A team led by researchers at Johns Hopkins University has constructed a high-resolution human phosphorylation network identifying 3,656 kinase-substrate reactions.
Detailed in a paper published last week in Molecular Systems Biology, the study represents one of the largest such networks developed to date and could prove a broadly useful tool for protein signaling research, Heng Zhu, a Johns Hopkins researcher and one of the leaders of the effort, told ProteoMonitor.
The work has commercial potential, as well, Zhu said, noting that he and his co-author and Johns Hopkins colleague Jiang Qian have signed a contract through the university with a large pharmaceutical firm to apply the methods used in the study to investigate substrates of certain kinases. Other pharma firms have also expressed interest, he noted.
The research is based on protein microarray technology developed by Zhu's lab that enabled the scientists to screen 289 individual kinases against 4,191 full-length proteins representing 12 major protein families. Using this approach they built a network connecting 230 kinases to 2,591 specific in vivo phosphorylation sites on 652 different proteins.
In addition to establishing the network, the researchers used it to explore B-cell receptor signaling, identifying a new role for protein kinase A in this system.
Heng's protein microarray platform is based on work he did while a postdoc at Yale University in the lab of researcher Michael Snyder, who is now at Stanford University. At that time, he and his colleagues in Snyder's lab used the platform for similar phosphorylation mapping work, looking at interactions between 87 yeast kinases and roughly 5,800 yeast proteins.
The technology has matured significantly since that time, however, Heng noted, citing in particular the bioinformatics work led by Qian, which, he said, was key to weeding out false positives inherent in such research.
An in vitro system like a protein microarray tests only whether two proteins interact when placed together. In vivo, however, such interactions are determined and limited by a number of factors including the proteins' cellular localization and the interplay of protein complexes. This leads to a high probability of false positives in interaction data generated by in vitro assays as, for instance, a kinase and a substrate that appear to interact on a microarray might never actually be present in the same part of a cell in vivo.
To identify and eliminate such false positives, Qian and his colleagues developed bioinformatics approaches that took account of biological data that could help them distinguish between true in vivo interactions and interactions likely to occur only in the microarray context.
"We included [information] like cellular localization, tissue specificity, [and] protein-protein interactions to try to enrich for the physiological relevance of the [observed] interactions," he told ProteoMonitor, adding that he believed that given this bioinformatic analysis, the rate of false positives in the data "should be very low."
False negatives, on the other hand, are likely higher, Qian said, given that the microarray does not offer various aspects of the in vivo environment, such as scaffold proteins that could be necessary for certain kinase-substrate interactions. Indeed, the researchers were unable "to recover a lot of known [kinase-substrate] interactions," he said.
Despite this limitation, Heng said he believed the network was the largest such developed to date for human phosphorylation.
Following identification of the kinase-substrate interactions, the researchers then performed three levels of validation. First, they examined whether these observed interactions took place in cellular systems, measuring kinase-dependent changes in the target proteins. In this work, Heng said, they validated roughly 75 percent of the interactions identified via their microarray data.
Second, they used phospho-specific antibodies to look at phosphorylation on the substrates upon interaction with a given kinase, validating roughly 80 percent of the interactions via this method.
Finally, they performed site-directed mutagenesis on a subset of phosphosites identified via microarray to see if this would eliminate the observed kinase-substrate interaction.
"In each of the cases that we tested, the mutation of the phosphorylation site abolished the kinase-dependent changes in those cells, so we think that at the end our data set is high quality," Heng said.
This ability to isolate kinase-substrate interactions at the level of specific phosphosites marks another advance over his previous efforts in yeast, Heng noted.
"The mass spec data [on phosphosites] was not available back then, but now there are so many people working [on phosphoproteomics] in humans that we were able to collect enough mass spec-identified sites to construct this high-resolution map," he said.
The researchers, Jiang said, have developed a database housing the results of the work that other groups can access to aid in their signaling work.
"Basically, people in the signaling field can use this database and look at their kinases of interest – their favorite kinases or substrates – and our data will provide clues that they can experimentally validate or [research] further," he said.
With Jin Zhang, a Johns Hopkins researcher and co-author on the paper, the team demonstrated such an application, using the network data to demonstrate that in B-cell receptor signaling the kinase PKA serves as an intermediary between the protein Bruton's tyrosine kinase, Btk, and the protein ARID3A.
"People have known that in B-cell receptor activation Btk is activated and then it recruits a transcription factor called ARID3A, and then that somehow gets phosphorylated," Heng said.
It was unknown, however, how Btk and ARID3A interacted given that Btk is a tyrosine kinase and the ARID3A phosphosites are not tyrosines.
"So, looking at our network, we identified PKA as a substrate of Btk and also as [sitting] upstream of a kinase that can phosphorylate ARID3A," Heng said. "PKA is a potential missing linking that connects Btk to ARID3A."
He noted that since they have begun presenting the work, other researchers have begun to approach them asking "if we can do this kinase reaction using their favorite kinases."
The team plans to expand the scope of the network to add data on some of the roughly 200 kinases not included in the MSB paper. They have expanded the number of proteins in their arrays, as well, Heng noted, raising the number from 4,196 to around 17,000.
Heng said he plans to offer kinase research services based on the MSB work through CDI Laboratories, a firm he launched several years ago to commercialize the protein microarray technology.