Naiwen Cui

A mix-and-read drop-based in vitro two-hybrid method for screening high-affinity peptide binders


Drop-based microfluidics have recently become a novel tool by providing a stable linkage between phenotype and genotype for high throughput screening. However, use of drop-based microfluidics for screening high-affinity peptide binders has not been demonstrated due to the lack of a sensitive functional assay that can detect single DNA molecules in drops. To address this sensitivity issue, we introduced in vitro two-hybrid system (IVT2H) into microfluidic drops and developed a streamlined mix-and-read drop-IVT2H method to screen a random DNA library. Drop-IVT2H was based on the correlation between the binding affinity of two interacting protein domains and transcriptional activation of a fluorescent reporter. A DNA library encoding potential peptide binders was encapsulated with IVT2H such that single DNA molecules were distributed in individual drops. We validated dropIVT2H by screening a three-random-residue library derived from a high-affinity MDM2 inhibitor PMI. The current drop-IVT2H platform is ideally suited for affinity screening of small-to-medium-sized libraries (10^3–10^6). It can obtain hits within a single day while consuming minimal amounts of reagents. Drop-IVT2H simplifies and accelerates the drop-based microfluidics workflow for screening random DNA libraries, and represents a novel alternative method for protein engineering and in vitro directed protein evolution.

Protein-protein interactions (PPIs) regulate cellular physiology by influencing interactome networks. Between 40,000 and 200,000 PPIs have been predicted to exist within the human interactome, and their malfunction is one of the fundamental causes of human diseases1,2. One promising therapeutic strategy involves the use of peptide drugs with high target-specific affinities to regulate certain PPIs. Compared with protein and small-molecule drugs, therapeutic peptides offer the advantages of better cell penetration, less immunogenicity and greater specificity. More than 60 synthetic therapeutic peptides have recently reached pharmaceutical markets. For example, Degarelix (Firmagon), a gonadotrophin-releasing hormone receptor blocker, has been shown effective for the treatment of men with advanced hormone-sensitive prostate cancer. Currently, molecular display represents the most widely used high-throughput techniques for screening high-affinity peptide binders. Taking the advantage of living cells’ ability to express a DNA library and display the protein or peptide products on their surfaces, in vivo display systems utilize phage, bacterium and yeast to establish a physical link between the binding affinity to a target molecule (phenotype) and the DNA sequence (genotype) of the displayed molecule. However, such in vivo systems often suffer from serious drawbacks, such as low transformation efficiency, expression bias, toxicity of fusion proteins and interference of other surface proteins during selection. In comparison, cell-free display systems, such as ribosome and mRNA display, contain only the essential elements for protein expression and thus provide an in vitro solution to address some of these issues. However, cell-free display systems also have their own drawbacks. For instance, in the ribosome display method, potential protein or peptide binders are expressed from a DNA library and through ribosomes form a linkage with their coding mRNAs. The binders with high affinities towards a target are selected through “biopanning”, a series of washing and amplification cycles. Such selection conditions can destabilize the mRNA-binder complex. The mRNA display method improves the stability by establishing a covalent linkage between the binder and its coding mRNA. Nevertheless, the RNA-binder complexes are inherently unstable which can severely restricts the screening conditions. Moreover, the selection of high-affinity binders using biopanning in both ribosome and mRNA display methods can be biased by the dissociation kinetics of the binder to the target molecule. A recently reported bead display method circumvents some of these drawbacks by displaying binder-DNA conjugates on monoclonal beads, which are subsequently screened using flow cytometer. It is unclear whether the binding to the target molecule is affected by the attachment of the binder to a large heterogeneous bead surface. Drop-based microfluidics has become a novel tool for high throughput screening in recent years. It provides a stable linkage between phenotype and genotype by partitioning single cells into picoliter drops and allows fluorescence-activated drop sorting (FADS). However, the use of such drop-based microfluidic method to screen high-affinity binders from a random DNA library has never been shown, largely due to the fact that there is no functional assay sensitive enough to detect single DNA molecules for protein binding in drops. Here we combined drop-based microfluidics with our recently developed in vitro two-hybrid system (IVT2H) – an in vitro assay for detection of protein-protein interaction, and demonstrated a simple and cost-effective screening platform, which we named “drop-IVT2H”. We encapsulated single DNA molecules of a peptide library in picoliter drops with the IVT2H reagents containing plasmids expressing the target protein. Drops were incubated off-chip to allow the expression of both binder and target proteins. The binding of a high-affinity binder to the target protein activated the GFP expression in situ, resulting in highly fluorescent drops (bright drops). These bright drops were isolated by the FADS device and the high-affinity binders were subsequently identified by DNA sequencing. We demonstrate that this mix-and-read drop-IVT2H has allowed successful enrichment of high-affinity peptide binders in a p53-MDM2 binding model.



Figure 1. A schematic workflow for screening high-affinity binders using drop-based in vitro twohybrid method (drop-IVT2H)

Figure 2. Fluorescence images of drops after off-chip incubation

Figure 3. Histograms of the normalized drop fluorescence of sorted drops containing PMI, p53 or no binder template.

Figure 4. Histograms of the normalized drop fluorescence of sorted drops containing the random PMI library, PMI, its variants (FWR and FSL), or no binder template.

Figure 5. Frequency histogram representation of deep sequencing data from collected drops.