Fine tuning of ddaPASEF methods improves the immunopeptidomics coverage and throughput, which is further boosted by ML-assisted rescoring

INTRODUCTION

Identifying ligands of the human leukocyte antigen (HLA), known as immunopeptides, is vital for developing vaccines and immunotherapies. LC-MS-based immunopeptidomics can directly identify and quantify HLA peptides. However, this approach faces significant challenges, requiring optimization at each step of the workflow, from sample preparation through MS analysis to data treatment. Throughout, the non-tryptic nature of HLA adds specific complications.

We optimized the MS acquisition method on the timsTOF Pro instrument but also implemented it on the timsTOF Ultra to achieve highest possible detection sensitivity and throughput. Additionally, we integrated MS²Rescore into the data analysis, which uses AI and machine learning-based peptide fragmentation (MS²PIP) and retention time (DeepLC) predictions to rescore search engine results.

METHODS

MS method optimizations were performed on a nanoElute timsTOF Pro and Ultra system (Bruker) with immunopeptides from 5.10e8 HL60 cells. Key acquisition parameters evaluated included ion mobility range (0.6 to 1.75 Vs/cm²), accumulation time (100 to 200 ms), collision energy (CE) ramp (10 to 55 eV and 20 to 59 eV), number of PASEF scans (6 or 10), and polygon filters (standard or HLA-tailored). We tested 5 cm and 25 cm Aurora3 columns (IonOptiks) with gradients ranging from 10 to 100 minutes. Peptides were identified using Proline or Sage software and rescored with MS²Rescore.

RESULTS

We benchmarked eight different MS methods using ddaPASEF acquisition on a timsTOF Pro to finely tune a method dedicated to non-tryptic immunopeptides. Crucial MS acquisition parameters were evaluated, including ion mobility, collision energy ranges, accumulation time, number of PASEF scans and tailored polygon filter. The optimized parameters increased identified immunopeptides from 4548 to 5793 compared to the classical method. Implementing Sage and MS²Rescore further boosted the number of identified peptides to 10257, a 77% increase. An HLA-tailored polygon filter improved the total number of identified peptides by an additional 10%, allowing the identification of singly charged peptides (14%), which had previously been discarded.

When applied to the timsTOF Ultra, similar performance could be achieved with only half the injection amount and only one-fifth of the gradient time. A 100-minute gradient using 25 cm Aurora3 columns identified as many as 19,376 peptides, showing the highest level of performance. More than 5000 peptides were still detected using a shorter 5 cm column in 10 minute gradient, demonstrating great potential for increased throughput.

CONCLUSION

Our results demonstrate a significant improvement in the immunopeptidomics workflow, achieving unprecedented performance with the latest timsTOF Ultra. Various gradient tests showed potential for increased throughput while maintaining adequate coverage. Overall, the enhanced coverage and depth of analysis are promising for detecting neo-antigens in cancer tissues.