Chemical Dimerizers in Three-Hybrid Systems for Small Molecule

Jun 7, 2016 - yeast and mammalian three-hybrid systems for small molecule−target protein profiling within the broader scope of target deconvolution ...
0 downloads 0 Views 2MB Size
Reviews pubs.acs.org/acschemicalbiology

Chemical Dimerizers in Three-Hybrid Systems for Small Molecule− Target Protein Profiling Dries J. H. De Clercq,†,∥ Jan Tavernier,‡,§ Sam Lievens,*,‡,§,⊥ and Serge Van Calenbergh*,† †

Laboratory for Medicinal Chemistry, Faculty of Pharmaceutical Sciences, Ghent University, 9000 Ghent, Belgium Department of Medical Protein Research, Vlaams Instituut voor Biotechnologie, 9000 Ghent, Belgium § Department of Biochemistry, Faculty of Medicine and Health Sciences, Ghent University, 9000 Ghent, Belgium ‡

ABSTRACT: The identification of the molecular targets and mechanisms underpinning the beneficial or detrimental effects of small-molecule leads and drugs constitutes a crucial aspect of current drug discovery. Over the last two decades, three-hybrid (3H) systems have progressively taken an important position in the armamentarium of small molecule− target protein profiling technologies. Yet, a prerequisite for successful 3H analysis is the availability of appropriate chemical inducers of dimerization. Herein, we present a comprehensive and critical overview of the chemical dimerizers specifically applied in both yeast and mammalian three-hybrid systems for small molecule−target protein profiling within the broader scope of target deconvolution and drug discovery. Furthermore, examples and alternative suggestions for typical components of chemical dimerizers for 3H systems are discussed. As illustrated, more tools have become available that increase the sensitivity and efficiency of 3H-based screening platforms. Hence, it is anticipated that the great potential of 3H systems will further materialize in important contributions to drug discovery.

D

one drug, one disease” core assumption that frames targetbased drug discovery is largely outdated.6 Therefore, small molecule−target profiling technologies contribute to deciphering the mechanism of action (MOA) of leads, thereby supporting chemical optimization programs while providing a polypharmacological framework6 for anticipating potential efficacy or toxicity-related later-stage attrition. Alternatively, these methods also allow identification of potential novel therapeutic applications of established drugs within the scope of drug repositioning projects. 7 However, to date, target identification often remains an important bottleneck in drug discovery, as a generally applicable methodology is still lacking.8,9 At present, the most frequently applied strategies to target identification include chemical proteomics-based methods, which have been extensively reviewed in recent years.10−13 Such methods mainly comprise affinity chromatography,14 activity-based protein profiling (ABPP),15 and label-free techniques (e.g., DARTS,16 SPROX,17 and TPP18). Despite significant technological advances in quantitative mass spectrometry (MS) approaches,19,20 these target deconvolution strategies traditionally continue to be labor-intensive and timeconsuming. In addition, sensitivity can be an important limiting factor, for instance, if the small molecule exhibits low binding affinity for its target protein or if the target of interest is low abundant. Additionally, the target proteins might suffer from

rug discovery research has mainly relied on two strategies to identify potential drug candidates, molecular and empirical approaches.1 The latter relate to phenotypic screening, which implies the measurement of phenotypic responses to small bioactive modulators in complex biological systems. Traditional unbiased phenotype-based drug discovery has been applied for over a century and successfully led to the development of numerous important drugs.2 However, with molecular biology and biochemistry moving to the center stage since the genomics era in the 1990s,3 drug discovery predominantly concentrated on molecular approaches. These include hypothesis-driven target-based approaches in which compounds that interact with a validated (protein) target are identified via high-throughput screening. Today, both screening paradigms are successfully applied side by side. This is illustrated by a recent analysis of the origins of first-in-class drugs approved by the US Food and Drug Administration (FDA) from 1999 to 2013, which indicated that although the majority (70%) were discovered through target-based approaches, phenotypic screening also contributes significantly to the pharma R&D pipelines.4 In both approaches, target deconvolution following lead identification is of critical importance. In the case of phenotypic screening, where the drug target is not a priori known, the need for target identification is obvious. However, also in the case of leads originating from a target-based screening campaign, target profiling is an essential element of the development track. Indeed, many clinically approved drugs have been found to be more promiscuous than originally thought, with every drug interacting on average with six targets.5 Thus, the “one gene, © XXXX American Chemical Society

Received: October 6, 2015 Accepted: June 7, 2016

A

DOI: 10.1021/acschembio.5b00811 ACS Chem. Biol. XXXX, XXX, XXX−XXX

Reviews

ACS Chemical Biology

Figure 1. Schematic overview of the yeast two-hybrid (Y2H) and three-hybrid (Y3H) systems. (a) Y2H: A chimeric protein (DBD-X) consisting of a DNA-binding domain (DBD) fused to a protein X is coexpressed with a second protein chimera (AD-Y) comprising a transcriptional activation domain (AD) fused to a protein Y. If protein X and Y bind to one another, a functional transcription factor is reconstituted, leading to transcription of a downstream reporter gene. (b) Y3H: Successful reconstitution of a functional trimeric complex through interaction of three hybrid components implies small molecule−target protein binding and results in the expression of a reporter gene (see text for more details). These hybrids include a hybrid protein that consists of a DBD fused to a ligand-binding domain (LBD, e.g. dihydrofolate reductase), a CID comprising an anchor moiety (e.g., methotrexate) linked to a small molecule of interest, and a second protein chimera composed of an AD fused to a library of candidate target proteins (Y).

stability issues under the classical in vitro conditions of chemical proteomics experiments, possibly resulting in the loss of their three-dimensional structure.8 Hence, to circumvent these limitations, several expression-cloning-based methods, such as yeast21 and mammalian22 three-hybrid systems, phage display,23 and mRNA display,24 have been developed in the last decades. Collectively, these functional cloning technologies exploit the artificial overexpression of target proteins from cDNA libraries to address the above-mentioned potential sensitivity problems. Moreover, rather than in vitro, three-hybrid systems operate in the physiologically relevant context of an intact cell. In the case of mammalian three-hybrid assays, this can be a human cell, which is the native background of the targeted protein(s) and the context in which the future drug will operate. A conditio sine qua non for successful three-hybrid analysis is the availability of appropriate bifunctional small-molecule probes, capable of artificially bringing two proteins together to form a stable ternary complextermed chemical inducers of dimerization (CIDs),25 or chemical dimerizers. Thorough reviews providing a broader overview on CID applications26−28 as well as those summarizing yeast “N”-hybrid systems29,30 have been recently published. However, reports elaborating on a combination of these themes, more specifically the use of CIDs in three-hybrid systems, are scarce and go back more than a decade.31 Therefore, in this review we aim to fill this gap, focusing on the design of CIDs and their specific application in both yeast and mammalian three-hybrid systems for small molecule−target protein profiling within the broader scope of target deconvolution and drug discovery.

of action studies on the immunosuppressive drugs FK506, rapamycin, and cyclosporin A (CsA) initiated the development of the concept of chemically induced dimerization.32 These studies revealed that the latter compounds act as naturally occurring heterodimerizers, since they first form a binary complex with their respective immunophilin, which subsequently competitively binds to and concomitantly inhibits a second protein enzyme. More specifically, FK506 and rapamycin both bind to FK506-binding protein 12 (FKBP12), after which the complex recruits calcineurin33 or the FKBP12-rapamycin-binding (FRB) domain of FRAP (FKBP12-rapamycin-associated protein),34 respectively. In the case of CsA, complex formation with cyclophilin is followed by interaction with calcineurin.33 The first application of the CID technology was introduced in 1993 by the Schreiber and Crabtree groups, reporting on the use of cell permeable, synthetic ligands to control an endogenous signal transduction pathway.25 In particular, an artificial dimer of FK506, FK1012, was constructed to homodimerize chimeric FKBP12 proteins fused to the cytoplasmic ζ subunit of the T-cell receptor, allowing ligandregulated signal transmission and specific target gene activation. Since this seminal paper, the chemical toolbox to induce protein−protein interactions has been greatly expanded,35 with prominent roles reserved for both FK506, rapamycin, and their synthetic analogues.36,37 A comprehensive general overview of the CID field is beyond the scope of this review. Rather, in the following sections we focus on the spectrum of dimerizer systems that has been applied in yeast as well as mammalian three-hybrid assays over the last two decades.





CHEMICALLY INDUCED DIMERIZATION The ability to bind two proteins or protein domains simultaneously represents the main characteristic of chemical inducers of dimerization. Homodimerizers are symmetrical bifunctional ligands that act to bring two identical proteins together. If two different proteins are recognized, the molecules are referred to as heterodimerizers.26 Historically, mechanism

YEAST THREE-HYBRID SYSTEMS The yeast three-hybrid (Y3H) system21 is an extension of the original yeast two-hybrid assay,38 which was introduced in 1989 by Fields and Song for the identification of protein−protein interactions. As the name suggests, the three-hybrid version includes a third hybrid component, i.e. a CID consisting of an B

DOI: 10.1021/acschembio.5b00811 ACS Chem. Biol. XXXX, XXX, XXX−XXX

Reviews

ACS Chemical Biology

Figure 2. Chemical structures of selected dimerizers used in yeast three-hybrid systems. Dexamethasone-FK506 (1); methotrexate-anecortave (2); trimethoprim-synthetic ligand for FKBP12 (3); and O6-benzylguanine-VC490004 (4).

ligand−receptor CID pairs, as systematically outlined in Table 1. Initially, the Schreiber and Crabtree laboratories exploited the use of the natural heterodimerizers FK506 and rapamycin to control transcription of integrated genes in yeast.40 In an effort to circumvent the limitations of these dimerizers related to interference with cellular functions of their endogenous binding partners, novel, orthogonal dimerizer systems were engineered following a “bump−hole” strategy.67 In this approach, an artificial cavity is introduced into the ligandbinding site of the protein to accommodate an additional, bulky substituent installed on the ligand and which inhibits binding to the wild-type protein. In the case of FK506, a Y3H screen was applied as a chemical genetic selection tool to identify compensatory mutant modular calcineurin A/B fusion protein domains (mCABs) that restored binding to a calcineurinresistant bumped analogue.41 Over the last two decades, these natural dimerizers and their synthetic derivatives found widespread use as powerful research tools in an investigational37,36 and therapeutic68,69,26 context, but they have only been scarcely applied in small molecule−target protein profiling research. An interesting yeast-based example is the concept of detecting protein−ligand interactions in stochastic nanodroplets on a large scale.42 Here, rapamycin is coupled to TentaGel resin beads via a photocleavable linker, and its timedependent photochemical release results in the dimerization of FKBP12-DBD and FRB-AD fusion proteins, which triggers transcription of a HIS3 reporter gene, allowing selective yeast growth in nanodroplets lacking histidine. In an extension of this initial design, the authors postulate the potential of the system to screen an anchored combinatorial library, consisting of large numbers of combinatorial small molecules covalently coupled to a constant anchor moiety, against a cDNA library of candidate target proteins in a single experiment. However, since this first proof of concept, to our knowledge no follow-up studies on this topic have been reported. Relying on the rather weak binding of dexamethasone to yeast-expressed GR proteins70 and its relatively low activity in recombinant yeast systems,71 Peterson and co-workers investigated steroidal estrogens as alternative immobilizing anchor moieties for profiling natural products in Y3H.43 They synthesized both a hybrid 7-α-substituted β-estradiol derivative

anchor moiety which is covalently coupled to a small molecule of interest via an appropriate linker. Accordingly, this adaptation of the Y2H assay shifts the application scope from protein−protein interactions to the detection of small molecule−target protein interactions. Similar to Y2H, the Y3H approach functions through reconstitution of an active transcription factor, which has been split into a separate DNAbinding domain (DBD) and a transcriptional activation domain (AD). In the assay, the small molecule of interest is displayed as “bait” through a high-affinity interaction of the anchor moiety of the CID with the ligand-binding domain (LBD) of a LBDDBD fusion protein. The latter is screened against a collection of chimeric candidate target proteins (referred to as “preys”; Y), encoded by a cDNA library fused to the AD. Positive interactions reconstitute a functional transcription-activating trimeric complex, which upon recruitment to specific DNAbinding sites in the promoter region of a reporter construct results in the activation of the downstream reporter gene, enabling, for example, yeast growth on a selective medium (Figure 1). Finally, sequencing of the prey fusion gene present in these positive clones completes the target identification process. The initial report describing the Y3H system was published by Licitra and Liu nearly two decades ago.21 In this pioneering work, they introduced the first chemical heterodimerizer, comprising a dexamethasone-FK506 hybrid ligand (1, Figure 2), for effectively bridging two fusion proteins: the hormonebinding domain of the rat glucocorticoid receptor (rGR) fused to the LexA DNA-binding domain and FKBP12 fused to the B42 transcriptional activation domain. Moreover, in a proof-ofprinciple experiment using this heterodimerizer, they demonstrated the potential of the Y3H system to uncover the target proteins of small organic compounds by identifying FKBP12 as the interaction partner of FK506 from a Jurkat cDNA library fused to the B42 activation domain. The application of Y3H as a screening tool for novel compound-binding proteins was later confirmed by identifying dihydrofolate reductase (DHFR) clones from a mouse cDNA library screen using the dexamethasone-methotrexate (Dex-MTX) heterodimer as a test compound in a GAL4-AD-based system.39 Hence, a major focus in the Y3H field has been the development of new C

DOI: 10.1021/acschembio.5b00811 ACS Chem. Biol. XXXX, XXX, XXX−XXX

Reviews

ACS Chemical Biology Table 1. Overview of Chemical Heterodimerizers Used in Yeast Three-Hybrid (Y3H) Systemsa Dimerizer

Description

DBD-fusion

AD-fusion

Comments

ref

Dex-FK506 Dex-MTX FK506 Rapamycin C40-PhOPhFK506

Dexamethasone and FK506 Dexamethasone and methotrexate Natural product Natural product C40-p-phenoxyphenyl bumped FK506 analogue

rGR rGR FKBP12 FKBP12 Mutant mCAB library FKBP12

FKBP12 mDHFR CNA FRB FKBP12

Used in Jurkat cDNA library screen Used in mouse cDNA library screen Using synthetic ligands as a transcriptional on−off switch

21 39 40

Screening for restored binding of compensatory mutant mCABs to a calcineurin-resistant FK506 derivative

41

FRB

42

Tetrameric SA (Y43A mutant) Tetrameric SA (Y43A mutant)

Toward large-scale screening of protein−ligand interactions in nanodroplets Enhanced potency and efficacy compared to Dex-biotin More readily synthesized than estradiol congeners

44

Rapamycin

Natural product

Estradiolbiotin Estrone oximebiotin MTX-Dex

7-α-substituted β-estradiol and biotin Estrone-17-(O-carboxymethyl) oxime and biotin

ER-α/β

Methotrexate and dexamethasone

eDHFR/ mDHFR eDHFR

rGR

Most commonly used anchor moiety

45−49

FKBP12 mutants

50

eDHFR

CDK

MTX-SLF

ER-α/β

43

MTX-PurBb

Methotrexate and synthetic ligand for FKBP12 Methotrexate and purvalanol B

MTX-PurBc

Methotrexate and purvalanol B

Cub-eDHFR

Nub-PCTK3

MTX-RGB286147

Methotrexate and a pyrazolopyrimidinone kinase inhibitor Methotrexate and anecortave acetate Methotrexate and an imidazopyridine PKG inhibitor Trimethoprim and dexamethasone Trimethoprim and synthetic ligand for FKBP12 O6-Benzylguanine and methotrexate O6-Benzylguanine and sulfasalazine O6-Benzylguanine and an anti-TB drug lead

eDHFR

CDK/CRK

KD of ligand−receptor interaction correlates with transcription read-out First successful identification of novel (CDK-like/unrelated) kinase targets Split-ubiquitin-based Y3H system for small-molecule-protein interaction analysis Application to MOA studies

eDHFR

PDE6D

Identification of new target of drug with unknown MOA

56

eDHFR

Application to MOA studies of antiparasitic drug

57, 58

eDHFR eDHFR

TgBRADIN/ GRA24 rGR FKBP12

No cross-reactivity with endogenous DHFR No endogenous (mammalian) targets

59, 60 61

hAGT

eDHFR

Selective covalently immobilizing anchor moiety (SNAP-tag)

62, 63

hAGT

SPR

64, 65

hAGT

CoxF

Identification of known and novel targets of clinically approved drugs Target deconvolution and validation of anti-TB drugs

MTX-AA MTXCmpd2 TMP-Dex TMP-SLF BG-MTX BG-Sulfad BGVC490004e

51, 52 53 54, 55

66

a

Abbreviations used: AD, transcription activation domain; CDK, cyclin-dependent kinase; CID, chemical inducer of dimerization; CNA, calcineurin A subunit; CoxF, carbon monoxide dehydrogenase F protein; CRK, CDK-related kinase; Cub/Nub, C/N-terminal half of ubiquitin; DBD, DNA-binding domain; eDHFR/mDHFR, Escherichia coli/mammalian dihydrofolate reductase; ER, estrogen receptor; FKBP12, FK506-binding protein with a mass of 12 kDa; FRB, FKBP12-rapamycin-binding domain of FRAP (FKBP12-rapamycin-associated protein); hAGT, human O6-alkylguanine-DNA alkyltransferase; mCAB, modular calcineurin A/B fusion protein domain; MOA, mechanism of action; PCTK3, PCTAIRE protein kinase 3; PDE6D, phosphodiesterase 6-delta; PKG, cGMP-dependent protein kinase; rGR, rat glucocorticoid receptor; SA, streptavidin; SPR, sepiapterin reductase; TB, tuberculosis; TgBRADIN, Toxoplasma gondii bradyzoite differentiation inhibitor. bAs well as (R)-roscovitine and indenopyrazole-1. cAs well as dexamethasone. dAs well as methotrexate, dasatinib, purvalanol B, erlotinib, atorvastatin, furosemide, and indomethacin. eAs well as methotrexate, fluphenazine, mefloquine, ofloxacin, clofazimine, and rifampicin.

reductase (DHFR) CID pair by the Cornish group,45 which represents the most commonly used anchor system to date. In an effort to characterize three-hybrid systems at the biochemical level, they examined the influence of the ligand−receptor pair on the transcriptional read-out.46 The latter was found to be much more sensitive to the structure of the fusion proteins than to variations in the MTX-Dex dimerizer design. Most surprisingly, using the bacterial E. coli DHFR (eDHFR), consistently higher transcription activation levels were obtained compared to its mammalian homologue (mDHFR), despite the fact that both exhibit similar binding affinities for MTX (Ki = 1.0−3.4 pM).73 Subsequently, this initial plasmid-based system was optimized to stabilize the transcriptional read-out in a manner that allows the straightforward variation of the AD chimera.47 Therefore, the genes encoding the DHFR anchor protein and the complementary LEU2 and lacZ reporters were integrated into the yeast chromosome. Given its remarkably reduced percentage of false negatives (from 20% to 3%) and a

fused to biotin, providing a simple model of more complex natural products, and its dexamethasone congener. To directly compare dimerizer-mediated activation of gene expression, yeasts were engineered to coexpress tetrameric streptavidin (SA)-AD and estrogen receptor (ER) α/β- or GR-DBD fusion proteins, respectively. Markedly, this new β-estradiol-based CID showed enhanced potency and efficacy compared to the previous Dex-GR CID pair. However, the routine use of these 7-α-substituted β-estradiol derivatives as CIDs is hampered by their elaborate 14-step synthesis.72,43 As a dramatically streamlined alternative, a facile two-pot synthesis starting from estrone afforded in good overall yield a novel, cellpermeable CID comprising estrone-17-(O-carboxymethyl)oxime linked to biotin, which displayed a dose-dependent activation of reporter gene expression in the ER-SA Y3H assay comparable to its estradiol congener.44 A significant evolution in the Y3H field was the introduction of the well-established methotrexate (MTX)-dihydrofolate D

DOI: 10.1021/acschembio.5b00811 ACS Chem. Biol. XXXX, XXX, XXX−XXX

Reviews

ACS Chemical Biology

anchored Y3H system (2, Figure 2), they identified phosphodiesterase 6-delta (PDE6D) as a new molecular target and validated it all the way from quantitative in vitro measurements to an animal model of the disease. So as to overcome the partial limitations of MTX-based CIDs related to their cross-reactivity with endogenous yeast DHFR and the ensuing potentially impaired transcriptional activation in Y3H, Cornish and colleagues introduced trimethoprim (TMP) as an alternative prokaryote-specific76 DHFR anchor moiety.59 This new CID pair (TMP-eDHFR) was originally established as an in vivo chemical protein labeling method, i.e. TMP-tag,77 which recently has been rendered both covalent and fluorogenic for high-resolution intracellular live cell imaging applications78 (for thorough reviews on chemical tags, see79−81). In the context of protein dimerization studies, TMP-based CIDs proved particularly useful in mammalian systems, as exemplified by the biocompatible TMP-SLF heterodimerizer (3, Figure 2)61 and further discussed in the Mammalian Three-Hybrid Systems section of this review. The latter CID 3 was successfully employed to activate transcription in an eDHFR-FKBP12 Y3H assay and to modulate a Golgiresident glycosyltransferase in Chinese hamster ovary (CHO) cells. While the spectrum of CIDs discussed so far exclusively relies on reversible interactions, some dimerizers have been developed to present the small molecule of interest in an irreversible, covalent fashion based on yet another protein labeling approach, i.e. SNAP-tag.82 This strategy is centered around the human DNA-repair protein O6-alkylguanine-DNA alkyltransferase (hAGT). The Johnsson laboratory exploited the low substrate specificity of this enzyme to covalently label SNAP-tag-fused proteins in vivo with a ligand of interest by conjugating the latter to the para position of O6-benzylguanine (BG). Extrapolating this method to the CID field, they introduced the first chemical inducer of hemicovalent dimerization comprising a BG-MTX bifunctional molecule.62 In a proof-of-principle hAGT-eDHFR Y3H system, this dimerizer induced transcriptional activation at a level comparable to that achieved with noncovalent CIDs.62 By optimizing this initial concept, the Johnsson group established a powerful yeast-based platform that enabled the identification of both known and novel interaction partners of clinically approved drugs covering a wide range of therapeutic areas.64 The methodology couples the identification of a candidate target to its subsequent unambiguous validation by combining a new, more sensitive SNAP-tag-based Y3H system with an effective pulldown or activity assay that utilizes the same benzylguanine drug derivative as in the Y3H screens. In this respect, they uncovered, for example, the first nonkinase target of erlotinib (ORP7) and characterized several off-targets of atorvastatin (such as PDE6D and NQO2). Most intriguingly, this milestone study revealed that the anti-inflammatory drug sulfasalazine and its metabolites inhibit sepiapterin reductase (SPR) and consequently decrease tetrahydrobiopterin (BH4) biosynthesis, thereby suggesting new and improved therapeutic applications for this long-standing drug. Finally, this SNAP-tagbased Y3H platform was recently adapted for target deconvolution of antituberculosis drugs by screening cDNA libraries encoding mycobacterial proteins with benzylguaninederivatized small molecules, such as the synthetic antituberculosis drug lead VC490004 (4, Figure 2).66 A comparable approach was followed by Odell et al. to identify T. gondii parasite proteins that interact with a MTX-functionalized

highly consistent reporter read-out, Cornish and colleagues assumed that this MTX-based integrated system would be suitable for the de novo identification of small molecule−target proteins in AD-fused prey cDNA library screens. In fact, in a follow-up paper, they showed that the MTXDHFR Y3H system has the requisite sensitivity for drug discovery.50 For this, a systematic study was undertaken to quantitatively characterize the correlation between ligand− receptor affinity and the transcription read-out using a heterodimer of MTX and the synthetic ligand for FKBP12 (SLF), which was previously described in the context of a bacterial RNA polymerase small molecule three-hybrid system.74 As a starting point, de Felipe et al. designed a series of FKBP12 point mutants spanning several orders of magnitude in their KD for SLF, as measured in vitro using a fluorescence polarization assay. Subsequently, the levels of transcriptional activation for these variants were determined in the Y3H assay. Interestingly, they observed that the strength of the SLFFKBP12 interaction does correlate with the transcriptional read-out, albeit over a small range of KD’s (1 order of magnitude on log scale). Furthermore, the authors concluded that in this MTX-based Y3H assay it is not possible to detect interactions whose KD’s are over ca. 50 nM, a cutoff considered high enough to realistically use the Y3H system for drug discovery. A question that may be asked, however, is whether such a system would indeed be sufficiently sensitive for detecting medium-/low-affinity interactions of different types of small molecules of interest with off-target proteins that contribute to unwanted side-effects or toxicity. In 2004 this major outstanding question was addressed by Kley and co-workers in a key publication reporting the first successful application of the Y3H assay to discover novel drug targets.51,52 More specifically, the target spectrum of cyclindependent kinase (CDK) inhibitors, including purvalanol B, was determined in a MTX-eDHFR-based system using both cDNA library and focused screening of yeast cell arrays displaying selected polypeptide open reading frames (ORFs). The target profile of the latter purine analogue proved to be significantly broader than originally anticipated, since in addition to a range of known targets (CDK1, CDK4, and CK1, among others) several novel candidate binders, such as CDC/CDK-like kinases, other types of serine/threonine kinases, and even receptor and nonreceptor tyrosine kinases, were identified. In vitro, these enzymes are inhibited by purvalanol B with IC50 values in the higher nanomolar to low micromolar range, defining a sensitivity limit of this Y3H system comparable to that of the Y2H system 75 and significantly broadening the range of detectable interactions in the context of target deconvolution and drug discovery. As such, this work represents a significant step forward by enlarging the scope of Y3H toward profiling of small-molecule kinase inhibitors, one of the main components of drug development pipelines to date. Another interesting application of Y3H are mechanism of action studies, which are essential for linking phenotypic responses with molecular targets of small molecules. This was first successfully demonstrated in 2005 by Caligiuri et al., exploring the potential applications of compounds derived from a trisubstituted pyrazolopyrimidinone kinase inhibitor scaffold (such as the putative CDK inhibitor RGB-286147).54,55 Recently, Shepard et al. elucidated the MOA of anecortave acetate (AA), an intraocular pressure-lowering cortisene for treating glaucoma.56 By displaying the latter in a MTXE

DOI: 10.1021/acschembio.5b00811 ACS Chem. Biol. XXXX, XXX, XXX−XXX

F

hEGFR

ER1 ER1

HMGCR ER1

SAR evaluation of fumagillin analogues via competition assay MASPIT cDNA library screening and competition assays for small molecule−target protein profiling in intact human cells Validation of MASPIT cell-array screening using CIDs comprising a clickable MTX anchor moiety Three-hybrid KISS setup for small molecule−target protein interaction mapping Increased MASPIT sensitivity and higher membrane permeability compared to MTX-based CIDs SNAP-tag MASPIT setup for selective covalent immobilization of CIDs to CR Photoaffinity label containing heterotrimeric CIDs for UV-cross-linking of small molecule−protein interactions Evaluation of contribution of covalent bait-prey interactions to MASPIT read-out

Pharmacological control of gene expression in human cells No affinity for endogenous FRB

No affinity for endogenous calcineurin

Inducible transcriptional activation in Jurkat cells Diversified library of cell-permeable CIDs to identify novel ligand−receptor pairs

Using synthetic ligands as a transcriptional on−off switch Reduced complexity, greater adaptability and more potent than natural product CIDs No affinity for endogenous FKBP12

Comments

110

102 110

101 102

100

98, 99 22

94−96 97, 69

41

92 93

40, 84, 85 86−88, 91 89, 90

ref

Abbreviations used: ABL, Abelson tyrosine kinase; AD, transcription activation domain; CID, chemical inducer of dimerization; CR, cytokine receptor; CyP, cyclophilin; DBD, DNA-binding domain; eDHFR, Escherichia coli dihydrofolate reductase; ER, estrogen receptor; FKBP12, FK506-binding protein with a mass of 12 kDa; FRB, FKBP12-rapamycin-binding domain of FRAP (FKBP12-rapamycinassociated protein); hAGT, human O6-alkylguanine-DNA alkyltransferase; hEGFR, human epidermal growth factor receptor; HMGCR, 3-hydroxy-3-methylglutaryl-CoA reductase; KISS, kinase substrate sensor; MASPIT, mammalian small molecule−protein interaction trap; mCAB, modular calcineurin A/B fusion protein domain; PfMetAP2, Plasmodium falciparum methionine aminopeptidase type 2; rGR, rat glucocorticoid receptor. bIn MASPIT, anchor protein (X) and prey (Y) denote the cytokine receptor (CR)-X and Y-gp130 fusion proteins, respectively, whereas, in KISS, the anchor protein (X) represents the tyrosine kinase 2 (TYK2)-X fusion protein (see Figure 4 for details). cAs well as C40-phenyl-FK506. dAs well as C16-(R)-isopropoxy rapamycin. eAs well as AP1867, E7070, RGB-285978, RGB-285961, and RGB-286147. fAs well as tamoxifen and reversine. gAs well as FK506. hAs well as reversine and simvastatin. iAs well as a benzophenone and aryl azide photo-cross-linker. jAs well as dihydro-afatinib and gefitinib.

a

eDHFR

hAGT eDHFR

O6-Benzylguanine and tamoxifen Trimethoprim, tamoxifen and trifluoromethyl phenyl diazirine Trimethoprim and afatinib

BG-TAM TMP-TAM-diazirinei

TMP-Afaj

eDHFR eDHFR

Methotrexate and simvastatin Trimethoprim and tamoxifen

eDHFR

FKBP12

PfMetAP2 ABL

rGR eDHFR

FKBP12 (3 copies) FKBP12 (3 copies)

mCAB VKMGC mutant FRB FRB LFP mutant

CyP (2 copies) ?

FKBP12 F36 V mutant (3 copies)

FKBP12 (3 copies) FKBP12 (3 copies)

AD/Preyb-fusion

FKBP12 (3 copies)

FKBP12 (3 copies) FKBP12 F36 V mutant

FKBP12 F36 V mutant (3 copies)

FKBP12 (3 copies) FKBP12 (3 copies)

DBD/Anchor proteinb-fusion

MTX-Simg TMP-TAMh

MTX-FK506f

Rapamycin C16-(R)-methallyld and other rapalogues Dex-Fum MTX-PD173955e

C40-PhOPh-FK506c

FK506 and cyclosporin A Synthetic FKBP12 ligand and substituted tetrahydrooxazepines C40-p-phenoxyphenyl bumped FK506 analogue Natural product C16-(R)-methallyl and other bumped rapamycin analogues Dexamethasone and fumagillin Methotrexate and a potent ABL kinase inhibitor Methotrexate and FK506

Bumped analogues of AP1510 series

AP1889 and congeners

Heterodimerizers FKCsA AP1867-THOXs

Dimeric FK506 Dimeric synthetic FKBP12 ligands

Description

Homodimerizers FK1012 AP1510 and congeners

Dimerizer

Table 2. Overview of Chemical Dimerizers Used in Mammalian Three-Hybrid (M3H) Systemsa

ACS Chemical Biology Reviews

DOI: 10.1021/acschembio.5b00811 ACS Chem. Biol. XXXX, XXX, XXX−XXX

Reviews

ACS Chemical Biology

Figure 3. Chemical structures of selected dimerizers used in mammalian three-hybrid systems. FK506-cyclosporin A (5); methotrexate-PD173955 (6); trimethoprim-tamoxifen (7); O6-benzylguanine-tamoxifen (8); and trimethoprim-tamoxifen-trifluoromethyl phenyl diazirine (9).

imidazopyridine PKG inhibitor (“Compound 2”).57 Of note, both TMP- and SNAP-tag-based Y3H assays are being applied as compound profiling tools as part of a commercial drug discovery service.83

step.41 Toward that end, these hits were fused to an AD domain and evaluated for their ability to restore binding to the bumped FK506 analogue, displayed by FKBP12-DBD chimeras. As a result, the VKMGC mutant was shown to most optimally respond to the calcineurin-resistant derivative. Previously, the group had followed a similar M3H-based screening approach for the discovery of novel ligand−receptor CID pairs based on rapamycin.97 As mentioned in the Yeast Three-Hybrid Systems section, next to their application as tool compounds for fundamental biology studies, rapamycin94 and its bumped analogues69 paved the way for the development of suitable systems for pharmacological control of therapeutic gene expression. In another study aiming at the identification of new ligand− receptor CID pairs, Koide et al. constructed a diversified library of heterodimerizers comprising an invariant synthetic FKBP12 ligand (i.e., AP1867) attached to 320 substituted tetrahydrooxazepines (THOXs).93 The latter system was selected as “privileged scaffold” for its structural novelty in the cyclic form and convertibility to a linear form through reduction of the N−O bond. Furthermore, the presence of the olefin and three chiral centers creates a further functionalization site and the possibility to exploit stereochemistry as a diversity element, respectively. Subsequently, the cell permeability of these candidate heterodimerizers was assessed by analyzing a representative panel of 25 library members spanning a broad range of hydrophobicities using an AP1889-based M3H competition assay. In this experiment, a cell-permeable heterodimerizer competes with the AP1889 homodimerizer for binding to FKBP12 F36 V chimeras, thereby disrupting the complex that activates SEAP expression. All of the tested compounds as well as an additional sample of six of the most polar library members were found to be freely permeable to human fibrosarcoma cells, as each gave strong inhibition of SEAP expression levels. Unfortunately, the authors did not report on the identification of any functional, novel CID pairs and only briefly referred to several initial promising “hits”. However, to our knowledge, no follow-up studies have been published ever since. In addition to the evaluation of cell



MAMMALIAN THREE-HYBRID SYSTEMS Analogous to the yeast three-hybrid systems described above, several classes of chemical dimerizers have been applied in a mammalian (human) context, as listed in Table 2. The interest in developing these mammalian systems was instigated by a number of limitations of Y3H, including its restricted membrane permeability, primitive cellular context, incapacity to monitor full size membrane proteins or interactions outside of the nucleus. Starting in 1996, FK1012 homodimerizers were shown to activate gene transcription in Jurkat cells40,84 and human skin keratinocytes and fibroblasts.85 In a next stage, wholly synthetic dimeric SLFs (such as AP1510) were developed by Ariad Gene Therapeutics, which combine a reduced complexity and greater adaptability with higher potency compared to the first generation natural productderived CIDs.86−88 Finally, in order to potentially reduce nonproductive interactions and further improve performance, a series of novel, homodimeric bumped AP1510 analogues (e.g., AP1889) was designed to bind specifically to the F36 V point mutant and only minimally to endogenous FKBP12.89,90 The first synthetic heterodimerizer implemented in a mammalian three-hybrid setting was described by Crabtree and Schreiber in 1996, and it consists of a fusion of the natural products FK506 and cyclosporin A (FKCsA 5, Figure 3).92 By dimerizing FKBP12-DBD and cyclophilin (CyP)-AD chimeras, the hybrid ligand strongly stimulated the expression of a GAL4responsive secreted alkaline phosphatase (SEAP) reporter gene in Jurkat cells. As for the Y3H system, initial attention was also paid to the well-established natural heterodimerizers FK506, rapamycin and derivatives thereof. Having identified a set of selected mutant mCAB hits that restored binding to C40-pphenoxyphenyl FK506 using a Y3H system (see the Yeast Three-Hybrid Systems section), the Schreiber group applied a reciprocally configured M3H assay in a subsequent validation G

DOI: 10.1021/acschembio.5b00811 ACS Chem. Biol. XXXX, XXX, XXX−XXX

Reviews

ACS Chemical Biology

Figure 4. Outline of the MASPIT and KISS mammalian three-hybrid (M3H) systems. (a) MASPIT: E. coli dihydrofolate reductase (D) is fused to a cytokine receptor (CR) containing mutated STAT3 recruitment sites in its cytoplasmic tail (gray dots). A prey protein (Y) is tethered to a gp130 cytokine receptor fragment containing functional STAT3 docking sites (black dots). When a methotrexate (MTX) fusion compound is added to the cells, MTX binds to eDHFR, resulting in the compound of interest (asterisk) being displayed as bait. Upon administration of the appropriate cytokine ligand (cyt), JAK2 kinases are activated through cross-phosphorylation (P). When the small-molecule bait and the prey-gp130 fusion interact, JAK kinases phosphorylate the STAT3 docking sites on the gp130 chain (P). STAT3 recruitment and STAT3 phosphorylation (P) yield activated STAT3 dimers that induce the expression of a reporter gene coupled to a STAT3-dependent promoter. (b) KISS: eDHFR (D) is tethered to the C-terminal (kinase-domain-containing) portion of tyrosine kinase 2 (TYK2) and coexpressed with a prey protein (Y) fused to a gp130 CR fragment. Upon physical interaction between the small molecule and the prey protein, TYK2 phosphorylates STAT3 docking sites on the prey chimera (P), which eventually triggers reporter gene activation as in the case of the MASPIT assay.

profiling in a physiological relevant environment and conserves normal protein conformation as well as post-translational modifications. Furthermore, this might reveal potential effects of the target’s association with additional proteins or other intracellular molecules on small-molecule binding. MASPIT is the three-hybrid component of the MAPPIT (mammalian protein−protein interaction trap) technology platform,106,107 an approach that uses the JAK-STAT pathway which mediates cytokine receptor signaling. MASPIT employs E. coli dihydrofolate reductase (eDHFR), fused to a mutated, functionally inactivated cytokine receptor to present a methotrexate (MTX) fusion compound (MFC) to the intracellular environment. This allows screening of a smallmolecule bait against a collection of chimeric prey proteins. As a result of the interaction between the small-molecule bait and a target prey protein, the JAK-STAT pathway is activated, resulting in the expression of a reporter gene (Figure 4A). The original MASPIT paper was published by Kley and coworkers in 2006, focusing on MTX-based CIDs incorporating diverse small-molecule chemotypes, such as the potent pyridopyrimidine ABL tyrosine kinase inhibitor PD173955 (6, Figure 3).22 The latter MTX fusion compound was applied in a MASPIT cDNA library screen aiming at the proteomewide de novo identification of its target proteins. Only kinases were detected in this screen; besides different known kinases (e.g., LYN, FYN and FGFR1), the authors also identified several novel kinase targets of this compound, such as various ephrin receptors and cyclin G-associated kinase. Furthermore, this study also describes for the first time the use of competition assays to compare and rank unmodified small molecules (e.g., PD173955 and imatinib) on the basis of their relative ability to displace a CID (MTX-PD173955) from its target protein (ABL) in intact mammalian cells (analogous to

permeability properties, M3H competition experiments with native unfused small molecules also allow for their prioritization with respect to in cell target-binding affinity. Accordingly, structure−activity relationships (SAR) can be mapped out based on the side-by-side comparison of analogue series. An interesting example of this strategy is provided by Chen et al., who performed dexamethasone-fumagillin-mediated M3H assays in the absence and presence of a series of potential competitors of this antimalarial lead.99,98 The possibility to perform such competition assays in a host system that is readily permeable to a wide variety of small molecules, offers an important advantage of mammalian over yeast three-hybrid systems. Namely, wild-type yeast are naturally resistant to many drugs, whose intracellular accumulation is hampered by constitutively active multidrugresistance (MDR) mechanisms,103 e.g. through the expression of ATP-binding cassette (ABC) transporters that actively mediate drug efflux.104,105 Yeast reporter strains have been engineered, by specific deletion of genes contributing to these drug efflux mechanisms, to enhance permeability to small molecules,64 thereby providing an artificial system that allows the same type of competition analysis as one can perform in mammalian cells. However, an inherent limitation to both the mammalian three-hybrid systems discussed so far and Y3H in general is their restriction to the detection of interactions with proteins or protein domains that can be functionally expressed and effectively translocated into the nucleus. MASPIT (mammalian small molecule−protein interaction trap) circumvents this disadvantage by shifting the analysis of binding events between small organic compounds and their target proteins to the cytosol of living human cells (Figure 4A).22 Additionally, the intact human cellular context (HEK293T) ensures compound H

DOI: 10.1021/acschembio.5b00811 ACS Chem. Biol. XXXX, XXX, XXX−XXX

Reviews

ACS Chemical Biology

Figure 5. Typical components of chemical dimerizers for three-hybrid systems. Schematic noncomprehensive overview of anchor moieties, linkers, and conjugation methods applicable to the synthesis of chemical dimerizers as exemplified by TMP-TAM 7. All anchor moieties and conjugation methods (except for hydrazone ligation) listed represent alternative examples that have not been applied in the context of three-hybrid studies to date. Anchor proteins are in parentheses. SA, streptavidin.

limitation that applies to both MASPIT and KISS is that compounds which exhibit affinity for any of the proteins involved in the JAK/STAT pathway might compromise the assay readout and hence are not compatible with the approach (e.g., a broad spectrum kinase inhibitor might inhibit JAK2 kinase activity). Recently we reported the evaluation of two alternative chemical dimerizer approaches aimed at increasing the sensitivity of MASPIT, based on tamoxifen (TAM) as model bait.102 To circumvent any potential limitations related to the tight binding of MTX to endogenous human DHFR, trimethoprim (TMP) was explored as an alternative prokaryote-specific eDHFR ligand. MASPIT evaluation of TMP fusion compounds with tamoxifen, reversine, and simvastatin as pilot baits, resulted in dose−response curves being shifted toward lower EC50 values than those of their MTX congeners. Additionally, a scalable TMP-azido reagent was synthesized that displayed a similar improvement in sensitivity (7, Figure 3), possibly owing to increased membrane permeability relative to the MTX anchor. Furthermore, so as to stabilize the ternary complex, the concept of covalent bonding was introduced into the MASPIT assay, which hitherto relied on reversible interactions of either component of the dimerizers. Recent reports support the notion that covalently ’locking‘ of small molecule−protein interfaces can increase assay performance significantly. This has been shown e.g. using the SNAP-tag approach in various assays, such as Y3H64,66 or time-resolved fluorescence resonance energy transfer (TR-FRET)108 for the high-throughput identification or characterization of ligand− receptor interactions, respectively. In a first attempt, we explored the use of the SNAP-tag in the MASPIT assay to selectively and covalently immobilize the bait fusion compound to the cytokine receptor anchor fusion protein. Unexpectedly, in this specific case study, this approach did not yield the hypothesized increase in sensitivity, as the O6-benzylguanineTAM heterodimerizer (8, Figure 3) produced an inferior readout compared to its MTX or TMP congeners. For example, the induction window turned out to be significantly lower (6-fold vs 177-fold) and the EC50 markedly higher (2.4 × 10−7 M vs 2.5 × 10−8 M) than in the case of the TMP-tag setup.102 There might be a number of reasons for this, including titration of the supplied benzylguanine fusion compound by endogenous species of O6-alkylguanine-DNA alkyltransferase (AGT), the enzyme employed in the SNAP system to mediate the covalent linkage,82 and cytokine receptor-AGT fusion protein degradation upon alkyl transfer.109 Therefore, further optimization efforts were initiated by implementing photoactivatable cross-

the SAR evaluation in ref 99, as discussed above). Interestingly and unlike in vitro KD or IC50 values, such quantitative cellular targeting potencies integrate many variables associated with compound cell-bioavailability and intracellular targeting efficiency. Recently, our research group presented a versatile, clickable MTX reagent that allows swift γ-selective conjugation to yield MTX fusion compounds appropriate for MASPIT.100 The general applicability of the reagent was demonstrated using three structurally diverse pharmacologically active compounds, i.e. tamoxifen, reversine, and FK506. In analytical mode, MASPIT produced concentration-dependent reporter signals for the established target proteins of these model baits. Furthermore, in a proof-of-concept experiment, the FK506 MTX fusion compound was explored in a cell-array screen for targets of FK506. Out of a pilot collection of nearly 2,000 fulllength human ORF preys, FKBP12 was selectively identified as an interaction partner of FK506, thereby validating the MASPIT system and showing its potential toward uncovering new intracellular targets of small molecules of interest. In contrast to classical target-based profiling, this M3H system can thus provide information regarding unanticipated small molecule−target protein interactions. Hence, MASPIT represents an unbiased screening approach, not restricted to a limited panel of related proteins but covering a broad area of the human proteome (currently ca. 15,000 ORFs). However, an important limitation of the system is its incompatibility with (full size) transmembrane proteins. Indeed, since the smallmolecule bait is recruited to a plasma membrane anchored cytokine receptor chimera, only interactions with soluble target proteins can be detected. The latter shortcoming was addressed by the introduction of KISS (kinase substrate sensor), a novel binary protein−protein interaction mapping approach that enables in situ analysis in living mammalian cells of protein interactions and additionally small molecule−target protein profiling in a three-hybrid setup.101 In this assay, small molecules are displayed as baits through recruitment of their corresponding MTX fusion compounds to an E. coli DHFR (eDHFR)-tyrosine kinase 2 (TYK2) fusion protein, and interactions with target protein prey fusions are detected through reporter induction as in the case of the MASPIT assay (Figure 4B). Contrary to MASPIT, the KISS 3H approach allows screening for integral membrane target proteins as this interaction sensor is freely diffusible. The concept was successfully applied to the detection of interactions between the drugs simvastatin and FK506 and their primary target proteins HMGCR and FKBP12, respectively. A I

DOI: 10.1021/acschembio.5b00811 ACS Chem. Biol. XXXX, XXX, XXX−XXX

Reviews

ACS Chemical Biology

Toward this end, the attachment position of the spacer, both on the anchor moiety and bait, must be meticulously chosen so as to minimize loss in binding affinity through perturbation of pharmacophores or induction of conformational changes. In this respect, linker studies are often guided by available SAR, structural, and molecular modeling data, which facilitate selection of the attachment site and incorporation of appropriate chemical handles. Indeed, in the case of DHFR ligands attachment positions are well established. For example, SAR studies of MTX derivatives have emphasized the importance of selective conjugation to the γ-carboxylic acid of the glutamate moiety to ensure high-affinity binding to eDHFR through interaction of this enzyme with the free α-carboxylic acid.118,119 However, appropriate derivatization of smallmolecule baits is often less obvious and is generally initiated from SAR information on their primary targets (as far as known). Of note, for comprehensive target deconvolution studies one should ideally rely on phenotypical SAR data, acquired from testing a panel of derivatives modified at different positions for retained activity in appropriate functional assays.64 Two main factors that govern the design of a particular linker are its length and chemical composition. Though crucial, no hard and fast rules are available for identifying the optimal linker length, which is typically determined empirically. For instance, one can envision that sufficient length and flexibility would be necessary to effectively bridge both binding sites and successfully heterodimerize the two fusion proteins, while on the other hand the linker should be short enough to ensure the desired proximity-related dimerization event to occur. Hence, this supports the development of modular synthesis strategies, enabling linker length and conjugation method to be swiftly mixed and matched as required (Figure 5).58,100 Previously, the effect of CID linker length modifications on the transcription read-out in both yeast and mammalian three-hybrid systems has been investigated, yet yielding contradictory outcomes. Piloting work by Cornish and colleagues suggested an increase in Y3H sensitivity by increasing the linker length of MTX-Dex CIDs.46 However, only subtle differences among five-, eight-, and tenmethylene linkers were revealed. Similarly, upon MASPIT evaluation of different tamoxifen MTX fusion compounds comprising four-, six-, or eight-PEG linkers, no significant differences in read-out were observed by our group.100 Yet, the latter trend was later confirmed in several other yeast-based studies,43,51,53 but contrasts sharply with the reports by Amara86 and Schreiber,84 observing a clear inverse correlation between linker length and activity. These discrepancies might be explained by the relative strength of the interactions, as well as geometrical and steric effects governed by the nature of the prey chimeras and hence frequently underscore the need for application-specific CID optimizations. Three main classes of linker types have been applied for the construction of CIDs in the context of 3H systems, i.e. straightchain aliphatic (methylene), 46 poly(ethylene glycol) (PEG),51,53,58,100 and mixed-type spacers (e.g., N,N′-(ethane1,2-diyl)bis(2-hydroxyacetamide)-derived).88,86 The chemical composition of the linker might significantly influence the stability, solubility, cell permeability and hence overall efficiency of CIDs. Clearly, one should keep away from reactive functional groups, such as esters, which could be cleaved in vivo. Johnsson and co-workers have shown that a benzylguanine-based CID comprising an aliphatic linker proved to be more effective than its PEG congener, most probably due to the fact that the

linkers on the bait-end of the dimerizer (e.g., diazirine 9, Figure 3) in order to capture transient and low-affinity bait-prey interactions.110 In parallel, the contribution of covalent baitprey interactions to the MASPIT read-out will be evaluated using a panel of irreversible and reversible EGFR tyrosine kinase inhibitors. It will be interesting to see whether covalent binding at the bait-end will enhance the sensitivity of the assay.113



DESIGN ASPECTS OF CHEMICAL DIMERIZERS Ideally a CID for three-hybrid applications should be accessible via a modular, versatile and practical synthesis in a minimal number of steps and acceptable overall yield. Furthermore, the CID should be readily soluble and cell permeable. In the following three subsections examples and alternative suggestions for anchor moieties, linkers and conjugation methodologies are discussed (Figure 5). Anchor Moieties. The introduction of O6-benzylguaninehAGT (SNAP-tag) in Y3H by the Johnsson group, enabling screening of covalently anchored small-molecule baits, has proven to significantly contribute to the performance of these assays (KD cutoff in low micromolar range).64,66 Interestingly, the broad application scope of this labeling technology enables additional or parallel uses for the benzylguanine-based CIDs, such as pulldown and other assays for detecting and analyzing ligand−receptor interactions.108 In this respect, and in line with a trend that has also been observed in the field of live-cell imaging and proteomics, we anticipate that the 3H field will further focus on the implementation of ligand−receptor CID pairs that enable establishing covalent immobilization of bait small molecules. Hence, in our opinion, a number of alternative anchor moieties might be considered as suitable immobilization reagents (Figure 5, left panel). These predominantly originate from a limited number of irreversible, orthogonal covalent chemical tags, which were introduced in recent years. In order to render the TMP-tag covalent, the Cornish group designed an eDHFR variant (eDHFR:L28C) with a unique cysteine nucleophile positioned just outside the TMP-binding pocket to react with an acrylamide electrophile installed on the TMP-probe (A-TMP) via a proximity-induced Michael addition.111−113 Furthermore, the substrate scope of the SNAP-tag labeling technology implemented in 3H systems could be extended to O4-benzyl-2-amino-6-chloropyrimidine (CP) conjugates, which have shown higher cell permeability compared to the corresponding benzylguanine derivatives.114 Alternatively, the specific labeling of fusion proteins based on the irreversible reaction of O2-benzylcytosine (BC) derivatives with an engineered AGT mutant, coined CLIP-tag, might be explored.115 Finally, yet another modular protein tagging system, i.e. HaloTag, should prove useful for covalently immobilizing small molecules of interest, originally attached to a chloroalkane linker, onto a modified haloalkane dehalogenase chimera.116 Lastly, one could rely on biotinbased heterodimerizers and their quasi irreversible interaction with streptavidin (SA) fusions, which implies a reciprocal setup as described by Muddana et al.44 or a comparable approach to the alternative Y3H system by Athavankar et al.117 Linkers. To discourage steric hindrance of fusion partners, which could cause the CID to bind suboptimally to its anchor protein or allow one to overlook targets that might interact with the unconjugated small-molecule bait, a linker is introduced to allow optimal interaction with prey chimeras. J

DOI: 10.1021/acschembio.5b00811 ACS Chem. Biol. XXXX, XXX, XXX−XXX

Reviews

ACS Chemical Biology

appropriate cell type. The resulting pool of cells, ideally each expressing a different prey protein, is treated with a CID containing the bait small molecule and cells harboring an interacting small-molecule bait−target prey pair can be recovered through selection for reporter gene activity. In yeast systems, growth selection through an auxotrophic marker potentially combined with a β-lactamase-based colorimetric read-out is typically used.21,39,51,56 In mammalian cells, a wider variety of growth selection (mostly based on antibiotic resistance markers), enzymatic (e.g., luciferase) or fluorescent read-outs are being applied.22,127 Although seemingly straightforward, cDNA library screening is quite labor- and timeconsuming: a screen for a single small-molecule bait typically requires several weeks, if not months, from the actual library screen up to the selection of validated clones. In addition, once these clones have been selected, target protein identity still needs to be retrieved through prey cDNA amplification and sequencing. An alternative format that accommodates higher throughputs is array screening. This entails parallel testing in an arrayed fashion of individual combinations of the small-molecule bait of interest with a (large) set of target preys, typically in microtiterplates or using (micro)arrays. Conceptually, an important benefit is that, because preys are assayed individually rather than in the bulk of a cDNA library-expressing cell pool, there is no competitive selection among different target proteins, which results in a higher level of sensitivity of an array format. Compared to cDNA library screening, setting up an array screening platform requires much more effort in terms of generation of the library of cDNA clones (which need to be cloned individually) and infrastructure for printing and scanning the arrays. Once operational, however, array replicas can be produced batch-wise (potentially in a largely automated way) and stored as a flexible screening resource, and an efficient arrayed 3H platform enables screening large collections of target proteins within days, whereby target identity is revealed by its position in the array. A focused yeast cell array covering a limited subset of protein kinases has been used to complement a cDNA library screening approach aimed at identifying novel targets of purvalanol B, as mentioned higher. Here, array replicas of yeast cell clones stably expressing different kinase prey fusion proteins were generated and screened with different concentrations of a number of small-molecule CIDs.51 Over the past couple of years, our group has invested heavily in adopting the MASPIT and KISS M3H assays to a cell array format. In the current setup, functionalized mixtures containing individual prey-encoding plasmids are being printed at high density and screened through reverse transfection. In this approach, a cell pool, bulk transfected with the cytokine receptor-eDHFR anchor protein and treated with the small-molecule-derived CID (see Figure 4), is grown as a monolayer on the array, resulting in local uptake and expression of the target prey. In cells growing on top of a plasmid encoding a target protein of the small molecule of interest, the reporter gene is activated, flagging the interaction. A pilot screen of a MTX-FK506 CID against a collection of close to 2,000 full size human ORFs arrayed in 384-well plates successfully identified the known FK506 target FKBP12.100 Based on this successful proof-of-concept study, the array has been developed further both with respect to library size and footprint. The arrayed collection now comprises close to 15,000 full size human ORFs, including the entire human ORFeome version 8.1 set,128 which are being screened

former spacer relatively increases the overall hydrophobicity of the CID, favoring its cellular uptake.62 However, this frequently observed enhanced permeability of methylene spacer-based CIDs is often at the expense of their solubility and too hydrophobic linkers could promote undesired aggregation in aqueous media.28 PEG linkers on the other hand were reported to yield dimerizers with generally good water solubility properties and, remarkably, ditto yeast as well as mammalian cell permeability.51,31,102 Conjugation Methodologies. Various (bio)conjugation methods are available for fusing both the anchor moiety and small-molecule bait to either end of the linker. Most synthetic strategies apply general anchor-spacer reagents (e.g., MTX(PEG)5-NH2 or BG-(PEG)4-NH2)51,64 for subsequent ligation to (appropriately functionalized) small molecules using standard peptide coupling, Mitsunobu or N-alkylation reactions, thereby yielding amide, ether or amine linkages, respectively.31 Remarkably, only by 2011, the highly powerful, regioselective and reliable copper(I)-catalyzed azide alkyne cycloaddition “click” reaction (CuAAC)120 found acceptance in the Y3H field.64 Two years later, our group100 and others58 independently reported on newly developed synthetic strategies, implementing “clickable” MTX reagents for the straightforward generation of CIDs for 3H applications. Besides the frequently employed conjugation chemistries listed above, a range of alternative click-type modular fusion reactions could be applied (Figure 5). For instance, the toolbox of thiol-click reactions121 (such as thiol−ene, thiol−yne, thiol-Michael) or (hetero)Diels−Alder ligations,122 encompassing [4 + 2] cycloadditions, might be exploited. Furthermore, one could envision the use of carbonyl-condensation reactions of the “nonaldol” type, such as imine-, oxime- and hydrazone-bond formation.123 An interesting recent example of the successful application of the latter conjugation reaction is provided by Moser and Johnsson,66 describing the synthesis of benzylguanine derivatized rifampicin, among others, for target identification using an adapted SNAP-tag-based Y3H system (see the Yeast Three-Hybrid Systems section). Finally, alternative amidation reactions, such as the Staudinger-Bertozzi ligation124 or thio acid/sulfonyl azide amidation,125 also denoted as sulfo-click, 126 could be appropriate for chemoselective CID construction.



TECHNOLOGICAL ADVANCES IN SCREENING PLATFORMS Adoption of 3H technologies in a small-molecule profiling platform, aimed at de novo identification of (on- or off-) target proteins, requires access to a set of potential target proteins (preferably proteome-wide) and an approach that enables efficient screening of this collection. Essentially two general methodologies can be pursued in high-throughput 3H applications: cDNA library and array screening. Traditionally, mainly the former has been applied because this requires relatively little resources with respect to both building and screening extensive libraries. Moreover, since the screening format is essentially the same as in the case of (yeast or mammalian) two-hybrid technologies for the identification of protein−protein interactions, tools and libraries, which are broadly distributed within the research community and can be acquired commercially, can be applied as such in a 3H setup. In brief, a mRNA sample derived from (a) relevant cell type(s) is cloned as cDNA in a suitable prey vector and the resulting cDNA library is coexpressed together with the anchor protein (e.g., the LBD-DBD fusion protein in the case of Y3H) in the K

DOI: 10.1021/acschembio.5b00811 ACS Chem. Biol. XXXX, XXX, XXX−XXX

Reviews

ACS Chemical Biology

raphy (pulldown) combined with high-resolution mass spectrometric analysis appears to be the most frequently applied.8 Despite significant technological advances in quantitative MS approaches,19,20 affinity-based proteomics, however, continues to be labor-intensive and prone to artifacts, as it often fails to reveal physiologically relevant interactions. In fact, significant triage and additional (de)validation is required to identify the actual target(s) from the extensive list of candidates typically resulting from such experiments.12 Threehybrid systems, on the other hand, operate against the physiological background of an intact (yeast or mammalian) cell, thereby overexpressing a target protein collection to overcome the sensitivity issues related to endogenous expression levels in proteomic methods. Another asset of 3H systems includes their generality since they function regardless of the nature of both the small-molecule drug and the candidate target proteins. Additionally, once a target protein for a small molecule of interest is identified, 3H systems provide an interesting binary small molecule−target assay, which allows for the evaluation of structural variations of either partner. More specifically, focusing first on the small molecule side, structure−activity relationships can be delineated, e.g. during lead optimization, based on the side-by-side comparison of analogue series via competition experiments. Hence, unfused small molecules can be ranked with respect to their in cell target-binding affinity, as exemplified earlier for a series of antiplasmodial fumarranol analogues (see the Mammalian Three-Hybrid Systems section).99 Alternatively, by creating and screening a systematic collection of target protein point mutants against a specific small-molecule bait, their interaction interface can be carefully mapped. This might provide additional structural information facilitating medicinal chemistry and drug design campaigns. However, as for every other target identification methodology, 3H approaches display some inherent limitations. For example, a major disadvantage of 3H methods includes the necessity to derivatize the small molecule of interest. Toward this end, extensive knowledge of the structure−activity relationship of the compound is needed and the process often requires elaborate chemistry efforts to identify a conjugation site on the molecule that tolerates modification without significant loss in binding affinity to the target protein(s). Additionally, the application of 3H systems is hampered when the conjugated small-molecule drug exhibits variable or limited cellular uptake. Furthermore, most 3H methods are not compatible with the detection of (full size) transmembrane proteins. While methotrexate-eDHFR has been the most extensively applied CID pair in the 3H field, currently two other prominent anchor moieties serve as useful starting points for compound profiling campaigns. First, based on our own experience in the design and synthesis of CIDs for the optimization of M3H systems, trimethoprim comprises a valuable alternative for MTX and, in fact, offers a number of important advantages over the latter, including significantly higher membrane permeability, increased solubility as well as compatibility with various reaction conditions, and lack of chirality.102 Second, the development of a sensitive SNAP-tag-based Y3H system by the Johnsson group has proven to have significant potential to reveal physiologically relevant interactions between smallmolecule drugs and their target proteins.64,66 Hence, as discussed in the Anchor Moieties section of this review, we expect that in the near future key roles will be

on a microarray platform, enabling a throughput of multiple screens a week.129 Although such an approach clearly exhibits vast potential as to scale and speed, besides the fact that it requires significant upfront investments to have such a platform in place, there are also a number of other points to consider. A potential disadvantage compared to cDNA library-based approaches is that, in particular when only full size ORFs are being used, interactions with specific subdomains that are not properly exposed (e.g., in cases where the small molecule binds to a subdomain of the target protein that is only accessible in specific protein conformers induced under particular physiologic conditions) might be missed. Additionally, a cDNA library might cover target protein isoforms not present in a full size ORF collection. These issues can be easily addressed however by growing the arrayed library beyond full size ORFs, by adding in additional protein isoforms and subdomains in order to move toward the complexity of classical cDNA libraries.



CONCLUSIONS AND PROSPECTS FOR DRUG DISCOVERY Although the number of published reports describing the application of three-hybrid systems in an actual drug discovery context is rather confined, these papers clearly demonstrate the potential of three-hybrid approaches to uncover the molecular targets of drugs and lead compounds. Major contributions to the field have been made by the Kley,22,51 Cornish,56 and Johnsson64,66 groups. Undoubtedly, these successful de novo small-molecule target identification studies opened up intriguing insights into the mechanisms of action of both drug candidates and clinically approved drugs, thereby potentially providing explanations for off-target-related side-effects and suggesting new or improved therapeutic applications. Moreover, 3H systems should prove useful for uncovering yet unidentified targets that contribute to the efficacy of known drugs. For example, the discovery that sulfasalazine interferes with BH4 metabolism might explain some of its adverse effects related to changes in the concentration of neurotransmitters dependent on this cofactor.64 Accordingly, this realization at the same time allows set up of hypotheses to alleviate these side-effects, for instance by adjunct therapy with agents that would restore a correct neurotransmitter balance. Furthermore, considering sulfasalazine’s specific inhibition of BH4 biosynthesis, this drug might be qualified for repurposing for other therapeutic applications, such as the suppression of chronic pain in cancer patients. Other examples highlighting the utility of 3H screening in drug MOA studies include the work by Caligiuri et al. and Odell et al.54,57 In the former study, the authors applied Y3H to explore the pharmacological properties and therapeutic potential of compounds derived from a trisubstituted pyrazolopyrimidinone CDK kinase inhibitor scaffold. Odell et al. recently used a Y3H screen to study the proteins and mechanisms involved in stage conversion in T. gondii, ultimately aiming at the development of new approaches to prevent the establishment of chronic infection. Y3H screenings are also being offered as a commercial compound profiling service, e.g. by the company Hybrigenics Services,83 who acquired the three-hybrid activities of Dualsystems Biotech AG, which had formerly significantly contributed to the field.29,56 Among the currently available methodologies for smallmolecule target deconvolution, classical affinity chromatogL

DOI: 10.1021/acschembio.5b00811 ACS Chem. Biol. XXXX, XXX, XXX−XXX

Reviews

ACS Chemical Biology reserved for anchor moieties, enabling covalent immobilization of small-molecule baits, such as those based on a O2benzylcytosine (CLIP-tag)115 or chloroalkane (HaloTag)116 scaffold (Figure 5, left panel). Finally, one can anticipate that protein-fragment complementation assays (PCAs), which have seen a veritable explosion of applications in the past decade in numerous fields, such as cell biology, drug discovery, and chemical biology,130−132 may ultimately make their way to the 3H field for small molecule−target protein profiling studies. It is clear that three-hybrid systems have a great potential, both for de novo identification of small-molecule targets and for analysis of specific compound-protein interactions. As more tools have become available that increase the sensitivity and efficiency of 3H-based platforms, there is no doubt that this potential will materialize in important contributions to drug discovery.





AUTHOR INFORMATION

Target deconvolution: the identification of the molecular targets that underlie an observed phenotypic response, an important facet of current drug discovery Three-hybrid (3H) systems: a derivative of two-hybrid (2H) systems to detect and screen for the interaction between small-molecule drugs and their protein targets. Similar to 2H systems, 3H systems typically comprise a “bait” fusion (protein-of-interest X attached to the DNAbinding domain of a transcription activator) and a “prey” fusion protein (protein-of-interest Y attached to the transcription activation domain of the same transcription activator), but also comprise an “anchor” and a “linker” molecule attached to the bait fusion protein, allowing the attachment of the small-molecule drug to the “bait protein scaffold”

REFERENCES

(1) Swinney, D. C., and Anthony, J. (2011) How were new medicines discovered? Nat. Rev. Drug Discovery 10, 507−519. (2) Drews, J. (2000) Drug discovery: A historical perspective. Science 287, 1960−1964. (3) International Human Genome Sequencing Consortium (2004) Finishing the euchromatic sequence of the human genome. Nature 431, 931−945. (4) Eder, J., Sedrani, R., and Wiesmann, C. (2014) The discovery of first-in-class drugs: origins and evolution. Nat. Rev. Drug Discovery 13, 577−587. (5) Mestres, J., Gregori-Puigjane, E., Valverde, S., and Sole, R. V. (2009) The topology of drug-target interaction networks: implicit dependence on drug properties and target families. Mol. BioSyst. 5, 1051−1057. (6) Hopkins, A. L. (2008) Network pharmacology: the next paradigm in drug discovery. Nat. Chem. Biol. 4, 682−690. (7) Jin, G. X., and Wong, S. T. C. (2014) Toward better drug repositioning: prioritizing and integrating existing methods into efficient pipelines. Drug Discovery Today 19, 637−644. (8) Ziegler, S., Pries, V., Hedberg, C., and Waldmann, H. (2013) Target Identification for Small Bioactive Molecules: Finding the Needle in the Haystack. Angew. Chem., Int. Ed. 52, 2744−2792. (9) Chan, J. N. Y., Nislow, C., and Emili, A. (2010) Recent advances and method development for drug target identification. Trends Pharmacol. Sci. 31, 82−88. (10) Schirle, M., Bantscheff, M., and Kuster, B. (2012) Mass Spectrometry-Based Proteomics in Preclinical Drug Discovery. Chem. Biol. 19, 72−84. (11) Rix, U., and Superti-Furga, G. (2009) Target profiling of small molecules by chemical proteomics. Nat. Chem. Biol. 5, 616−624. (12) Raida, M. (2011) Drug target deconvolution by chemical proteomics. Curr. Opin. Chem. Biol. 15, 570−575. (13) Wang, K., Yang, T., Wu, Q., Zhao, X., Nice, E. C., and Huang, C. H. (2012) Chemistry-based functional proteomics for drug target deconvolution. Expert Rev. Proteomics 9, 293−310. (14) Saxena, C., Higgs, R. E., Zhen, E., and Hale, J. E. (2009) Smallmolecule affinity chromatography coupled mass spectrometry for drug target deconvolution. Expert Opin. Drug Discovery 4, 701−714. (15) Cravatt, B. F., Wright, A. T., and Kozarich, J. W. (2008) Activity-based protein profiling: From enzyme chemistry. Annu. Rev. Biochem., pp 383−414, Annual Reviews, Palo Alto. (16) Lomenick, B., Hao, R., Jonai, N., Chin, R. M., Aghajan, M., Warburton, S., Wang, J. N., Wu, R. P., Gomez, F., Loo, J. A., Wohlschlegel, J. A., Vondriska, T. M., Pelletier, J., Herschman, H. R., Clardy, J., Clarke, C. F., and Huang, J. (2009) Target identification using drug affinity responsive target stability (DARTS). Proc. Natl. Acad. Sci. U. S. A. 106, 21984−21989. (17) West, G. M., Tucker, C. L., Xu, T., Park, S. K., Han, X. M., Yates, J. R., and Fitzgerald, M. C. (2010) Quantitative proteomics approach for identifying protein-drug interactions in complex mixtures using

Corresponding Authors

*E-mail: [email protected]. *E-mail: [email protected]. Present Addresses ∥

(D.D.C.) Department of Cancer Biology, Dana-Farber Cancer Institute, Boston, Massachusetts 02115, USA, and Department of Biological Chemistry and Molecular Pharmacology, Harvard Medical School, Boston, Massachusetts 02115, USA. ⊥ (S.L.) Orionis Biosciences, 9052 Ghent, Belgium. Notes

The authors declare the following competing financial interest(s): J.T. is scientific co-founder of Orionis Biosciences, which holds a license to the MAPPIT, MASPIT, and KISS technology platforms. J.T. and S.L. are employees of Orionis Biosciences.



ACKNOWLEDGMENTS J.T. was supported by a grant from IUAP P6/36, and is a recipient of an ERC Advanced grant (CYRE,340941). S.V.C. and J.T. received support from the Fund for Scientific Research-Flanders (FWO-V). D.D.C. obtained a Ph.D. grant from the Special Research Fund of Ghent University (BOF).



KEYWORDS Chemical inducer of dimerization (CID): bifunctional small-molecule probes, capable of artificially bringing two proteins together to form a stable ternary complex; also referred to as “chemical dimerizer” or “hybrid ligand” Conjugation: term derived from a Latin word that means “to link together”, e.g. linking a small-molecule bait to a linker moiety that is connected to the anchor moiety Methotrexate (MTX): most popular anchor part of CIDs to date. MTX binds with high affinity to the dihydrofolate reductase (DHFR) hook and is amenable to chemical modification to present small-molecule baits to the cytosolic environment Off-target protein: unintended protein target bound by a particular therapeutic molecule, potentially mediating the side-effect caused by that drug Small-molecule drug: an organic drug compound having a molecular weight below 1000 Da, typically between 300 and 700 Da M

DOI: 10.1021/acschembio.5b00811 ACS Chem. Biol. XXXX, XXX, XXX−XXX

Reviews

ACS Chemical Biology protein stability measurements. Proc. Natl. Acad. Sci. U. S. A. 107, 9078−9082. (18) Savitski, M. M., Reinhard, F. B. M., Franken, H., Werner, T., Savitski, M. F., Eberhard, D., Molina, D. M., Jafari, R., Dovega, R. B., Klaeger, S., Kuster, B., Nordlund, P., Bantscheff, M., and Drewes, G. (2014) Tracking cancer drugs in living cells by thermal profiling of the proteome. Science 346, 1255784. (19) Bantscheff, M., Lemeer, S., Savitski, M. M., and Kuster, B. (2012) Quantitative mass spectrometry in proteomics: critical review update from 2007 to the present. Anal. Bioanal. Chem. 404, 939−965. (20) Cox, J., and Mann, M. (2011) Quantitative, High-Resolution Proteomics for Data-Driven Systems Biology. in Annu. Rev. Biochem., Vol 80 (Kornberg, R. D., Raetz, C. R. H., Rothman, J. E., and Thorner, J. W., Eds.), pp 273−299, Annual Reviews, Palo Alto. (21) Licitra, E. J., and Liu, J. O. (1996) A three-hybrid system for detecting small ligand-protein receptor interactions. Proc. Natl. Acad. Sci. U. S. A. 93, 12817−12821. (22) Caligiuri, M., Molz, L., Liu, Q., Kaplan, F., Xu, J. P., Majeti, J. Z., Ramos-Kelsey, R., Murthi, K., Lievens, S., Tavernier, J., and Kley, N. (2006) MASPIT: Three-hybrid trap for quantitative proteome fingerprinting of small molecule-protein interactions in mammalian cells. Chem. Biol. 13, 711−722. (23) Sche, P. P., McKenzie, K. M., White, J. D., and Austin, D. J. (1999) Display cloning: functional identification of natural product receptors using cDNA-phage display. Chem. Biol. 6, 707−716. (24) McPherson, M., Yang, Y. F., Hammond, P. W., and Kreider, B. L. (2002) Drug receptor identification from multiple tissues using cellular-derived mRNA display libraries. Chem. Biol. 9, 691−698. (25) Spencer, D. M., Wandless, T. J., Schreiber, S. L., and Crabtree, G. R. (1993) Controlling signal-transduction with synthetic ligands. Science 262, 1019−1024. (26) Fegan, A., White, B., Carlson, J. C. T., and Wagner, C. R. (2010) Chemically Controlled Protein Assembly: Techniques and Applications. Chem. Rev. 110, 3315−3336. (27) Clackson, T. Controlling Protein−Protein Interactions Using Chemical Inducers and Disrupters of Dimerization. In Chemical Biology; Wiley-VCH Verlag GmbH: Weinheim, 2008; pp 227−249. (28) Corson, T. W., Aberle, N., and Crews, C. M. (2008) Design and Applications of Bifunctional Small Molecules: Why Two Heads Are Better Than One. ACS Chem. Biol. 3, 677−692. (29) Rezwan, M., and Auerbach, D. (2012) Yeast ″N″-hybrid systems for protein-protein and drug-protein interaction discovery. Methods 57, 423−429. (30) Hamdi, A., and Colas, P. (2012) Yeast two-hybrid methods and their applications in drug discovery. Trends Pharmacol. Sci. 33, 109− 118. (31) Kley, N. (2004) Chemical dimerizers and three-hybrid systems: Scanning the proteome for targets of organic small molecules. Chem. Biol. 11, 599−608. (32) Crabtree, G. R., and Schreiber, S. L. (1996) Three-part inventions: Intracellular signaling and induced proximity. Trends Biochem. Sci. 21, 418−422. (33) Liu, J., Farmer, J. D., Lane, W. S., Friedman, J., Weissman, I., and Schreiber, S. L. (1991) Calcineurin is a common target of cyclophilincyclosporine-A and FKBP-FK506 complexes. Cell 66, 807−815. (34) Choi, J. W., Chen, J., Schreiber, S. L., and Clardy, J. (1996) Structure of the FKBP12-rapamycin complex interacting with the binding domain of human FRAP. Science 273, 239−242. (35) Rutkowska, A., and Schultz, C. (2012) Protein Tango: The Toolbox to Capture Interacting Partners. Angew. Chem., Int. Ed. 51, 8166−8176. (36) Gestwicki, J. E., and Marinec, P. S. (2007) Chemical control over protein-protein interactions: Beyond inhibitors. Comb. Chem. High Throughput Screening 10, 667−675. (37) Putyrski, M., and Schultz, C. (2012) Protein translocation as a tool: The current rapamycin story. FEBS Lett. 586, 2097−2105. (38) Fields, S., and Song, O. K. (1989) A novel genetic system to detect protein protein interactions. Nature 340, 245−246.

(39) Henthorn, D. C., Jaxa-Chamiec, A. A., and Meldrum, E. (2002) A GAL4-based yeast three-hybrid system for the identification of small molecule-target protein interactions. Biochem. Pharmacol. 63, 1619− 1628. (40) Ho, S. N., Biggar, S. R., Spencer, D. M., Schreiber, S. L., and Crabtree, G. R. (1996) Dimeric ligands define a role for transcriptional activation domains in reinitiation. Nature 382, 822−826. (41) Clemons, P. A., Gladstone, B. G., Seth, A., Chao, E. D., Foley, M. A., and Schreiber, S. L. (2002) Synthesis of calcineurin-resistant derivatives of FK506 and selection of compensatory receptors. Chem. Biol. 9, 49−61. (42) Borchardt, A., Liberles, S. D., Biggar, S. R., Crabtree, G. R., and Schreiber, S. L. (1997) Small molecule-dependent genetic selection in stochastic nanodroplets as a means of detecting protein-ligand interactions on a large scale. Chem. Biol. 4, 961−968. (43) Hussey, S. L., Muddana, S. S., and Peterson, B. R. (2003) Synthesis of a β-Estradiol-Biotin Chimera that Potently Heterodimerizes Estrogen Receptor and Streptavidin Proteins in a Yeast Three-Hybrid System. J. Am. Chem. Soc. 125, 3692−3693. (44) Muddana, S. S., and Peterson, B. R. (2004) Facile synthesis of CIDs: Biotinylated estrone Oximes efficiently heterodimerize estrogen receptor and streptavidin proteins in yeast three hybrid systems. Org. Lett. 6, 1409−1412. (45) Lin, H., Abida, W. M., Sauer, R. T., and Cornish, V. W. (2000) Dexamethasone−Methotrexate: An Efficient Chemical Inducer of Protein Dimerization In Vivo. J. Am. Chem. Soc. 122, 4247−4248. (46) Abida, W. M., Carter, B. T., Althoff, E. A., Lin, H. N., and Cornish, V. W. (2002) Receptor-dependence of the transcription readout in a small-molecule three-hybrid system. ChemBioChem 3, 887− 895. (47) Baker, K., Sengupta, D., Salazar-Jimenez, G., and Cornish, V. W. (2003) An optimized dexamethasone-methotrexate yeast 3-hybrid system for high-throughput screening of small molecule-protein interactions. Anal. Biochem. 315, 134−137. (48) Wingler, L. M., and Cornish, V. W. (2011) A Library Approach for the Discovery of Customized Yeast Three-Hybrid Counter Selections. ChemBioChem 12, 715−717. (49) Harton, M. D., Wingler, L. M., and Cornish, V. W. (2013) Transcriptional regulation improves the throughput of three-hybrid counter selections in Saccharomyces cerevisiae. Biotechnol. J. 8, 1485− 1491. (50) de Felipe, K. S., Carter, B. T., Althoff, E. A., and Cornish, V. W. (2004) Correlation between ligand-receptor affinity and the transcription readout in a yeast three-hybrid system. Biochemistry 43, 10353−10363. (51) Becker, F., Murthi, K., Smith, C., Come, J., Costa-Roldan, N., Kaufmann, C., Hanke, U., Degenhart, C., Baumann, S., Wallner, W., Huber, A., Dedier, S., Dill, S., Kinsman, D., Hediger, M., Bockovich, N., Meier-Ewert, S., Kluge, A. F., and Kley, N. (2004) A three-hybrid approach to scanning the proteome for targets of small molecule kinase inhibitors. Chem. Biol. 11, 211−223. (52) Lefurgy, S., and Cornish, V. (2004) Finding Cinderella after the ball: A three-hybrid approach to drug target identification. Chem. Biol. 11, 151−153. (53) Dirnberger, D., Unsin, G., Schlenker, S., and Reichel, C. (2006) A small-molecule-protein interaction system with split-ubiquitin as sensor. ChemBioChem 7, 936−942. (54) Caligiuri, M., Becker, F., Murthi, K., Kaplan, F., Dedier, S., Kaufmann, C., Machl, A., Zybarth, G., Richard, J., Bockovich, N., Kluge, A., and Kley, N. (2005) A proteome-wide CDK/CRK-specific kinase inhibitor promotes tumor cell death in the absence of cell cycle progression. Chem. Biol. 12, 1103−1115. (55) Toogood, P. L. (2005) The kinome is not enough. Chem. Biol. 12, 1057−1058. (56) Shepard, A. R., Conrow, R. E., Pang, I. H., Jacobson, N., Rezwan, M., Rutschmann, K., Auerbach, D., SriRamaratnam, R., and Cornish, V. W. (2013) Identification of PDE6D as a Molecular Target of Anecortave Acetate via a Methotrexate-Anchored Yeast ThreeHybrid Screen. ACS Chem. Biol. 8, 549−558. N

DOI: 10.1021/acschembio.5b00811 ACS Chem. Biol. XXXX, XXX, XXX−XXX

Reviews

ACS Chemical Biology (57) Odell, A. V., Tran, F., Foderaro, J. E., Poupart, S., Pathak, R., Westwood, N. J., and Ward, G. E. Yeast Three-Hybrid Screen Identifies TgBRADIN/GRA24 as a Negative Regulator of Toxoplasma gondii Bradyzoite Differentiation. PLoS One 2015, 10, e012033110.1371/journal.pone.0120331 (58) Tran, F., Odell, A. V., Ward, G. E., and Westwood, N. J. (2013) A Modular Approach to Triazole-Containing Chemical Inducers of Dimerisation for Yeast Three-Hybrid Screening. Molecules 18, 11639− 11657. (59) Gallagher, S. S., Miller, L. W., and Cornish, V. W. (2007) An orthogonal dexamethasone-trimethoprim yeast three-hybrid system. Anal. Biochem. 363, 160−162. (60) Bronson, J. E., Mazur, W. W., and Cornish, V. W. (2008) Transcription factor logic using chemical complementation. Mol. BioSyst. 4, 56−58. (61) Czlapinski, J. L., Schelle, M. W., Miller, L. W., Laughlin, S. T., Kohler, J. J., Cornish, V. W., and Bertozzi, C. R. (2008) Conditional glycosylation in eukaryotic cells using a biocompatible chemical inducer of dimerization. J. Am. Chem. Soc. 130, 13186−13187. (62) Gendreizig, S., Kindermann, M., and Johnsson, K. (2003) Induced Protein Dimerization in Vivo through Covalent Labeling. J. Am. Chem. Soc. 125, 14970−14971. (63) Gronemeyer, T., Chidley, C., Juillerat, A., Heinis, C., and Johnsson, K. (2006) Directed evolution of O-6-alkylguanine-DNA alkyltransferase for applications in protein labeling. Protein Eng., Des. Sel. 19, 309−316. (64) Chidley, C., Haruki, H., Pedersen, M. G., Muller, E., and Johnsson, K. (2011) A yeast-based screen reveals that sulfasalazine inhibits tetrahydrobiopterin biosynthesis. Nat. Chem. Biol. 7, 375−383. (65) Chidley, C., Haruki, H., Pedersen, M. G., Fellay, C., Moser, S., and Johnsson, K. (2011) Searching for the Protein Targets of Bioactive Molecules. Chimia 65, 720−724. (66) Moser, S., and Johnsson, K. (2013) Yeast Three-Hybrid Screening for Identifying Anti-Tuberculosis Drug Targets. ChemBioChem 14, 2239−2242. (67) Clackson, T. (1998) Redesigning small molecule-protein interfaces. Curr. Opin. Struct. Biol. 8, 451−458. (68) Pollock, R., and Clackson, T. (2002) Dimerizer-regulated gene expression. Curr. Opin. Biotechnol. 13, 459−467. (69) Pollock, R., Giel, M., Linher, K., and Clackson, T. (2002) Regulation of endogenous gene expression with a small-molecule dimerizer. Nat. Biotechnol. 20, 729−733. (70) Garabedian, M. J., and Yamamoto, K. R. (1992) Genetic dissection of the signaling domain of a mammalian steroid-receptor in yeast. Mol. Biol. Cell 3, 1245−1257. (71) Sitcheran, R., Emter, R., Kralli, A., and Yamamoto, K. R. (2000) A genetic analysis of glucocorticoid receptor signaling: Identification and characterization of ligand-effect modulators in Saccharomyces cerevisiae. Genetics 156, 963−972. (72) Hussey, S. L., He, E., and Peterson, B. R. (2002) Synthesis of Chimeric 7α-Substituted Estradiol Derivatives Linked to Cholesterol and Cholesterylamine. Org. Lett. 4, 415−418. (73) Appleman, J. R., Prendergast, N., Delcamp, T. J., Freisheim, J. H., and Blakley, R. L. (1988) Kinetics of the formation and isomerization of methotrexate complexes of recombinant human dihydrofolate-reductase. J. Biol. Chem. 263, 10304−10313. (74) Althoff, E. A., and Cornish, V. W. (2002) A bacterial smallmolecule three-hybrid system. Angew. Chem., Int. Ed. 41, 2327−2330. (75) Estojak, J., Brent, R., and Golemis, E. A. (1995) Correlation of 2-hybrid affinity data with in-vitro measurements. Mol. Cell. Biol. 15, 5820−5829. (76) Baccanari, D. P., Daluge, S., and King, R. W. (1982) Inhibition of dihydrofolate-reductase - effect of reduced nicotinamide adeninedinucleotide phosphate on the selectivity and affinity of diaminobenzylpyrimidines. Biochemistry 21, 5068−5075. (77) Miller, L. W., Cai, Y. F., Sheetz, M. P., and Cornish, V. W. (2005) In vivo protein labeling with trimethoprim conjugates: a flexible chemical tag. Nat. Methods 2, 255−257.

(78) Jing, C. R., and Cornish, V. W. (2013) A Fluorogenic TMP-Tag for High Signal-to-Background Intracellular Live Cell Imaging. ACS Chem. Biol. 8, 1704−1712. (79) Hinner, M. J., and Johnsson, K. (2010) How to obtain labeled proteins and what to do with them. Curr. Opin. Biotechnol. 21, 766− 776. (80) Wombacher, R., and Cornish, V. W. (2011) Chemical tags: applications in live cell fluorescence imaging. J. Biophotonics 4, 391− 402. (81) Jing, C. R., and Cornish, V. W. (2011) Chemical Tags for Labeling Proteins Inside Living Cells. Acc. Chem. Res. 44, 784−792. (82) Keppler, A., Gendreizig, S., Gronemeyer, T., Pick, H., Vogel, H., and Johnsson, K. (2002) A general method for the covalent labeling of fusion proteins with small molecules in vivo. Nat. Biotechnol. 21, 86− 89. (83) Hybrigenics Services. http://www.hybrigenics-services.com/ contents/our-services/discover/ultimate-ychemh. (84) Diver, S. T., and Schreiber, S. L. (1997) Single-step synthesis of cell-permeable protein dimerizers that activate signal transduction and gene expression. J. Am. Chem. Soc. 119, 5106−5109. (85) Freiberg, R. A., Ho, S. N., and Khavari, P. A. (1997) Transcriptional control in keratinocytes and fibroblasts using synthetic ligands. J. Clin. Invest. 99, 2610−2615. (86) Amara, J. F., Clackson, T., Rivera, V. M., Guo, T., Keenan, T., Natesan, S., Pollock, R., Yang, W., Courage, N. L., Holt, D. A., and Gilman, M. (1997) A versatile synthetic dimerizer for the regulation of protein-protein interactions. Proc. Natl. Acad. Sci. U. S. A. 94, 10618− 10623. (87) Rivera, V. M. (1998) Controlling gene expression using synthetic ligands. Methods 14, 421−429. (88) Keenan, T., Yaeger, D. R., Courage, N. L., Rollins, C. T., Pavone, M. E., Rivera, V. M., Yang, W., Guo, T., Amara, J. F., Clackson, T., Gilman, M., and Holt, D. A. (1998) Synthesis and activity of bivalent FKBP12 ligands for the regulated dimerization of proteins. Bioorg. Med. Chem. 6, 1309−1335. (89) Yang, W., Keenan, T. P., Rozamus, L. W., Wang, X. R., Rivera, V. M., Rollins, C. T., Clackson, T., and Holt, D. A. (2003) Regulation of gene expression by synthetic dimerizers with novel specificity. Bioorg. Med. Chem. Lett. 13, 3181−3184. (90) Pollock, R., and Rivera, V. M. (1999) in Regulation of gene expression with synthetic dimerizers. Methods in Enzymology (Glorioso, J. C., and Schmidt, M. C., Eds.), pp 263−281, Academic Press, Waltham. (91) Senner, V., Sotoodeh, A., and Paulus, W. (2001) Regulated gene expression in glioma cells: A comparison of three inducible systems. Neurochem. Res. 26, 521−524. (92) Belshaw, P. J., Ho, S. N., Crabtree, G. R., and Schreiber, S. L. (1996) Controlling protein association and subcellular localization with a synthetic ligand that induces heterodimerization of proteins. Proc. Natl. Acad. Sci. U. S. A. 93, 4604−4607. (93) Koide, K., Finkelstein, J. M., Ball, Z., and Verdine, G. L. (2001) A Synthetic Library of Cell-Permeable Molecules. J. Am. Chem. Soc. 123, 398−408. (94) Rivera, V. M., Clackson, T., Natesan, S., Pollock, R., Amara, J. F., Keenan, T., Magari, S. R., Phillips, T., Courage, N. L., Cerasoli, F., Holt, D. A., and Gilman, M. (1996) A humanized system for pharmacologic control of gene expression. Nat. Med. 2, 1028−1032. (95) Natesan, S., Molinari, E., Rivera, V. M., Rickles, R. J., and Gilman, M. (1999) A general strategy to enhance the potency of chimeric transcriptional activators. Proc. Natl. Acad. Sci. U. S. A. 96, 13898−13903. (96) Go, W. Y., and Ho, S. N. (2002) Optimization and direct comparison of the dimerizer and reverse tet transcriptional control systems. J. Gene Med. 4, 258−270. (97) Liberles, S. D., Diver, S. T., Austin, D. J., and Schreiber, S. L. (1997) Inducible gene expression and protein translocation using nontoxic ligands identified by a mammalian three-hybrid screen. Proc. Natl. Acad. Sci. U. S. A. 94, 7825−7830. O

DOI: 10.1021/acschembio.5b00811 ACS Chem. Biol. XXXX, XXX, XXX−XXX

Reviews

ACS Chemical Biology (98) Chen, X. C., Chong, C. R., Shi, L. R., Yoshimoto, T., Sullivan, D. J., and Liu, J. O. (2006) Inhibitors of Plasmodium falciparum methionine aminopeptidase 1b possess antimalarial activity. Proc. Natl. Acad. Sci. U. S. A. 103, 14548−14553. (99) Chen, X. C., Xie, S., Bhat, S., Kumar, N., Shapiro, T. A., and Liu, J. O. (2009) Fumagillin and Fumarranol Interact with P. falciparum Methionine Aminopeptidase 2 and Inhibit Malaria Parasite Growth In Vitro and In Vivo. Chem. Biol. 16, 193−202. (100) Risseeuw, M. D. P., De Clercq, D. J. H., Lievens, S., Hillaert, U., Sinnaeve, D., Van den Broeck, F., Martins, J. C., Tavernier, J., and Van Calenbergh, S. (2013) A ″Clickable″ MTX Reagent as a Practical Tool for Profiling Small-Molecule-Intracellular Target Interactions via MASPIT. ChemMedChem 8, 521−526. (101) Lievens, S., Gerlo, S., Lemmens, I., De Clercq, D. J. H., Risseeuw, M. D. P., Vanderroost, N., De Smet, A.-S., Ruyssinck, E., Chevet, E., Van Calenbergh, S., and Tavernier, J. (2014) Kinase Substrate Sensor (KISS), a Mammalian In Situ Protein Interaction Sensor. Mol. Cell. Proteomics 13, 3332−3342. (102) De Clercq, D. J. H., Risseeuw, M. D. P., Karalic, I., De Smet, A. S., Defever, D., Tavernier, J., Lievens, S., and Van Calenbergh, S. (2015) Alternative Reagents for Methotrexate as Immobilizing Anchor Moieties in the Optimization of MASPIT: Synthesis and Biological Evaluation. ChemBioChem 16, 834−843. (103) Balzi, E., and Goffeau, A. (1995) Yeast multidrug-resistance the PDR network. J. Bioenerg. Biomembr. 27, 71−76. (104) Kolaczkowski, M., Kolaczkowska, A., Luczynski, J., Witek, S., and Goffeau, A. (1998) In vivo characterization of the drug resistance profile of the major ABC transporters and other components of the yeast pleiotropic drug resistance network. Microb. Drug Resist. 4, 143− 158. (105) Rogers, B., Decottignies, A., Kolaczkowski, M., Carvajal, E., Balzi, E., and Goffeau, A. (2001) The pleitropic drug ABC transporters from Saccharomyces cerevisiae. J. Mol. Microbiol. Biotechnol. 3, 207− 214. (106) Eyckerman, S., Verhee, A., Van der Heyden, J., Lemmens, I., Van Ostade, X., Vandekerckhove, J., and Tavernier, J. (2001) Design and application of a cytokine-receptor-based interaction trap. Nat. Cell Biol. 3, 1114−1119. (107) Lievens, S., Peelman, F., De Bosscher, K., Lemmens, I., and Tavernier, J. (2011) MAPPIT: A protein interaction toolbox built on insights in cytokine receptor signaling. Cytokine Growth Factor Rev. 22, 321−329. (108) Haruki, H., Gonzalez, M. R., and Johnsson, K. (2012) Exploiting Ligand-Protein Conjugates to Monitor Ligand-Receptor Interactions. PLoS One 7, 11. (109) Keppler, A., Pick, H., Arrivoli, C., Vogel, H., and Johnsson, K. (2004) Labeling of fusion proteins with synthetic fluorophores in live cells. Proc. Natl. Acad. Sci. U. S. A. 101, 9955−9959. (110) Declercq, D. J. H., Tavernier, J., Lievens, S., and Van Calenbergh, S. Unpublished results. (111) Gallagher, S. S., Sable, J. E., Sheetz, M. P., and Cornish, V. W. (2009) An In Vivo Covalent TMP-Tag Based on Proximity-Induced Reactivity. ACS Chem. Biol. 4, 547−556. (112) Chen, Z., Jing, C., Gallagher, S. S., Sheetz, M. P., and Cornish, V. W. (2012) Second-Generation Covalent TMP-Tag for Live Cell Imaging. J. Am. Chem. Soc. 134, 13692−13699. (113) Wang, T. Y., Friedman, L. J., Gelles, J., Min, W., Hoskins, A. A., and Cornish, V. W. (2014) The Covalent Trimethoprim Chemical Tag Facilitates Single Molecule Imaging with Organic Fluorophores. Biophys. J. 106, 272−278. (114) Correa, I. R., Baker, B., Zhang, A. H., Sun, L., Provost, C. R., Lukinavicius, G., Reymond, L., Johnsson, K., and Xu, M. Q. (2013) Substrates for Improved Live-Cell Fluorescence Labeling of SNAP-tag. Curr. Pharm. Des. 19, 5414−5420. (115) Gautier, A., Juillerat, A., Heinis, C., Correa, I. R., Kindermann, M., Beaufils, F., and Johnsson, K. (2008) An engineered protein tag for multiprotein labeling in living cells. Chem. Biol. 15, 128−136. (116) Los, G. V., Encell, L. P., McDougall, M. G., Hartzell, D. D., Karassina, N., Zimprich, C., Wood, M. G., Learish, R., Ohane, R. F.,

Urh, M., Simpson, D., Mendez, J., Zimmerman, K., Otto, P., Vidugiris, G., Zhu, J., Darzins, A., Klaubert, D. H., Bulleit, R. F., and Wood, K. V. (2008) HaloTag: A novel protein labeling technology for cell imaging and protein analysis. ACS Chem. Biol. 3, 373−382. (117) Athavankar, S., and Peterson, B. R. (2003) Control of gene expression with small molecules: Biotin-mediated acylation of targeted lysine residues in recombinant yeast. Chem. Biol. 10, 1245−1253. (118) Bolin, J. T., Filman, D. J., Matthews, D. A., Hamlin, R. C., and Kraut, J. (1982) Crystal-structures of Escherichia-coli and Lactobacillus-casei dihydrofolate-reductase refined at 1.7 Å resolution. 1. General features and binding of methotrexate. J. Biol. Chem. 257, 13650−13662. (119) Wells, X. E., Bender, V. J., Francis, C. L., He-Williams, H. M., Manthey, M. K., Moghaddam, M. J., Reilly, W. G., and Whittaker, R. G. (1999) Tris and the ready production of drug-fatty acyl conjugates. Drug Dev. Res. 46, 302−308. (120) Rostovtsev, V. V., Green, L. G., Fokin, V. V., and Sharpless, K. B. (2002) A Stepwise Huisgen Cycloaddition Process: Copper(I)Catalyzed Regioselective “Ligation” of Azides and Terminal Alkynes. Angew. Chem., Int. Ed. 41, 2596−2599. (121) Hoyle, C. E., Lowe, A. B., and Bowman, C. N. (2010) Thiolclick chemistry: a multifaceted toolbox for small molecule and polymer synthesis. Chem. Soc. Rev. 39, 1355−1387. (122) Kolb, H. C., Finn, M. G., and Sharpless, K. B. (2001) Click chemistry: Diverse chemical function from a few good reactions. Angew. Chem., Int. Ed. 40, 2004−2021. (123) Ulrich, S., Boturyn, D., Marra, A., Renaudet, O., and Dumy, P. (2014) Oxime Ligation: A Chemoselective Click-Type Reaction for Accessing Multifunctional Biomolecular Constructs. Chem. - Eur. J. 20, 34−41. (124) Kö hn, M., and Breinbauer, R. (2004) The Staudinger LigationA Gift to Chemical Biology. Angew. Chem., Int. Ed. 43, 3106−3116. (125) Shangguan, N., Katukojvala, S., Greenburg, R., and Williams, L. J. (2003) The reaction of thio acids with azides: A new mechanism and new synthetic applications. J. Am. Chem. Soc. 125, 7754−7755. (126) Rijkers, D. T. S., Merkx, R., Yim, C.-B., Brouwer, A. J., and Liskamp, R. M. J. (2010) ‘Sulfo-click’ for ligation as well as for sitespecific conjugation with peptides, fluorophores, and metal chelators. J. Pept. Sci. 16, 1−5. (127) Lievens, S., Van der Heyden, J., Vertenten, E., Plum, J., Vandekerckhove, J., and Tavernier, J. (2004) in Design of a Fluorescence-Activated Cell Sorting-Based Mammalian ProteinProtein Interaction Trap. Flow Cytometry Protocols (Hawley, T., and Hawley, R., Eds.) pp 293−309, Humana Press, New York. (128) Yang, X. P., Boehm, J. S., Yang, X. P., Salehi-Ashtiani, K., Hao, T., Shen, Y., Lubonja, R., Thomas, S. R., Alkan, O., Bhimdi, T., Green, T. M., Johannessen, C. M., Silver, S. J., Nguyen, C., Murray, R. R., Hieronymus, H., Balcha, D., Fan, C. Y., Lin, C. W., Ghamsari, L., Vidal, M., Hahn, W. C., Hill, D. E., and Root, D. E. (2011) A public genomescale lentiviral expression library of human ORFs. Nat. Methods 8, 659−661. (129) Lievens, S., Declercq, D., Van Calenbergh, S., and Tavernier, J. Unpublished results. (130) Remy, I., and Michnick, S. W. (2007) Application of proteinfragment complementation assays in cell biology. BioTechniques 42, 137−145. (131) Michnick, S. W., Ear, P. H., Manderson, E. N., Remy, I., and Stefan, E. (2007) Universal strategies in research and drug discovery based on protein-fragment complementation assays. Nat. Rev. Drug Discovery 6, 569−582. (132) Shekhawat, S. S., and Ghosh, I. (2011) Split-protein systems: beyond binary protein-protein interactions. Curr. Opin. Chem. Biol. 15, 789−797.

P

DOI: 10.1021/acschembio.5b00811 ACS Chem. Biol. XXXX, XXX, XXX−XXX