A research team led by scientists at the European Molecular Biology Laboratory has developed a large-scale network of functionally associated protein post-translational modifications.
As detailed in a study published in the current edition of Molecular Systems Biology, the researchers developed a method for measuring PTM co-evolution and used this to predict functional links between 13 frequent PTM types, identifying a number of well-established forms of PTM crosstalk and providing a research tool for further investigations into PTM associations, Peer Bork, senior group leader for bioinformatics at EMBL and an author on the paper, told ProteoMonitor.
To build the network, the researchers collected more than 420,000 experimentally verified PTMs spanning 89 different varieties across 2,485 different species from public data sets. They then preprocessed this data to remove redundancies, leaving 11,149 PTM residues comprising 13 types from eight eukaryotic species.
Taking this data, they developed measures of the conservation status of the PTM residues, which they then used to analyze the co-evolution of pairs of modified residues, using this measure as a proxy for their functional association. This analysis found that 74,386 residues in 10,325 proteins are co-evolving, suggesting that they are functionally linked.
Via this analysis, the team identified a number of known interactions, Bork said, including various PTMs linked to histones and epigenetic processes. The map also predicted a number of more surprising associations, he added, particularly in secreted and membrane-associated proteins.
"For example, between phosphorylation and N-glycosylation," he said. "That at first seems strange because phosphorylation usually occurs in a cell and N-glycosylation outside the cell, but in the Golgi [apparatus], for instance, these things can happen together."
Because the researchers built the network using co-evolution as a proxy for functional linkage, it identifies not only tight physical interactions between PTMs, but more indirect associations, as well, Bork noted.
"We take [the network] as a broad description of functional coupling," he said. "There must be a pretty strong correlation [between the associated PTMs] because co-evolution is about selection, and if it didn't make sense that two residues stuck together, they would be mutated away. So we have a strong argument for [co-evolution as a proxy for] functional coupling, but in a loose sense – not necessarily in [the sense of] physical crosstalk."
"Functional association within a protein can be very indirect – [linked] by another regulator, for example," he said. "It doesn't need to be physical interactions where they sit next to each other [on a protein] and compete with each other."
Bork observed, however, that the analysis did demonstrate that PTMs located close to each other on a protein tend to demonstrate more co-evolution, supporting the notion, he said, that "there is some indication that a number of these associations will be physical ones."
Despite the network's large size, Bork said he believes the PTM interactions his team identified and predicted are just "the tip of the iceberg."
"It's very clear that many PTMs have not been discovered yet," he said, noting that improving mass spectrometry instruments and techniques will continue to add to the ranks of reported PTMs.
The network, Bork said, is intended as a guide for further exploration of these interactions, offering a framework for future investigations.
"My hope is that this brings a new angle to research in the PTM world," he said. "Everyone has talked about the protein [PTM] code for a long time, but mostly in a hypothetical kind of way. What we have here is a network in hand that we can continue to fill in with [experimental] data."
In the paper, the authors provide an example of such an application, using the network to identify a series of phosphorylation sites that are likely linked to regulation of NFκβ by its subunit RelA.
Bork compared the PTM research to large-scale efforts to map protein-protein interactions, which, he said, were driven significantly by bioinformatics work done more than a decade ago.
The effort to develop large protein-protein interaction maps "started at the end of the [1990s], and bioinformaticians made many of the first connections," he said. "And then the experimental data came afterwards. So I hope we see the same thing [in the case of PTM interactions.]"
Moving forward, Bork and his team are pursuing experimental validation of their predictions, collaborating with EMBL senior scientist Anne-Claude Gavin, a co-author on the MSB paper.
Gavin also co-authored with Bork a paper in the March edition of MSB exploring crosstalk between protein phosphorylation and lysine acetylation in Mycoplasma pneumoniae (PM 3/2/2012). That study, which used a dimethyl labeling approach combined with analysis on a Thermo Scientific LTQ Orbitrap, identified a broader-than-thought role for protein phosphorylation in post-transcriptional regulation as well as considerable crosstalk between phosphorylation and acetylation, showing significant linkage between the two modifications.
In addition to validating the predictions made by the PTM network, the researchers are also exploring interactions between PTMs on different proteins, Bork said.
"We're looking at pairs of different proteins with different PTMs – so how phosphorylation on one protein affects acetylation on another protein, for instance," he said, noting that this effort "was a bit more complicated from a statistical point of view."