Developing a massively parallel yeast two-hybrid assay to characterize thousands to millions of pairwise protein-protein interactions in a single pot

Abstract

The interaction between proteins drives nearly all important processes inside living cells. The mechanisms underlying protein-protein interactions remain not fully understood, which necessitates experimental verification of functional de novo designed binders for the purposes of synthetic biology and therapeutics. This, in turn, necessitates a technology to measure lots of protein-protein interactions in parallel. While existing massively parallel approaches have already been implemented to generate interactome-size datasets, they are still either labor intensive or too sophisticated and expensive to be implemented by most labs. We have developed and extensively characterized an easy-to-use high-throughput method to measurebinary protein-protein interactions (PPIs). In our approach, we can measure all pairwise interactions between a library of proteins in a single one-pot experiment. We make use of the conventional yeast two-hybrid assay’s genetic selection principle in which the expression level of a growth-essential gene is determined by the strength of interaction between two proteins of interest; one protein is fused to the DNA binding domain and the other one is fused to the transcriptional activation domain of a split transcription factor driving the expression of growth- essential gene and thus a growth selection can be used to differentiate stronger interacting protein pairs from weaker ones. We construct a population of yeast cells where in each cell only one distinct pair of proteins from a protein library is expressed in the form of fusions to the binding and activation domains. This is achieved by expressing both fusion proteins from the same yeast centromere plasmid maintained at a single copy inside each yeast cell. This cell population is then subjected to the growth selection to differentiate cells with stronger interacting proteins from the weaker ones. By integrating unique DNA barcodes into distinct plasmids and comparing the abundance of each distinct barcode before and after the selection process by means of next-generation sequencing, a fitness score for every protein pair is calculated and used as a proxy for the interaction strength. We show that measured fitness scores exhibit log-linear correlation with independently measured dissociation constants reported in literature, ranging from R2=0.55 for a set pro-survival BCL2 protein family and their de novo designed inhibitors to R2=0.95 for a set of parallel heterodimeric alpha-helical coiled coils with variable length.