Introduction

Most biochemical processes are mediated through specific interactions between proteins, and a primary goal of functional genomics is to elucidate complete protein interaction networks. One consequence of this focus is the potential for development of protein interaction antagonists that could target highly specific biochemical pathways. The diversity of small molecule protein interaction inhibitors that have been described has expanded considerably over the past several years (1,2), lending support to the feasibility of pursuing this strategy for drug discovery. Consequently, for this purpose, a variety of in vitro biochemical (3,4), as well as bacterial (5), yeast (6,7,8), and mammalian cell (9,10) based genetic strategies have been designed to enable high-throughput screening of small molecule compound libraries for protein interaction inhibitors. The majority of yeast-based strategies are derived from variations of the two-hybrid and interaction trap systems in which interaction between DNA binding domain bait and activation domain prey fusions causes activation of a reporter gene (6,7). Conversely, disruption of the bait-prey fusion interaction by an antagonist prevents recruitment of the activation domain prey and thereby causes loss of reporter gene expression. One inherent difficulty when using these systems for screening small molecule libraries is that hits representing potential interaction inhibitors produce a negative result (loss of signal), which need to be distinguished from effects due to toxicity or nonspecific inhibition of gene activation.

We have previously described a modified yeast two-hybrid strategy for use with transcriptional activator bait proteins, designated the repressed transactivator (RTA) system (11). Proteins that activate transcription typically cannot be used as bait fusions in the two-hybrid system, because they cause reporter gene expression in the absence of an interacting prey fusion protein. In the RTA system, the bait fusion protein must be capable of activating transcription, while the prey proteins are fused to the N-terminal transcriptional repression domain (RD) of Tup1, a general repressor of the budding yeast Saccharomyces (see Figure 1). In the absence of an interacting prey, activation of a URA3 reporter gene by the bait renders yeast sensitive to 5-fluoroorotic acid (5-FOA), which is converted to a toxic metabolite in a pathway dependent upon orotidine-5′-phosphate (OMP) decarboxylase, the catalytic product of URA3 (see Figure 1A). Interaction between the activator bait and a Tup1 RD prey fusion causes repression of the URA3 reporter gene, thereby enabling growth on 5-FOA (see Figure 1B). This strategy has been used to identify novel interactions with a variety of transcriptional activators, including herpes simplex virus 1 (HSV-1) regulatory protein VP16 (11), c-myc (12), and the androgen receptor (13).

Figure 1. Representation of the repressed transactivator (RTA) system and its use for identification of protein interaction inhibitors. (Click to enlarge)

Because interaction between the bait and prey fusions causes repression of reporter gene expression, we reasoned that the RTA system would be suited for identifying protein interaction inhibitors. Disruption of the interaction would cause reactivation of reporter genes, thereby producing a positive readout for potential hits in small molecule screens (see Figure 1C). In this report, we describe modifications that facilitate identification of protein interaction inhibitors and demonstrate feasibility of the strategy using several well-defined protein interactions and previously characterized inhibitors. To demonstrate utility, we adapted the system for high-throughput screening of small molecule compound libraries and have identified novel potential interaction inhibitors for four independent protein interactions. Compounds discovered that inhibit interaction between FKBP12 and the transforming growth factor β receptor (TGFβ-R) C terminus were shown to inhibit signaling in vivo, demonstrating effectiveness of this system for identifying potential novel lead pharmacological agents.

Materials and Methods

Yeast Strains and Plasmids

The bait plasmids expressing GAL4 DNA binding domain fusions from an ARS-CEN-TRP1 backbone are derived from pY1 (14). The activated pYV bait plasmids were derived from pY1 by inserting GAL4 DNA binding domain and VP16 encoding NcoI/BglII and BglII/EcoRI PCR fragments, respectively. The multicopy LEU2 vector expressing Tup1 RD from the MET3 promoter (pIOX008) was produced by inserting an ApaI/SphI fragment (made blunt) from pBD1 (11) into YEpLac181 (15) cut with PvuII/NarI (also made blunt). Plasmid pIOX274 (LEU2, ARS-CEN prey) was produced by inserting a PGK1 promoter fragment amplified from yeast genomic DNA and a Tup1 RD—multiple cloning site (MCS)—ADH1 terminator fragment from pBD1 into the PvuII/DraIII sites of pRS315 (16). The GAL4-VP16-TGFβ-RC terminus (GV-TGFβ-R) and GAL4-VP16-thyroid hormone receptor α (GV-TRα) bait and FKBP12 and nuclear receptor co-repressor 1 (Ncor1) prey plasmids contain inserts produced by amplification from cDNA clones in pYV1 and pIOX008 or pIOX274. The ATF4 and Nrf2 activator bait plasmids were recovered from an activator bait T cell cDNA library screen (K. Lund et al., unpublished data) and the p300/ CBP Tup1 prey fusion plasmid from RTA interactions screens using these constructs as bait fusions.