NEW YORK (GenomeWeb) – A team led by researchers at Dana-Farber Cancer Institute and the University of Toronto has developed a DNA barcoding-based method for high-throughput mapping of protein-protein interactions.
Detailed in a paper published last week in Molecular Systems Biology, the method is roughly tenfold cheaper than the conventional yeast two-hybrid screens used in protein interaction research, Frederick Roth, a University of Toronto researcher and senior author on the study, told GenomeWeb.
In the MSB paper, the researchers used the method to test interactions between roughly 2.5 million protein pairs. They are currently working to use the method to map the entire yeast interactome, which would involve testing some 36 million protein pairs, Roth said.
Yeast two-hybrid screens are among the most common approaches for studying protein-protein interactions. Essentially, the method screens "bait" proteins against "prey" proteins by fusing the former to the DNA-binding domain of the transcription factor Gal4 and the latter to Gal4's activation domain. When a "bait" and "prey" protein interact, the two portions of Gal4 come together to create a functional protein that allows the strain to grow.
Initially, the method was used to screen single bait proteins against single prey proteins. More recently, it has become possible to screen single baits against hundreds of preys at a time. The readout of these interactions has been quite tedious, however, Roth noted. For each pair, researchers must isolate the resulting colony and then perform various PCR or sequencing assays to determine which two proteins were involved.
"So, you have to pick the colonies, and it's still kind of old school," he said. "You go in with a toothpick, and you pick the colony."
In the method presented in the MSB paper, termed Barcode Fusion Genetics-Yeast Two-Hybrid (BFG-Y2H), Roth and his colleagues used Cre recombination to fuse DNA barcodes linked to interaction pairs, meaning that each interaction generated a specific barcode fusion that could then be quantified via next-generation sequencing to identify the proteins involved.
This allows the researchers to simultaneously screen multiple bait proteins against multiple prey proteins. Interactions are still signaled by Gal4-enabled growth, meaning that readout still involved plating and incubating of the strains to see which interactions took place. But, Roth said, instead of individually picking each colony for analysis, readout can be done in bulk.
"To figure out what is there, you basically scrape the plates and then sequence these now fused barcodes," he said.
The approach is an example of next-generation sequencing's ability to transform traditional assays, Roth said. "Because the sequencing technology is so powerful, a lot of the advances in technology lately has been, 'How do you adapt the assay you used to do before so that sequencing is the readout?'"
In the case of the BFG-Y2H method, an NGS-based approach means a tenfold reduction in the cost of protein-protein interaction assays compared to current methods, Roth said, adding that he believes the performance will improve even more with future developments that could eliminate the need for plating the cells prior to NGS analysis.
"The bottleneck right now is plating the cells," he said. "We need to give them a little bit of real estate for the colonies to come up, and I'm not sure that is going to be essential in the future."
Roth declined to go into detail regarding how he and his colleagues hoped to eliminate this requirement, but he said he expected they would manage it in the relatively near future.
The approach does suffer in comparison to conventional methods in terms of sensitivity, Roth said, noting that in the MSB study its sensitivity was roughly 50 percent to 80 percent of standard methods. However, he added, it provides higher confidence results, eliminating the need for the sort of follow-up confirmation that is necessary in the case of some interactions identified using traditional assays.
"In conventional yeast two-hybrid, in the first pass there are still some errors in there, and you have to do follow-up retesting to confirm that the signal is still there," he said. "We actually have better accuracy with the interactions we got such that we wouldn't need to do that sort of pairwise retesting."
The expectation is also that because BFG-Y2H assays are much cheaper, the higher volume enabled by the approach will more than make up for any losses in sensitivity.
"Our hope is that it being 10 times cheaper, if we can afford to run 10 times the number of assays, our net gain will be far more than [what we lose in sensitivity]," Roth said. Though, he added, it had yet to be determined "how many assays we will have to run to sort of claw back that sensitivity."
While previous yeast two-hybrid screens have maxed out at around one bait against 200 prey proteins, Roth and his colleagues managed in their study to screen roughly 2,000 bait against 2,000 prey. Now they are tackling the full yeast interactome, which involves screening around 6,000 bait against 6,000 prey proteins.
They will likely do this as a series of nine 2,000 versus 2,000 protein screens, each of which will take around two weeks, Roth said. This is due not so much to some difficulty inherent in scaling up from a 2,000 by 2,000 screen to a 6,000 by 6,000 screen, he said, but rather to the fact that he doesn't have enough incubator space to plate all the cells that such an experiment would generate.
"The bottleneck is basically plating the cells out, and we have only so many incubators," he said. "So, if I could get more space and more incubators we probably could do a yeast interactome in two weeks."
Ultimately, Roth said he hopes to use the method to map the human interactome. Though, he noted as well that its potential lies not only in mapping larger and larger sets of interactions, but also running assays under many different conditions.
"It may be that there are changes in which [protein] pairs interact if you change the pH of the cell or the temperature or if a drug comes along and the cell has a physiological response to the drug," he said. "So this opens up a dimension of dynamics potentially."
In his own research, Roth said he is particularly interested in looking at how protein interactions change under DNA damage conditions as well as cell cycle-dependent and protein kinase-dependent interactions.