Topology-Driven Discovery of Transmembrane Protein S -Palmitoylation

Michael T. Forrester, Jacob R. Egol, Sinan Ozbay, Rohit Singh, Purushothama Rao TatabioRxiv 2024 · DOI

TL;DR

This paper introduces TopoPalmTree, a novel machine learning tool that leverages protein topology to predict S-palmitoylation sites on transmembrane proteins (TMPs). By overcoming the limitations of traditional proteomics, TopoPalmTree successfully identifies novel S-palmitoylation sites and even enables the rational design of new ones, offering a powerful new approach for studying this crucial post-translational modification.

What Problem Does This Paper Address?

S-palmitoylation is a reversible lipid modification that plays a critical role in the function of transmembrane proteins (TMPs), influencing everything from signaling pathways to viral infections. While many studies have observed S-palmitoylation occurring at cysteine residues near the membrane-cytoplasmic interface, large-scale discovery efforts often miss these sites. This is because standard proteomics techniques, which rely on trypsin digestion, struggle to efficiently analyze the hydrophobic juxtamembrane regions where these cysteines are located. This gap in our ability to comprehensively identify S-palmitoylation sites on TMPs limits our understanding of their biological roles.

Methods

The core of this work is the development of TopoPalmTree, a gradient boosted tree (GBT) machine learning model. Unlike previous methods that primarily relied on sequence information, TopoPalmTree incorporates topological features derived from UniProt annotations. These features describe the location and environment of cysteine residues within the protein, such as their proximity to transmembrane domains and their local hydrophobicity. The GBT algorithm was chosen for its interpretability and ability to handle complex datasets. The model was trained on a curated dataset of known S-palmitoylated sites and then validated on an independent dataset of viral TMPs.

Datasets

  • Training Dataset: A curated dataset of 446 S-palmitoylation sites was compiled, primarily from site-directed mutagenesis and ³H-palmitate radiolabeling experiments. This dataset was augmented with data from LC-MS and included topological information from UniProt.
  • Viral Holdout Dataset: A dataset of 36 viral TMPs (82 S-palmitoyl sites) was used for validation. This dataset was chosen for its lack of sequence homology to the training data, providing a rigorous test of the model's generalizability.
  • Murine TMP Proteome: The entire murine transmembrane proteome (5009 TMPs) was analyzed to identify novel putative S-palmitoylation sites.

Key Insights and Ideas

  • Juxtamembrane sites are underrepresented in traditional proteomics: The study highlights that a significant portion of known S-palmitoylation sites on TMPs are located in hydrophobic, juxtamembrane regions that are difficult to analyze with standard trypsin-based methods.
  • Topology is key for TMP S-palmitoylation: The research strongly suggests that the location and environment of cysteine residues within the membrane, rather than just their sequence, are critical determinants of S-palmitoylation on TMPs.
  • Machine learning can bridge the experimental gap: TopoPalmTree demonstrates that machine learning, by incorporating topological features, can effectively predict S-palmitoylation sites that are missed by experimental techniques.
  • The model enables rational design: Beyond prediction, TopoPalmTree can be used to design new S-palmitoylation sites into proteins, opening up avenues for functional studies.

Key Results

  • Novel site discovery: TopoPalmTree identified 1591 novel putative S-palmitoyl sites in the murine TMP proteome.
  • Experimental validation: Two predicted sites on Vamp5 (Cys69) and Aquaporin-5 (Cys6) were experimentally confirmed as S-palmitoylated using Acyl-RAC assays. These sites were previously unreported, underscoring the model's ability to find "missed" sites.
  • Rational design success: A new S-palmitoylation site was successfully engineered into KDEL Receptor 2 (KdelR2) at position 200, as confirmed by Acyl-RAC.
  • Performance comparison: TopoPalmTree outperformed the widely used GPS-Palm tool in predicting S-palmitoylation sites on viral TMPs, showing higher and more consistent Matthews Correlation Coefficient (MCC).

Limitations and Future Work

  • TMP-specific: TopoPalmTree is currently limited to transmembrane proteins due to its reliance on topological features. It is not applicable to soluble proteins.
  • Data dependency: The model's performance is dependent on the availability of topological annotations in databases like UniProt. A significant portion of TMPs still lack complete annotation.
  • Future Directions: The authors suggest expanding TopoPalmTree by integrating it with chemoproteomic data or other post-translational modification (PTM) datasets to understand cross-talk. Further efforts to improve topological annotation of TMPs would also enhance the model's reach.

Summary updated

Abstract

Protein <i>S</i>-palmitoylation is a reversible lipophilic posttranslational modification regulating a diverse number of signaling pathways. Within transmembrane proteins (TMPs), <i>S</i>-palmitoylation is implicated in conditions from inflammatory disorders to respiratory viral infections. Many small-scale experiments have observed <i>S</i>-palmitoylation at juxtamembrane Cys residues. However, most large-scale <i>S</i>-palmitoyl discovery efforts rely on trypsin-based proteomics within which hydrophobic juxtamembrane regions are likely underrepresented. Machine learning- by virtue of its freedom from experimental constraints - is particularly well suited to address this discovery gap surrounding TMP <i>S</i>-palmitoylation. Utilizing a UniProt-derived feature set, a gradient boosted machine learning tool (TopoPalmTree) was constructed and applied to a holdout dataset of viral <i>S</i>-palmitoylated proteins. Upon application to the mouse TMP proteome, 1591 putative <i>S</i>-palmitoyl sites (i.e. not listed in SwissPalm or UniProt) were identified. Two lung-expressed <i>S</i>-palmitoyl candidates (synaptobrevin Vamp5 and water channel Aquaporin-5) were experimentally assessed. Finally, TopoPalmTree was used for rational design of an <i>S</i>-palmitoyl site on KDEL-Receptor 2. This readily interpretable model aligns the innumerable small-scale experiments observing juxtamembrane <i>S</i>-palmitoylation into a proteomic tool for TMP <i>S</i>-palmitoyl discovery and design, thus facilitating future investigations of this important modification.
BibTeX
@misc{forrester2024topology,
  author = {Michael T. Forrester and Jacob R. Egol and Sinan Ozbay and Rohit Singh and Purushothama Rao Tata},
  title = {{Topology-Driven Discovery of Transmembrane Protein <i>S</i> -Palmitoylation}},
  year = {2024},
  doi = {10.1101/2024.09.08.611865},
  howpublished = {bioRxiv},
}

Loading related content...

Figure