. Jaccard distance between antibody repertories of 322 samples longitudinally followed 4 years apart and between unrelated samples. E. Jaccard distance between antibody repertories of 26 family trios and between unrelated participants.

Figure 2.. Peptide co-occurrence highlight the presence of motifs driving antibody cross-reactivity.

A. Correlation heatmap between peptides that belonged to co-occurrence modules of at least 10 peptides using 1,443 individuals. Annotation displays the taxonomic origin of each peptide and the cluster assigned by WGCNA. Module 5 is highlighted. B. Module 5 motif discovery. At left, a hierarchical clustering (average method) based on sequence similarity between the peptides belonging to module 5. At right, their multiple sequence alignment (each colored line represents an amino acid, gray indicates an alignment gap). Peptides’ colors indicate their taxonomic origin. C. Logo of the most significant motif from the module 5 sequences (MEME, E-value 7.1×10⁻⁶⁰). Y-axis represents bits of information for each position and amino acid. B/C Amino acid residues are colored according to their chemical properties represented in the same legend.

Figure 3.. Genetics contribute to antibody-bound peptide variability

A. Manhattan plot from genome-wide association study meta-analysis of 2,798 antibody-bound peptides in 1,745 participants (490 IBD). Genome-wide association threshold (5×10⁻⁸, blue) and study-wide significance (7×10⁻¹¹, red) are shown as horizontal lines. Labels indicate the three major loci identified. Colored dots represent a recessive model. Gray dots represent additive models. B. Peptide motif deconvolution maps of DR3, DQ2.5 and DR14 (amino acids code: negatively charged = red, positively charged = blue, polar uncharged = green, hydrophobic = black) compared with the Streptococcus agalactiae C5a peptidase peptide core and percentage of elution score (%Rank_EL: strong binding ≤ 2.0, weak binding 2.0–10.0, no binding > 10) predicted by NetMHCIIpan-4.0 . Predicted binding mode, polar molecular interactions (dashes, hydrogen bonds: green, salt bridges: yellow), binding energy and dissociation constant (Kd) of the Streptococcus agalactiae C5a peptidase peptide core (red cartoon and sticks) into HLA-II receptors (chain A in green and chain B in blue).

Figure 4.. Phenotype-antibody-bound peptide associations.

A. Bar plot displaying the number of associations per phenotype (FDR < 0.05). Phenotypes are grouped in categories. Peptides associated with > 5 phenotypes are grouped. Peptides associated with < 5 phenotypes are labeled ‘Other’. B. Smoking-linked antibody-bound peptide prevalence. X-axis shows prevalence of peptides in smokers. Y-axis shows the prevalence in non-smokers. Colors of dots depict peptide taxonomy. C,D. Autoimmune- and allergy-specific association counts of antibody-bound peptides, per category. Bacterial peptides are binned as “Bacteria”. Viral peptides are binned as “Virus”. Autoantigens or antigens to casein are binned as “Mammal”. Plant peptides are binned as “Plant”. Anti-SSA: anti–Sjögren’s-syndrome-related antigen A autoantibodies. Anti-CTD: anti-connective tissue diseases screening ratio. Anti-CCP: anti-cyclic citrullinated peptide.