Quantifying homologous proteins and proteoforms

Dmitry Malioutov, Tianchi Chen, Jacob D. Jaffe, Edoardo M. Airoldi, Steve Carr, Bogdan Budnik, Nikolai SlavovbioRxiv 2017 · DOI

TL;DR

This paper introduces HIquant, a novel computational method that accurately quantifies the relative abundance of protein isoforms (proteoforms) and homologous proteins, even in the presence of complex post-translational modifications. By leveraging shared peptides as internal standards, HIquant overcomes the peptide-specific biases inherent in traditional mass spectrometry approaches, enabling more reliable proteomic analysis without the need for expensive external standards.

What Problem Does This Paper Address?

Understanding the diverse functions of proteins within a cell often requires quantifying different versions of the same protein, known as proteoforms. These proteoforms can arise from alternative splicing, post-translational modifications (PTMs), or even different genes encoding similar proteins (paralogs). Existing methods for quantifying these proteoforms using bottom-up mass spectrometry (MS) are plagued by "peptide-specific biases." These biases stem from variations in how efficiently peptides are generated during sample preparation (e.g., protein digestion) and how well they are detected by the mass spectrometer (e.g., ionization efficiency). These variations mean that the measured abundance of a peptide doesn't always accurately reflect the abundance of the proteoform it came from, making it difficult to determine the true relative amounts of different proteoforms. This is particularly challenging for PTMs, where even small differences in modification can lead to vastly different biological outcomes.

Methods

The core of the paper is the development and implementation of a first-principles model called HIquant. HIquant tackles the problem of peptide-specific biases by modeling peptide measurements as a superposition of protein levels, scaled by unknown peptide-specific biases (nuisances). The key innovation is the use of shared peptides – peptides that are common to multiple proteoforms or homologous proteins – as internal standards. These shared peptides act as anchors, coupling the measurements from different peptides and allowing the model to disentangle the true protein abundance from the peptide-specific biases.

The model is formulated as a system of linear equations that can be solved using convex optimization techniques. HIquant infers both the protein levels and the peptide-specific biases simultaneously. The paper also characterizes the conditions under which HIquant can find a unique and optimal solution, even when unique peptides are scarce.

Datasets

The paper evaluates HIquant using two main experimental setups:

  1. Spiked-in proteoform mixtures: A standard protein mixture (UPS2) was chemically modified (alkylated) to create different proteoforms. These were then mixed in known ratios and spiked into yeast lysate. This allowed for a direct comparison between the known mixing ratios and the ratios inferred by HIquant.
  2. Endogenous histone H3 PTMs: The study also applied HIquant to quantify naturally occurring post-translational modifications (methylation and acetylation) on histone H3 peptides. These results were compared to estimates obtained using previously established external standards (MasterMix).

Key Insights and Ideas

  • Shared peptides are powerful internal standards: The paper demonstrates that peptides common to multiple proteoforms can effectively act as internal standards, allowing for the deconvolution of peptide-specific biases from true proteoform abundance.
  • Decoupling protein abundance from peptide detection: HIquant successfully separates the signal originating from the actual abundance of a protein from the noise introduced by variations in sample processing and MS detection.
  • Accurate quantification without external standards: A major breakthrough is the ability of HIquant to achieve high accuracy in quantifying proteoform stoichiometries and PTM occupancies without relying on expensive and often difficult-to-prepare external standards.
  • Broad applicability: While demonstrated for PTMs and paralogs, the underlying model is generalizable to any set of proteins that share peptides, suggesting wide applicability in proteomics.

Key Results

  • High accuracy in proteoform ratio inference: For the spiked-in proteoform mixtures, HIquant achieved a median error below 10% for inferring proteoform ratios, significantly outperforming methods relying solely on unique peptide ratios.
  • Accurate PTM site occupancy estimation: HIquant accurately inferred fractional site occupancies for histone H3 modifications, showing good agreement with estimates derived from external standards.
  • Robustness to biases: The method demonstrated robustness to systematic biases like incomplete digestion and co-isolation interference, as these are absorbed into the inferred nuisance parameters.

Limitations and Future Work

  • Dependence on relative quantification accuracy: HIquant's performance is contingent on the accuracy of the underlying relative peptide quantification. While technological advancements are mitigating this, extreme differences in abundance (over 3-6 orders of magnitude) might still be better inferred from unique peptides.
  • Handling very low abundance proteoforms: Proteoforms with abundances significantly lower (over 1000-fold) than main proteoforms, and their corresponding peptides, were omitted from the inference due to the unrealistically high accuracy required for their quantification. Future work could explore methods to better handle these low-abundance species.
  • Potential for further generalization: The authors suggest that the general form of the HIquant model is not limited to proteoforms and could be applied to any set of proteins sharing peptides, opening avenues for broader applications in comparative proteomics.

Summary updated

Abstract

Many proteoforms – arising from alternative splicing, post-translational modifications (PTMs), or paralogous genes – have distinct biological functions, such as histone PTM proteoforms. However, their quantification by existing bottom-up mass–spectrometry (MS) methods is undermined by peptide-specific biases. To avoid these biases, we developed and implemented a first-principles model (HI quant ) for quantifying proteoform stoichiometries. We characterized when MS data allow inferring proteoform stoichiometries by HI quant , derived an algorithm for optimal inference, and demonstrated experimentally high accuracy in quantifying fractional PTM occupancy without using external standards, even in the challenging case of the histone modification code. A HI quant server is implemented at: https://web.northeastern.edu/slavov/2014_HIquant/
BibTeX
@misc{malioutov2017quantifying,
  author = {Dmitry Malioutov and Tianchi Chen and Jacob D. Jaffe and Edoardo M. Airoldi and Steve Carr and Bogdan Budnik and Nikolai Slavov},
  title = {{Quantifying homologous proteins and proteoforms}},
  year = {2017},
  doi = {10.1101/168765},
  howpublished = {bioRxiv},
}
Figure