Maximizing Immunopeptidomics-Based Bacterial Epitope Discovery by Multiple Search Engines and Rescoring

Abstract

Mass spectrometry-based discovery of bacterial immunopeptides presented by infected cells allows untargeted discovery of bacterial antigens that can serve as vaccine candidates. However, reliable identification of bacterial epitopes is challenged by their extremely low abundance. Here, we describe an optimized bioinformatic framework to enhance the confident identification of bacterial immunopeptides. Immunopeptidomics data of cell cultures infected with Listeria monocytogenes were searched by four different search engines, PEAKS, Comet, Sage and MSFragger, followed by data-driven rescoring with MS²Rescore. Compared with individual search engine results, this integrated workflow boosted immunopeptide identification by an average of 27% and led to the high-confidence detection of 18 additional bacterial peptides (+27%) matching 15 different Listeria proteins (+36%). Despite the strong agreement between the search engines, a small number of spectra (<1%) had ambiguous matches to multiple peptides and were excluded to ensure high-confidence identifications. Finally, we demonstrate our workflow with sensitive timsTOF SCP data acquisition and find that rescoring, now with inclusion of ion mobility features, identifies 76% more peptides compared to Q Exactive HF acquisition. Together, our results demonstrate how integration of multiple search engine results along with data-driven rescoring maximizes immunopeptide identification, boosting the detection of high-confidence bacterial epitopes for vaccine development.

Keywords: Listeria monocytogenes; TIMS2Rescore; immunopeptides; ion mobility; mass spectrometry; search engines.

Figure 1

Integration of multiple search engines and rescoring boost immunopeptide identification. (A) Listeria-infected and uninfected HeLa or HCT-116 cells were subjected to both label-free LC-MS/MS on a Q Exactive HF instrument as well as TMT-labeling and LC-MS/MS analysis on a Fusion Lumos instrument after prefractionation. Resulting MS data was searched by four search engines in parallel and the search results of each engine were rescored by MS²Rescore v3 independently. All peptide and PSM identifications with a peptide q-value <1% were aggregated and further subjected to quantification, HLA binding prediction and Gibbs clustering. Created with BioRender.com. (B) Rescoring by MS²Rescore boosts the number of identified peptide sequences per search engine and across engines. (C) Annotated MS2 spectrum (HeLa label-free replicate 2, scan 52812) matched to both ‘RAAPLLQLL’ by PEAKS Studio 12 and ‘RAALPLQLL’ by Comet. The experimental spectrum is displayed on top and the MS²PIP-predicted spectrum at the bottom, indicating the Pearson spectrum correlation calculated by MS²Rescore (‘spec_pearson’ feature). (D) Integrating the results of four search engines after rescoring further boosts the immunopeptide detection. The dotted line indicates the total number of unique, identified peptide sequences after rescoring per sample.

Figure 2

Integration of multiple search engines and rescoring yields complementary immunopeptide identifications. (A) Number of unique peptide sequences identified per amino acid length. Peptide sequences predicted as strong binder (SB, %Rank <0.5) or weak binder (WB, %Rank <2) by NetMHCpan-4.1 are indicated in red and blue, respectively. Other peptides (nonbinder, NB) are indicated in gray. (B) Venn diagram showing the overlap of identified peptide sequences predicted as NetMHCpan-4.1 strong binders per search engine. (C) Sequence logo of peptides predicted as strong binder (SB, %Rank <0.5) by NetMHCpan-4.1. Logos were made by Logomaker using the peptide 9-mer binding core predicted by NetMHCpan-4.1.

Figure 3

Detection of additional high confidence Listeria immunopeptides. (A) Filters applied for selecting bacterial immunopeptides, resulting in a final number of 86 high-confidence Listeria peptides. (B) Annotated MS2 spectrum (HCT-116 cells TMT fraction 1, scan 19403) matched to ‘TMT-TIDELAGK_TMTI’ by Sage or matched to ‘TMT-TIDEIQK_TMTL’ by MSFragger, Comet, and PEAKS. The experimental spectrum is displayed on top and the MS²PIP-predicted spectrum at the bottom with their respective Pearson correlation. (C) Venn diagram showing the overlap of high-confidence and filtered out Listeria peptides in this study to the high-confidence peptides described in Mayer et al. (D) Imputed and unimputed intensity heatmaps of high confidence Listeria immunopeptides for all four experimental conditions. Z-scored log₂ intensities were displayed after FlashLFQ label-free quantification (LFQ) and TMT reporter ion intensities outputted by IsobaricAnalyzer. Missing values were imputed by Perseus.

Figure 4

Listeria virulence factors are represented by multiple immunopeptides. (A) The number of unique immunopeptides identified per Listeria protein is shown as a histogram. The number of identified peptides per sample is displayed in a heatmap. (B) Genome view of the Listeria Pathogenicity Island 1 region (NC_025568 sequence, numbers in kilobases) with identified immunopeptides for five out of six genes. The plcB peptide ‘YKLGLAIHY’ was a novel immunopeptide identified in this study (*). (A–B) Immunopeptides predicted as strong binders by NetMHCpan-4.1 (%Rank <0.5) are indicated in red, weak binders (% rank <2) in blue and nonbinders (%Rank >2 or no 8- to 12-mers) in gray.

Figure 5

Improved immunopeptide identification by DDA-PASEF acquisition. (A) MS²PIP timsTOF model performance, displaying a histogram of the Pearson correlation for all PSMs. (B) Correlation of DeepLC-predicted and experimental retention time (RT). (C) Correlation of the IM2Deep-predicted and experimental collision cross section (CCS). (D) TIMS²Rescore rescoring and integration of multisearch engine results boosts immunopeptide identification. (E) Representative peptide ion identifications plotted across the inversed ion mobility (1/K₀) and m/z dimensions. Identified 9-mer peptides are plotted in red, and peptides of other lengths are plotted in blue.