An Integrated Approach for Discovering Noncanonical MHC-I Peptides Encoded by Small Open Reading Frames

Lei Chen, Yuanliang Zhang, Ying Yang, Huihui Li, Xuan Dong, Hongwei Wang, Zhi Xie, Qian ZhaoJournal of the American Society for Mass Spectrometry 2021 · DOI

Abstract

MHC-I peptides are a group of important immunopeptides presented by major histocompatibility complex (MHC) on the cell surface for immune recognition. The majority of reported MHC-I peptides are derived from protein coding sequences, and noncanonical peptides translated from small open reading frames (sORF) are largely unknown due to the lack of accurate and sensitive detection methods. Herein we report an efficient approach that implements complementary bioinformatic strategies to improve the identification of noncanonical MHC-I peptides. In a database search strategy, noncanonical immunopeptides mapping was optimized by combining three complementary pipelines to construct predicted sORF databases from Ribo-seq data. In a de novo peptide sequencing strategy, MS data search results were filtered against sORF databases to pin down additional noncanonical immunopeptides. In total, 308 noncanonical immunopeptides were identified from two tumor cell lines with selected ones vigorously validated. Our approach is a handy solution to identify noncanonical MHC peptides with Ribo-seq and MS data. Meanwhile, the novel noncanonical immunopeptides identified with this method could shed insights on fundamental immunology as well as cancer immunotherapies.

Introduction

The immune system's ability to distinguish between "self" and "non-self" relies heavily on Major Histocompatibility Complex class I (MHC-I) molecules. These molecules present small protein fragments, or peptides, on the cell surface for inspection by T-cells. While traditional research has focused on peptides derived from the 1.5% of the human genome that officially codes for proteins, a significant portion of the "non-coding" genome is transcribed and translated into small peptides via Small Open Reading Frames (sORFs). These non-canonical peptides represent a potential goldmine for cancer neoantigens but remain largely invisible to standard proteomic methods due to their absence from reference databases and low abundance.

Methodology

To address this "blind spot," researchers developed an integrated pipeline combining ribosome profiling (Ribo-seq) with mass spectrometry (MS). The approach utilizes two primary strategies:

  1. Custom Database Construction: Ribo-seq data from HCT116 (colorectal carcinoma) and Jurkat (acute T-cell leukemia) cell lines were processed using three complementary pipelines—RiboTISH, PRICE, and sORF.org—to predict active sORFs and build cell-line-specific databases.
  2. Integrated MS Search: MS data were analyzed through a dual-pathway search. A standard database search matched signals against the custom sORF library, while de novo sequencing using the PEAKS algorithm allowed for the identification of peptides without a reference. These de novo results were subsequently filtered against the sORF database to ensure genomic validity.
Integrated workflow for identifying non-canonical MHC-I peptides using Ribo-seq and MS data.

Results

The integrated pipeline successfully identified 308 unique non-canonical MHC-I peptides across the two tumor cell lines. Interestingly, there was minimal overlap between the different sORF prediction tools, underscoring the necessity of an integrated algorithmic approach. Mapping the source of these peptides revealed they primarily originate from long non-coding RNAs (lncRNAs) and 5’ untranslated regions (UTRs) of messenger RNA.

Despite their non-canonical origins, these peptides exhibited biochemical properties—such as length, hydrophobicity, and binding motifs—strikingly similar to traditional canonical peptides. Analysis showed that over 90% of the identified non-canonical peptides were predicted to be strong MHC binders (using a stringent de novo score cutoff of ≥85). Experimental validation using synthetic standards confirmed six out of eight selected novel peptides, including the high-affinity binder ETDIEMETRY.

Venn diagram showing the overlap of canonical and non-canonical peptides, and bar charts illustrating the percentage of predicted MHC-I binders.

Discussion and Conclusion

This study demonstrates that "non-coding" regions are biologically active and contribute significantly to the immunopeptidome. By using custom sORF databases as a post-sequencing filter for de novo MS data, the researchers effectively reduced noise and expanded the scope of peptide discovery.

However, limitations remain. A portion of de novo identified peptides still cannot be mapped to known genomic sources, indicating that current sORF databases are not yet exhaustive. Future research will likely focus on incorporating machine learning to refine de novo sequencing accuracy and investigating the therapeutic potential of these non-canonical peptides in cancer vaccines. This integrated approach provides a robust framework for uncovering the hidden landscape of the immunopeptidome, offering new targets for immunotherapy.

Summary updated

Abstract

MHC-I peptides are a group of important immunopeptides presented by major histocompatibility complex (MHC) on the cell surface for immune recognition. The majority of reported MHC-I peptides are derived from protein coding sequences, and noncanonical peptides translated from small open reading frames (sORF) are largely unknown due to the lack of accurate and sensitive detection methods. Herein we report an efficient approach that implements complementary bioinformatic strategies to improve the identification of noncanonical MHC-I peptides. In a database search strategy, noncanonical immunopeptides mapping was optimized by combining three complementary pipelines to construct predicted sORF databases from Ribo-seq data. In a <i>de novo</i> peptide sequencing strategy, MS data search results were filtered against sORF databases to pin down additional noncanonical immunopeptides. In total, 308 noncanonical immunopeptides were identified from two tumor cell lines with selected ones vigorously validated. Our approach is a handy solution to identify noncanonical MHC peptides with Ribo-seq and MS data. Meanwhile, the novel noncanonical immunopeptides identified with this method could shed insights on fundamental immunology as well as cancer immunotherapies.
BibTeX
@article{chen2021integrated,
  author = {Lei Chen and Yuanliang Zhang and Ying Yang and Yang Yang and Huihui Li and Xuan Dong and Hongwei Wang and Zhi Xie and Qian Zhao},
  title = {{An Integrated Approach for Discovering Noncanonical MHC-I Peptides Encoded by Small Open Reading Frames}},
  journal = {Journal of the American Society for Mass Spectrometry},
  year = {2021},
  doi = {10.1021/jasms.1c00076},
}
Figure