A plasma peptidomic signature reveals extracellular matrix remodeling and predicts prognosis in alcohol-related hepatitis

Khaled Sayed, Christine E. Dolin, Daniel W. Wilkey, Jiang Li, Toshifumi Sato, Juliane I. Beier, Josepmaria Argemí, Ramón Bataller, Abdus S. Wahed, Michael L. Merchant, Panayiotis V. Benos, Gavin E. ArteelmedRxiv 2023 · DOI

TL;DR

Researchers have identified a "plasma peptidomic signature"—a specific set of protein fragments—that reflects active liver tissue remodeling in patients with alcohol-related hepatitis (AH). By applying causal graphical modeling to these fragments, the team developed a three-peptide assay that predicts 90-day mortality more accurately than the current clinical gold standard, the MELD score.

What Problem Does This Paper Address?

Alcohol-related hepatitis (AH) is a severe, inflammatory liver condition with a staggering 90-day mortality rate of 30–50%. Clinicians currently rely on the Model for End-Stage Liver Disease (MELD) score to predict patient outcomes and guide treatment. However, MELD is primarily based on systemic markers (like bilirubin and creatinine) and often fails to accurately stratify risk in patients with "moderate" disease.

The authors hypothesize that the liver undergoes intense extracellular matrix (ECM) remodeling during AH. This process releases specific peptide fragments—the "degradome"—into the bloodstream. By capturing this molecular "exhaust," we can gain a real-time view of liver injury and find more precise, non-invasive biomarkers for prognosis.

Methods

The study utilized a multi-omic approach to map the AH peptidome:

  • Peptidomics: The team used high-resolution, high-mass-accuracy UPLC-MS (Liquid Chromatography-Mass Spectrometry) to analyze EDTA plasma. Unlike standard proteomics, which looks at whole proteins, peptidomics focuses on naturally occurring fragments generated by protease activity in the body.
  • Bioinformatics Pipeline: They employed PEAKS X Pro for peptide identification and the MatrisomeAnalyzer to categorize fragments derived from the ECM.
  • Protease Prediction: Using Proteasix, the researchers back-calculated which enzymes (proteases) likely created the observed fragments.
  • Causal Graphical Modeling (PGM): To move beyond simple correlation, the authors used Probabilistic Graphical Models to identify peptides directly linked to 90-day mortality (within the "Markov blanket" of the outcome variable). This helped them prune 1,600+ features down to a robust 3-peptide signature.

Datasets

  • Primary Cohort: 70 individuals total, including 63 AH patients (stratified into moderate and severe based on MELD) and 7 healthy controls.
  • Validation Data: Survival data was available for 58 of the AH patients for the causal modeling.
  • Orthogonal Validation: Publicly available bulk RNA-seq data from AH patient livers was used to confirm the expression of the proteases predicted by the peptidomic data.

Key Insights and Ideas

  • The Matrisome Dominates: Over 60% of the peptide signal in AH patients came from the "matrisome" (the ECM and associated proteins). This confirms that AH isn't just about cell death; it's about a massive, systemic restructuring of liver architecture.
  • The Fibrinogen-to-Collagen Shift: As AH severity increases, there is a distinct shift in the plasma. Moderate cases show higher levels of fibrinogen A fragments, while severe cases are dominated by collagen 1A1 and 1A2 fragments, likely reflecting advanced fibrosis and active tissue breakdown.
  • Identifying the "Executioners": The study identified CAPN1 (Calpain-1) and MMP12 (Matrix Metallopeptidase 12) as the dominant proteases driving this degradome. This provides new mechanistic targets for potential therapeutic intervention.
  • Causal vs. Correlative: By using PGMs, the authors identified three specific peptides—APOC1, CALD1, and BIN2—that provide independent prognostic information. Interestingly, CALD1 and BIN2 are cytoskeletal proteins not previously linked to AH prognosis.

Key Results

  • Superior Accuracy: The 3-peptide signature (Model 2) and a composite model (MELD + 3 peptides; Model 3) both outperformed the MELD-only model.
  • Quantitative Gains: In a 5-fold cross-validation, the composite model achieved an average accuracy of 0.8055, compared to 0.7444 for MELD alone.
  • Sensitivity/Specificity: The peptidomic models were significantly better at identifying patients who would survive ("Alive" class), with the composite model reaching a sensitivity of 80% in validation sets.
  • Survival Curves: Kaplan-Meier plots showed that the peptidomic signature provided a much clearer separation between high-risk and low-risk groups than MELD.

Limitations and Future Work

The authors acknowledge several hurdles before clinical adoption:

  • Sample Size: The cohort (n=63 AH patients) is relatively small and imbalanced compared to healthy controls. Larger, multi-center prospective studies are needed to validate the signature.
  • Matrix Effects: The study specifically used K3EDTA plasma. Because certain peptides (like vimentin) have affinities for heparin, these results might not translate directly to studies using heparinized blood.
  • Site Effects: While samples came from multiple institutions, the potential for "clinical site effects" on the peptidome remains a factor to be explored.
  • Mechanistic Deep-Dive: The roles of CALD1 and BIN2 in AH pathophysiology are currently unknown and warrant further biological investigation.

Reference: Sayed, K., et al. (2023). A plasma peptidomic signature reveals extracellular matrix remodeling and predicts prognosis in alcohol-related hepatitis. medRxiv preprint.

Summary updated

Abstract

We used degraded proteins found the blood of alcohol-related hepatitis patients to identify new potential mechanisms of injury and to predict 90 day mortality.
BibTeX
@misc{sayed2023plasma,
  author = {Khaled Sayed and Christine E. Dolin and Daniel W. Wilkey and Jiang Li and Toshifumi Sato and Juliane I. Beier and Josepmaria Argemí and Ramón Bataller and Abdus S. Wahed and Michael L. Merchant and Panayiotis V. Benos and Gavin E. Arteel},
  title = {{A plasma peptidomic signature reveals extracellular matrix remodeling and predicts prognosis in alcohol-related hepatitis}},
  year = {2023},
  doi = {10.1101/2023.12.13.23299905},
  howpublished = {medRxiv},
}
Figure