In-depth characterization of the serum antibody epitope repertoire in inflammatory bowel disease using phage-displayed immunoprecipitation sequencing

Arno R. Bourgonje, Sergio Andreu‐Sánchez, Thomas Vogl, Shixian Hu, Arnau Vich Vila, Ranko Gaćeša, Sigal Leviatan, Alexander Kurilshikov, Shelley Klompus, Iris Kalka, Hendrik M. van Dullemen, Adina Weinberger, Marijn C. Visschedijk, Eleonora A. Festen, Klaas Nico Faber, Cisca Wijmenga, Gerard Dijkstra, Eran Segal, Jingyuan Fu, Alexandra Zhernakova, Rinse K. WeersmabioRxiv 2021 · DOI

TL;DR

This study uses a high-throughput method called PhIP-seq to map the antibody "fingerprint" in patients with Inflammatory Bowel Disease (IBD). They discovered distinct antibody patterns associated with IBD, particularly Crohn's disease (CD), and found that even a small set of antibodies can accurately distinguish CD patients from healthy individuals, opening doors for new diagnostic and therapeutic strategies.

What Problem Does This Paper Address?

Inflammatory Bowel Disease (IBD), encompassing Crohn's disease (CD) and ulcerative colitis (UC), is a chronic immune-mediated condition affecting the gut. While the immune system's role is clear, the specific targets and mechanisms driving IBD are complex and not fully understood. Specifically, the vast landscape of antibodies produced by the immune system and their potential involvement in IBD pathogenesis has been difficult to comprehensively map. Traditional methods are limited in scale, making it challenging to get a detailed picture of the antibody epitope repertoire – the specific parts of antigens that antibodies bind to. This gap hinders the discovery of new biomarkers for diagnosis and potential therapeutic targets.

Methods

The core of this study is the application of Phage-Displayed Immunoprecipitation Sequencing (PhIP-seq). This is a high-throughput technique that allows researchers to simultaneously screen for antibodies against a massive library of antigens. Here's how it works:

  1. Antigen Library: A library of 344,000 different antigen peptides (derived from bacterial genes, phages, allergens, etc.) is displayed on bacteriophages.
  2. Incubation: Human blood serum is incubated with these phages. If antibodies are present in the serum that recognize specific antigens, they will bind to the corresponding phages.
  3. Immunoprecipitation: The antibody-bound phages are captured.
  4. Sequencing: The DNA of these captured phages is sequenced to identify which antigens the antibodies were binding to.

This allows for a comprehensive characterization of the serum antibody epitope repertoire. The study also employed Principal Component Analysis (PCA) and k-means clustering to visualize and group patients based on their antibody profiles, and logistic regression with elastic net penalty for classification tasks.

Datasets

  • 1000IBD cohort: A large, multi-omics IBD cohort based in Groningen, Netherlands, comprising 497 patients with established IBD (256 with CD, 207 with UC, and 34 with undetermined IBD).
  • Lifelines-DEEP cohort: An independent population-based cohort from the northern Netherlands, used as a control group, consisting of 1,326 healthy individuals.
  • External dataset: A publicly available dataset from a previous population-based cohort study (997 healthy individuals) was used to test the external generalizability of the classification models.

Key Insights and Ideas

  • IBD has a distinct antibody "fingerprint": The study clearly shows that patients with IBD, particularly CD, have a unique antibody repertoire compared to healthy individuals. This suggests that the immune system's response to various antigens is altered in IBD.
  • Flagellins are key players in CD: Antibodies against bacterial flagellins, especially from the Lachnospiraceae family, are significantly overrepresented in CD patients, particularly those with ileal involvement and more severe disease. This highlights a potential link between gut bacteria and the immune response in CD.
  • A small antibody panel can diagnose CD: Surprisingly, even a panel of just ten specific antibody-bound peptides can accurately discriminate CD patients from healthy controls, achieving an AUC of 0.87. This is almost as good as using the full repertoire. This is a major breakthrough for potential diagnostic tools.
  • CMV antibodies drive clustering: The clustering observed in the antibody repertoire data was largely driven by antibodies against Cytomegalovirus (CMV), suggesting a significant role for this virus in shaping the immune response in some IBD patients.
  • Microbiome link is complex: While the study investigated the connection between gut microbiome composition and antibody responses, no clear associations were found at the individual microbial taxon or dysbiosis level. This suggests the relationship is more intricate than a simple one-to-one correlation.

Key Results

  • 373 differentially abundant antibodies: The study identified 373 antibody-bound peptides that were significantly different between IBD patients and healthy controls. Of these, 202 were overrepresented and 171 were underrepresented in IBD.
  • CD vs. UC specificity: 55% of the differentially abundant antibodies were unique to CD, while 28% were unique to UC, and 17% were shared.
  • High diagnostic accuracy for CD: Antibody epitope repertoires accurately discriminated CD from healthy controls with an Area Under the Curve (AUC) of 0.89 and an F1-score of 0.80.
  • Ten antibodies are sufficient for CD diagnosis: Using only ten antibody-bound peptides, discrimination between CD and healthy controls achieved an AUC of 0.87 and an F1-score of 0.77.
  • UC discrimination is good, but less so than CD: Discrimination between UC and healthy controls had an AUC of 0.80 and an F1-score of 0.70.
  • CD vs. UC discrimination is moderate: The ability to distinguish between CD and UC patients using antibody repertoires was moderate, with an AUC of 0.68 and an F1-score of 0.66.
  • External validation successful for CD: Classification models for CD showed good performance on an external dataset, demonstrating generalizability.

Limitations and Future Work

  • Disease remission: The majority of IBD patients in the cohort were in remission, which limited the ability to study differences in antibody responses related to active disease or new-onset IBD.
  • Disease activity assessment: The assessment of disease activity was primarily based on clinical and serological parameters, with limited endoscopic data or fecal calprotectin levels.
  • PhIP-seq technical aspects:
    • Peptide length restriction (up to 64 amino acids) and the absence of conformational epitopes might limit the full characterization of antibody responses.
  • Only protein antigens are included; non-protein antigens (lipids, glycans) are not represented, though these typically elicit T-cell-independent responses.
  • Interpretation of single peptide associations: Associations with single peptides should be interpreted cautiously; associations covering multiple peptides for the same protein are considered more reliable.
  • Coagulation-associated peptides: Comparisons involving serum vs. plasma could lead to underrepresentation of certain peptides, requiring cautious interpretation.

Future work could focus on studying patients with active disease, incorporating more detailed disease activity assessments, and exploring the role of non-protein antigens. The PhIP-seq technology is highlighted as a powerful tool for discovering systemic immune-based biomarkers and immunological targets for IBD, potentially aiding in disease prevention and personalized treatment strategies.

Summary updated

Abstract

Summary Inflammatory bowel diseases (IBD), e.g. Crohn’s disease (CD) and ulcerative colitis (UC), are chronic immune-mediated inflammatory diseases. A comprehensive overview of an IBD-specific antibody epitope repertoire is, however, lacking. We leveraged a high-throughput phage-displayed immunoprecipitation sequencing (PhIP-seq) workflow to identify antibodies against 344,000 antimicrobial, immune and food antigens in 497 IBD patients as compared to 1,326 controls. IBD was characterized by 373 differentially abundant antibodies (202 overrepresented and 171 underrepresented), with 17% shared by both IBDs, 55% unique to CD and 28% unique to UC. Antibodies against bacterial flagellins dominated in CD and were associated with ileal involvement, fibrostenotic disease and anti- Saccharomyces cerevisiae antibody positivity, but not with fecal microbiome composition. Antibody epitope repertoires accurately discriminated CD from controls (AUC=0.89), and similar discrimination was achieved when using only ten antibodies (AUC=0.87). IBD patients thus show a distinct antibody repertoire against selected peptides, allowing patient stratification and discovery of immunological targets.
BibTeX
@misc{bourgonje2021depth,
  author = {Arno R. Bourgonje and Sergio Andreu‐Sánchez and Thomas Vogl and Shixian Hu and Arnau Vich Vila and Ranko Gaćeša and Sigal Leviatan and Alexander Kurilshikov and Shelley Klompus and Iris Kalka and Hendrik M. van Dullemen and Adina Weinberger and Marijn C. Visschedijk and Eleonora A. Festen and Klaas Nico Faber and Cisca Wijmenga and Gerard Dijkstra and Eran Segal and Jingyuan Fu and Alexandra Zhernakova and Rinse K. Weersma},
  title = {{In-depth characterization of the serum antibody epitope repertoire in inflammatory bowel disease using phage-displayed immunoprecipitation sequencing}},
  year = {2021},
  doi = {10.1101/2021.12.07.471581},
  howpublished = {bioRxiv},
}

Loading related content...

Figure