full text (pdf) - plant · pdf filetopical review on protein-protein interaction techniques...

Topical Review on Protein-Protein Interaction Techniques

Techniques for the Analysis of Protein-ProteinInteractions in Vivo1[OPEN]

Shuping Xing, Niklas Wallmeroth, Kenneth W. Berendzen, and Christopher Grefen*

University of Tübingen, ZMBP Developmental Genetics (S.X., N.W., C.G.) and ZMBP Central Facilities(K.W.B.), D–72076 Tuebingen, Germany

ORCID ID: 0000-0002-5820-4466 (C.G.).

Identifying key players and their interactions is fundamental for understanding biochemical mechanisms at the molecular level.The ever-increasing number of alternative ways to detect protein-protein interactions (PPIs) speaks volumes about the creativity ofscientists in hunting for the optimal technique. PPIs derived from single experiments or high-throughput screens enable thedecoding of binary interactions, the building of large-scale interaction maps of single organisms, and the establishment of cross-species networks. This review provides a historical view of the development of PPI technology over the past three decades,particularly focusing on in vivo PPI techniques that are inexpensive to perform and/or easy to implement in a state-of-the-artmolecular biology laboratory. Special emphasis is given to their feasibility and application for plant biology as well as recentimprovements or additions to these established techniques. The biology behind each method and its advantages and disadvantagesare discussed in detail, as are the design, execution, and evaluation of PPI analysis. We also aim to raise awareness about thetechnological considerations and the inherent flaws of these methods, which may have an impact on the biological interpretation ofPPIs. Ultimately, we hope this review serves as a useful reference when choosing the most suitable PPI technique.

From protein synthesis, maturation, and degradationto vesicle budding, trafficking, and fusion and fromreceptor dimerization, signaling cascades, and generegulation to metabolism and catabolism, almost allcellular functions depend on and are executed by com-plex protein-protein interactions (PPIs). While com-plete biochemical pathways and even some proteincomplexes involved in these processes had beendeciphered by the 1960s, the knowledge about PPIsat that time was limited. Owing to the fledgling stateofmolecular biology in those days,methods for studyingPPIs and the development of such techniques haveemerged only slowly. The advent of gene fusion andprotein labeling, together with the engineering of flu-orescent probes covering the entire spectrum of light,provided the initial spark for a technological revolutionin the study of PPIs (Braun and Gingras, 2012).

In this Topical Review, we provide an overview ofcurrent tried-and tested-tools for investigating themechanistic relationships between potential proteinpartners that are easily accessible to a plant scientistworking in a standard molecular biology laboratory.Each of these methods comes with its own set of ben-efits and drawbacks. A detailed introduction to the bi-ology underlying each method, novel improvements,and potential pitfalls are included, allowing readers tochoose the most suitable method for their research.Some actual examples from the plant sciences havebeen included, but a detailed discussion of the biological

questions that were addressed has been omitted, as thetechniques take center stage in this review.

In order to help researchers decide which method ismost suitable, we summarize the main properties andissues inherent to each technique featured in this review(Table I; Fig. 1). In general, the following points shouldbe considered when designing a PPI analysis method.

(1) Technology. Each PPI technique utilizes a uniqueconcept for combining probes and reporting inter-actions. The mode of action of a PPI technique caninfluence the likelihood of a positive readout of agiven protein couple as well as its interpretation.

(2) Interaction dynamics. Interactions between proteinscan be weak or strong, transient or stable, non-permanent or permanent. Kinetics, thermodynamics,stoichiometry, and cofactors can affect a PPI andtherefore are important factors to consider prior toPPI analysis and when choosing an appropriate de-tection system.

(3) Environmental conditions. Non-native environments(e.g. yeast) can severely affect the expression, trans-lation, subcellular localization, posttranslationalmodifications, lifetime, and turnover of a proteinand therefore may preclude positive interaction anal-ysis. Verifying protein expression by immunoblotanalysis is imperative, especially when a negative in-teraction is reported.

1 This work was supported by an EmmyNoether fellowship of theDeutsche Forschungsgemeinschaft (DFG) to C.G. (GR 4251/1-1).

* Address correspondence to [email protected].

[OPEN] Articles can be viewed without a subscription.www.plantphysiol.org/cgi/doi/10.1104/pp.16.00470

Critical discussion of technological limitations andadvantages of the most prominent in vivo protein-protein interaction techniques emphasizing their

application to plant research

Plant Physiology�, June 2016, Vol. 171, pp. 727–758, www.plantphysiol.org � 2016 American Society of Plant Biologists. All Rights Reserved. 727 www.plantphysiol.orgon May 22, 2018 - Published by Downloaded from

Copyright © 2016 American Society of Plant Biologists. All rights reserved.

http://orcid.org/0000-0002-5820-4466

mailto:[email protected]

mailto:[email protected]

http://www.plantphysiol.org/cgi/doi/10.1104/pp.16.00470

http://www.plantphysiol.org

(4) Expression control. The overexpression of a proteinmight be necessary to enable the detection of anotherwise weak or transient interaction, but it alsomight lead to artificial results. Balancing expressionthrough induction or repression is crucial for mean-ingful analysis.

(5) Reporter output. Information about an interactionis quickly conveyed through the action of a reporter(directly through the output of probes, such asfluorescence) or with a temporal delay (indirectlythrough transcriptional regulation, such as reportergene expression).

(6) Controls. A positive interaction is only credible if acorresponding negative control using a closely re-lated ortholog or point mutation of the original pro-tein is performed. Control proteins should localizeto identical compartments, turnover and maturationmust not be different, and their spatiotemporal dis-tribution with the binding partners must be equalcompared with the positive interactor. It is crucialto analyze the same protein couple using at leasttwo (or preferably three) different PPI techniques.

The PPI methods described in this review includemethods that are preferentially used in the plant sci-ences (Figs. 1 and 2C), namely, the yeast two-hybridsystem, the split-ubiquitin system, FRET-related tech-niques, BiFC, and the split-luciferase complementationassay. CoIP, an ex vivo approach that is a viable alter-native to any of these techniques, is also described indetail.

BACKGROUND

Figure 2A highlights the increase in the number ofscientific articles that feature PPIs as either a topic ortitle keyword and were published from the last decadeof the 20th century to the present. While the mention ofPPIs in scientific publications (PPI as topic) has in-creased almost exponentially in the past 20 years, thenumber of articles featuring the term PPI in the title(likely articles that focus on technology development)has increased in a steady, linearmanner. The increase inpublications coincides with three major technologicaland scientific advances: (1) the invention of the yeast(Saccharomyces cerevisiae) two-hybrid system in 1989; (2)the cloning ofGFP in 1992; and (3) the publication of theyeast genome sequence in 1996.

The yeast two-hybrid method, developed by StanleyFields and Ok-kyu Song, was the first and remains oneof the most prominent systems for detecting PPIs(Fields andSong, 1989; Fields, 2009). As of February 2016,a keyword search in Thomson Reuters Web of Scienceyielded almost 60,000 publications with the topic yeasttwo-hybrid (Table II). The invention of the yeast two-hybrid technique not only triggered thousands ofstudies on PPIs but also spearheaded the development

of alternative PPI methods and, in conjunction withthe availability of fully sequenced genomes and high-throughput PPI techniques, initiated the subdisciplineof interactomics (see Box I). While the annual numberof publications featuring yeast two-hybrid methodsappears to have stalled recently (Fig. 2B), this techniqueremains the uncontested number-one PPI method ofchoice in plant research (Fig. 2C; Table II).

An even bigger technological breakthrough occurredat the beginning of the 1990s, which was a game changerfor the development of in vivo PPI techniques and cellbiological methods in general. Toward the end of the1960s, the mechanisms underlying and components fa-cilitating the phenomenon of bioluminescence were dis-covered. Among the first substances to be isolated wasthe luciferase aequorin, a light-producing enzyme fromthe jellyfish Aequorea victoria (Shimomura et al., 1962).Intriguingly, while the living jellyfish emits green light,the Ca2+-dependent oxidation of the substrate coelenter-azine, which aequorin catalyzes, leads to the emission ofblue light in vitro. From the extracts of jellyfish, OsamuShimomura also purified a very brightly autofluorescingprotein, which he named GFP. He postulated “an energytransfer from the light emitter of aequorin to GFP.” Inother words, the energy of the blue light produced bythe chemiluminescent aequorin is transferred to GFP,which in turn emits red-shifted green light (Moriseet al., 1974). In fact, this proposed mechanism proba-bly represents the first description of bioluminescenceresonance energy transfer, which is currently used as atechnique to study PPIs (Boute et al., 2002). Biolumi-nescence resonance energy transfer combines a light-producing luciferase, an energy-receiving fluorescentprotein, and the phenomenon of FRET, a theory pos-tulated some 30 years earlier by the physicist TheodorFörster (Förster, 1946).

Work in the 1980s and 1990s focused on cloningand recombinant expression of aequorin (Prasheret al., 1985, 1992). Doug Prasher, together with RogerTsien and Roger Heim, started engineering artificialfluorescent proteins by introducing mutations in theoriginal GFP (Heim et al., 1994; Heim and Tsien,1996). These novel fluorescent proteins revolution-ized our understanding of subcellular processes,enabling biologists to label organelles, perform liveimaging of subcellular protein movement, measureCa2+ concentrations, pH, redox state, and membranevoltage, and, most importantly, monitor PPIs in vivoand in real-time (Ehrhardt, 2003; Fricker et al., 2006).

The engineering of fluorescent proteins with differentspectral properties also enabled FRET to be used tostudy PPIs in vivo some 50 years after its initial de-scription. As shown in Figure 2B, by the end of the 20thcentury and in the wake of the innovations that fol-lowed the development of fluorescent proteins, thenumber of publications featuring FRET as a keywordincreased rapidly and continued to do so. Notably, FRETis exploited not only to detect PPIs but also to determinethe properties of the protein environment as unimolec-ular protein sensors. FRET has a range of applications in

728 Plant Physiol. Vol. 171, 2016

Xing et al.

www.plantphysiol.orgon May 22, 2018 - Published by Downloaded from Copyright © 2016 American Society of Plant Biologists. All rights reserved.


physics, chemistry, and materials science, which alsoaccounts for the large number of publications.

Categories of PPI Techniques

Since these groundbreaking discoveries, dozens ofnovel tools or improvements to established techniques

have been developed for the detection of PPIs. Thesetechniques have been categorized in various ways, suchas in vivo versus in vitro, biophysical versus biochemi-cal, split complementation versus two hybrid, and (withregard to the reporter output) enzymatic (e.g. chromo-genic or prototrophic) versus fluorescent markers. Flu-orescent reporters refer to inherent or complemented

Table I. Properties of the in vivo PPI techniques discussed in this review

MethodOrganism/

SystemReporter

Risk of Artifactsa

Quantificationb

Large Scaleb Comments

False

Positives

False

Negatives HT

Unbiased

Screening Pro Contra

Yeast two-hybrid Yeast Survival ondepletedmedium orenzymatic

++ ++ ++ +++ +++ Easy to perform,inexpensive

Truncation andmislocalizationof proteinsnecessary,limited tonucleus

Split-ubiquitinsystem

Yeast Survival ondepletedmedium orenzymatic

++ ++ ++ +++ +++ Easy to perform,inexpensive,virtually allproteins can betested,

Heterologousexpression canprevent proteinexpression,probes need toface cytosol

FRET-SE Plant Fluorescence + ++ ++ + + Interaction sitevisualized,easy toperform

Spectral bleedthrough,bleaching,concentrationdependence

FRET-AB Fluorescenceincrease indonor uponacceptorbleaching

2 +++ ++ 2 2 Interaction sitevisualized,intrinsiccontrol

Not suitable foranalyzing fast-movingspecimens,cytotoxicitythroughbleaching

FRET-FLIM Fluorescencedecay

2 +++ +++ 2 2 Interaction sitevisualized, PPIindependent ofconcentration,dynamicanalysispossible

Expensiveequipment andconsiderabletrainingnecessary

BiFC Plant Fluorescence +++ + +c ++ + Easy to perform,allowstopologyprediction

Many effectsoverlay PPI, invivo cross-linking

Split-luciferase Plant Bioluminescence,enzymatic

+ ++ ++ ++ + Reversibility ofprobeinteractionallows kineticand dynamicanalysis

Exogenoussubstrate

application

CoIP Plant Immunostaining ++ ++ + 2 +++d Easy to perform,theoreticallypossible toperformwithout tags

Not suitable fortransientinteractions,difficult to usemembraneproteins

HT, High-throughput; SE, sensitized emission; AB, acceptor photobleaching aEstimates range from low (2), moderate (+), and high (++) to veryhigh (+++). bRanges are from not applicable (2), possible (+), and suitable (++) to highly suitable (+++). cCredible, unbiased quantificationimpossible using classical BiFC; reliable semiquantitative detection of PPIs possible using the ratiometric BiFC (Grefen and Blatt, 2012a).dScreening for novel interactors possible when combining CoIP with mass spectrometry (MS) analysis.

Plant Physiol. Vol. 171, 2016 729

In Vivo Protein-Protein Interaction Techniques



Figure 1. Protein-protein interaction techniques. I, Two-hybrid techniques (2H) use two functional proteins or domains asprobes. Their inherent functionality is symbolized by dashed arrows, which can represent either DNA-binding or tran-scriptional activity for the GAL4 DNA-binding and transactivation domains in yeast two-hybrid analysis (A) or fluorescence


Xing et al.



fluorescence as opposed to enzymatic, chromogeniccolor, or fluorescence output, which requires thegeneration of substrate turnover (luciferase or lacZexpression).The in vivo approach (the focus of our attention) as-

sumes that the host organism inwhich the PPI takes placeis still alive when the analysis is performed. In sometechniques involving the use of prototrophic reporters,such as in yeast-based systems, the survival of the orga-nism is both a prerequisite and proof of the detection of aninteraction. However, the greatest advantage of in vivotechniques is that they preserve the native surroundingsin which the interaction takes place and is monitored.PPI techniques that involve fluorescent reporter

output are often performed in living cells as well, but inprinciple, the measurements do not necessarily requirelive-cell imaging. Immunolabeling is a feasible way offixing an otherwise fast-moving or transient interaction,but it involves terminating the life of the cell. Clearly,this particular finishing touch should not alter thetechnique’s in vivo label. After all, the interaction isallowed to take place in vivo, and only during the de-tection of this interaction is death of the cell acceptable.This argument is self-evident when we look at CoIP,which can be seen as a chimera of in vitro and in vivotechniques: while the detection of an interaction clearlytakes place in vitro in the true sense of the word, theestablishment of the interaction itself is allowed to oc-cur in an in vivo (i.e. living organism) environment.Accordingly, this approach is referred to as ex vivo.Inherent to most PPI techniques is the fusion of two

separate molecules (probes) attached to the proteins ofinterest (POIs), leading to their reporting of the inter-action. There are two categories of probes based on thenature of their reassembly or interaction: probes usedfor PCA versus two-hybrid analyses (Fig. 1). The latteris naturally connected to the popular yeast two-hybridmethod (Fields and Song, 1989). As the name implies, theprobes for the two-hybrid system contain two discretedomains, each of which is functional in its own right,whereas the PCA probes consist of two nonfunctionalfragments of a reporter protein; its complementation in-dicates a positive interaction (Stynen et al., 2012).

The split-ubiquitin system is the pioneering, classicexample of a PCA method (Johnsson and Varshavsky,1994; Müller and Johnsson, 2008), a categorization that isimportant for obtaining an in-depth understanding of themethods, their underlying biology, and potential limita-tions. While the two-hybrid system in principle can beconverted to a yeast one-hybrid system, allowing thedetection of protein-DNA interactions (Gstaiger et al.,1995), the PCA approach requires both split reporterfragments for a meaningful output; each fragment byitself is non-functional, an important difference. In recentyears, the split-ubiquitin system has been referred to as amembrane yeast two-hybrid method (Kittanakom et al.,2009),which is biologicallymisleading, as the split-ubiquitinsystem is aPCAandnot a two-hybrid technique.Worse still,this label downplays the split-ubiquitin system as anindependent method that is superior in some aspects tothe two-hybrid approach.

General Considerations about PPIs

The expression of POIs in yeast, as well as in in planta-based assays, is often driven by constitutive, strong pro-moters. Aberrant amounts of protein in the cell typeexamined may be a consequence of the use of such strongpromoters. For yeast-based techniques, this problem maybe negligible, considering that the PPI is monitored in aheterologous environment and is already flawed due toissues with codon bias, posttranslational modifications,and protein folding, stability, and turnover. However, forin planta approaches,where the advantage of homologousexpression is implicit, using endogenouspromoters shouldhelp reduce artifacts associated with aberrant expression.

From a thermodynamics perspective, it is evidentthat the concentration of a protein is a major drivingforce for any PPI (see Box II); hence, when artificiallyraising cellular protein concentrations, non-interactorsmight be forced to associate. Interaction data based onthe use of protein overexpression should always becritically evaluated and compared between at least two(or better yet three) different methods. Nevertheless,overexpression or molecular crowding might still be

Figure 1. (Continued.)upon excitation such as in Forster resonance energy transfer (FRET) sensors (B). The proximity of the probes suffices forreporter output to be detected, as indicated by the black arrow. A, Example of the yeast two-hybrid technique and its reporter output(i.e. growth on depleted [interaction-selective] medium). AD, Activation domain; BD, DNA-binding domain; UAS, upstream acti-vating sequence. B, Cartoon of a typical FRETinteraction and its visualization through sensitized emission (see “FRET: LifetimeVersusIntensity” for details). C, Strictly speaking, coimmunoprecipitation (CoIP) can be counted as a two-hybrid technique when both baitandpreyare fused to epitope tags (e.g.GFP). An interaction is reported through immunoblotting of the extract prior to column loading(L) and, after washing, the eluate (IP). M, Marker. II, Protein fragment complementation assays (PCAs), such as the split-ubiquitinsystem (D), bimolecular fluorescence complementation (BiFC; E), and split-luciferase complementation (F), use two non-functionalprotein fragments as probes. Only upon reassembly is their intrinsic function (enzymatic, fluorescence, and so on) restored, which isused to reveal an interaction (black arrow). D, Diagram of a split-ubiquitin assay; interactions are monitored through growth ondepleted medium, which is similar to yeast two-hybrid analysis. E, Cartoon of BiFC analysis and its (suboptimal) detection throughconfocal microscopy. See “BiFC: From Cherry Picking to Quantification” for further details on the appropriate analysis, quantifi-cation, and controls for this particular PPI technique. F, Split-luciferase complementation assay and its application on a 96-well plate,which enables high-throughput analysis of plant protoplasts by directly measuring luminescence after the addition of substrate(luciferin). + and 2 refer to typical positive and negative interactions, respectively, throughout.





circumvented in an in vivo environment through genesilencing or targeted degradation, attenuating potentialoccurrences of false-positive interactions.

It is crucial to check the native spatiotemporal ex-pression pattern of a protein and its supposed partner.This includes its expression in particular tissues and atspecific developmental stages as well as in particularsubcellular compartments. Only if the two POIs have thechance tomeet at some point in time or space is it sensibleto perform PPI analysis under non-native conditions.

We now summarize a few general points that shouldbe considered before initiating work with any of the de-scribed PPI techniques. Particular attention is given to thefollowing factors, which are likely to influence the de-tection of a PPI. Organismal factors are particularly im-portant, as heterologous expression can be influenced bycodon bias and posttranslational modifications or thelack thereof. Construct generation (cloning system), ex-pression strength (choice of promoters, replication ori-gin), maintenance in the organism of choice (marker),and probe-related issues (linkers, tags and their orienta-tion) can be summarized as structural factors. Equallyimportant for successful and reliable interaction analysisare experimental factors, such as the need to includerigorous controls, assess their suitability, and choose al-ternative methods to corroborate the results.

Organismal Factors

Expressing a plant protein heterologously in yeast, forexample, can result in low translation efficiency due tosuboptimal codon usage (Grefen, 2014). Codon optimi-zation, however, can be a two-edged sword, as it was re-cently shown that rare codons found in gene sequences ofsecretory and membrane proteins can slow down proteinsynthesis, thereby improving cotranslational translocation(Chartier et al., 2012; Pechmann et al., 2014; Yu et al., 2015).More worrying still is the recent finding that using dif-ferent synonymous codons modulates translation rates

Table II. Total number of publications using keywords referring to PPItechniques discussed in this review as of February 2016 (Fig. 2C)

Keyword search string abbreviations are as follows: Y2H, yeast two-hybrid; SUS, split-ubiquitin; FRET, Forster resonance energy transfer orfluorescence resonance energy transfer; SLCA, split-luciferase com-plementation assay.

Source Y2H SUS FRET BiFC SLCA

Plant science journalsa 3,318 220 506 1,174 39Web of Science 58,153 381 77,375 5,372 141

aUsing the search functions of the following journals: The Plant Cell,Plant Physiology, The Plant Journal, Molecular Plant, Journal of Exper-imental Botany, New Phytologist, Plant and Cell Physiology, andPlant, Cell & Environment.

Figure 2. Publication statistics on PPI methods (as of February 2016). A,Total number of publications that feature the search string “protein-protein interaction(s)” either as the topic keyword (gray surface plot,axis on left) or in the title (dotted line, axis on right), as analyzed usingthe Thomson Reuters Web of Science database. Vertical arrows markimportant technological or scientific discoveries and their publicationdate by year. B, Number of scientific publications per year in which anyof the five PPI techniqueswerementioned as a topic keyword (ThomsonReutersWeb of Science). The arrow highlights the year in which the firstGFP-derived FRET probes were reported (Heim and Tsien, 1996). Thegray-shaded area corresponds to the total number of all publicationsper year (S). C, Number of scientific publications featuring any of thecorresponding PPI techniques in the top eight plant science journals(TPC = The Plant Cell, PP = Plant Physiology, TPJ = The Plant Journal,MP = Molecular Plant, JXB = Journal of Experimental Botany, NP =New Phytologist, PCP = Plant and Cell Physiology, and PCE = Plant,Cell & Environment). The following keyword search strings for the ab-breviated methods in B and C were used: Y2H = yeast two-hybrid;

SUS = split-ubiquitin; FRET = Forster resonance energy transfer or flu-orescence resonance energy transfer; BiFC = bimolecular fluorescencecomplementation; and SLCA = split-luciferase.


Xing et al.



and alters the cotranslational folding of protein domains(Buhr et al., 2016). These points emphasize the importanceof biochemical verification of protein expression especially(but not only) in negative controls, as the failure to reportan interaction does not necessarily mean that a proteincouple does not interact. Negative results could simply bedue to modified expression or even a lack thereof.Equally problematic in heterologous expression sys-

tems is the need for correct posttranslational modifi-cations (Duan and Walther, 2015; Hurst and Hemsley,2015). A plethora of posttranslational modificationsare available, such as phosphorylation, glycosylation,and lipidation, all of which depend on recognition ofthe host organism by the molecular machinery. Post-translational modifications can influence a PPI eitherdirectly through the modification of interaction sites(e.g. phosphorylation) or indirectly by shifting lo-calization (e.g. palmitoylation; Hurst and Hemsley,2015). Another important posttranslational modifi-cation is the glycosylation pattern. Each eukaryotickingdom has its own glycosylation signature, whichlikely influences the functionality and interaction pat-terns of recombinantly expressed proteins (Dell et al.,

2010). This may result in the failure to detect a potentialPPI due to either the absence or incorrect glycosylationsignature in a POI or to its mistargeting, misfolding,and degradation, highlighting the fundamental im-portance of testing PPIs with (at the very least) twodifferent techniques.

Structural Factors

Vector Systems

Numerous vector systems are available for bothyeast- and plant-based PPI techniques. Comprehensivelists of vector systems for each technique can be foundin reviews that focus on individual methods. Ratherthan trying to assemble a list of available vectors foreach system, we would like to emphasize that the in-troduction of novel cloning techniques has made itquite easy to construct expression cassette/vector sys-tems for any of the PPI techniques listed here. Ulti-mately, it might be less time consuming to producesuch systems than to receive outdated plasmids with

Box I: An Omics Perspective of PPIs

The relatively low cost for sequencing whole genomes and comparing them between ecotypes and species allows for rapid data accumulation.Changes in transcript levels or epigenetic signatures enable in-depth views of evolutionary or developmental effects to be obtained. Thesuitability of some PPI techniques for high-throughput and/or large-scale approaches, such as CoIP-MS (Miernyk and Thelen, 2008) and yeasttwo-hybrid analysis, has led to the emergence of another omics discipline: interactomics.

This term, which was originally coined by Bernard Jacq in 1999 (Sanchez et al., 1999), describes the effort to assemble virtually all binary PPIsin an organism and/or compare them with those of other species (Kiemer and Cesareni, 2007). Whether it is sufficient to look exclusively atpairwise interactions is debatable, but such analysis is the easiest way to begin monitoring network structures. Interactome data sets can beacquired in silico through prediction algorithms (based on physicochemical properties, evolutionary conservation, coexpression analysis, or acombination of these, including integration of functional data) or experimentally via high-throughput PPI techniques (Morsy et al., 2008; Braunet al., 2013).

Again, the yeast two-hybrid system has pioneered this particular side branch of PPI studies. Yeast is an extremely powerful organism for high-throughput PPI analysis either through mating-based approaches or sequential cloning of cDNA libraries. It is also easy to collect and identifypositive interactions, since survival of the yeast can be used as a reporter. Nonetheless, every method comes with flaws such as the detection offalse positives (which will be discussed in detail below), and applying these techniques at a large scale inadvertently multiplies inherent issues,raising valid concerns over the authenticity of large lists of interactors (Hart et al., 2006).

While the initial screens and arrays (Uetz et al., 2000; Ito et al., 2001) were predicted to contain almost 50% false positives (von Mering et al.,2002), recent improvements have yielded more reliable interaction data (Yu et al., 2008). Nevertheless, interactome data should be treatedcritically: they should be verified experimentally for individual candidates, and only when confirmed should they be used to inform hypothesisgeneration. There are a number of online resources where interactome data can be accessed (Table Box I; Szklarczyk and Jensen, 2015). Forexample, the interactomes of Arabidopsis (Geisler-Lee et al., 2007; Lee et al., 2015), rice (Oryza sativa; Gu et al., 2011; Sapkota et al., 2011), andmaize (Zea mays; Zhu et al., 2016) have been assembled through prediction algorithms and experimental evidence or a combination of the two.

Screening Approaches

Using PPI techniques to screen for novel interaction partners of a POI is many scientists’ preferred way to venture into new avenues ofresearch. Unbiased approaches to obtain novel interactors of a POI include cDNA library screens using the yeast two-hybrid system (Gietz,2006), the split-ubiquitin system (Dirnberger et al., 2008), and the (now more affordable) CoIP-MS technique (Pertl-Obermeyer et al., 2014). It isimperative, however, to critically test an interaction couple individually and independently (with alternative techniques) before investing toomuch time and effort in analysis.

Hunting for novel binding partners also can begin with a limited number of putative candidates using large but defined collections (arrays) ofopen reading frames as potential preys (Miller et al., 2005; Lalonde et al., 2010). These arrays feature a defined choice of protein groups orsubfamilies, the screening of which has been executed successfully with different techniques, notably the membrane proteome of Arabidopsis usingthe mating-based split-ubiquitin system (Jones et al., 2014) and CoIP-MS to analyze SNARE protein interactions (Fujiwara et al., 2014b).





incomplete sequence information or verification. Pro-ducing such vectors also would enable the researcher tocombine the latest improvements (tags, fluorophores)and choose the best promoter (endogenous, constitu-tive, repressible, or inducible) while, most importantly,enabling dual or triple expression from one stretch ofDNA (plasmid or transfer DNA [T-DNA]), optimizingthe desired PPI technique for the POIs. With a range ofnovel, rapid, easy-to-use multicloning techniques tochoose from, such as GreenGate (Lampropoulos et al.,2013), GoldenGate (Engler et al., 2008), Gibson Assembly(Gibson et al., 2009), Multisite Gateway (Sasaki et al.,2004), and the 2in1 system (Grefen and Blatt, 2012a),building a multiexpression cassette has become an at-tractive option. In general, multiple gene expression froma single plasmid should be the preferred choice, espe-cially in any transient (plant) transformation. For exam-ple, mixing of two individual Agrobacterium tumefaciensstrains only yields approximately 80% cotransfection, inaddition to high expression variability, which is mostlikely a consequence of unequal gene dosage (see Fig. 6Abelow; Hecker et al., 2015).

In yeast, it is possible to coexpress two fusion proteinsfrom two individual plasmids through the use of twodifferent prototrophic selection markers. Additionally,

as yeast is capable of homologous recombination (invivo cloning), it is possible to create almost any desiredvector or construct (Joska et al., 2014), which is ex-tremely useful when working with genes that appear tobe toxic. For example, for unknown reasons, someDNAsequences, such as that of Arabidopsis thaliana HistidineKinase 5 (AHK5; Mira-Rodado et al., 2012), causeproblems during cloning in Escherichia coli, resulting intransposon-inserted or point-mutated open readingframes (Horák et al., 2008). Strategies to counter theseissues include using an E. coli strain that reduces thecopy number during log-phase growth (CopyCutter orBiozym) or simply performing in vivo cloning in yeast.

Expression Control: Promoters and Replication Origins (ORIs)

Choosing a promoter to control heterologous geneexpression is of great importance, particularly in yeast,where overexpression can potentially interfere withmetabolism, leading to lethality or other strong phe-notypes. A number of different promoter systems havebeen established for yeast, ranging from repressible orinducible to weak or strong constitutive expression.

Most systems feature the use of the alcohol dehydro-genase promoter, a strong constitutive promoter that is

Box I Table: Large-scale interactome analyses and databases

Plant Species Method Objective of the Study No. of Interactions Reference

Arabidopsis thaliana Yeast two-hybrid Aux/IAA-ARF 213 Piya et al. (2014)Affinity purification-MS Interactome of A. thaliana Qa-SNAREs 518 Fujiwara et al. (2014)Yeast two-hybrid Abscisic acid signaling network .500 Lumba et al. (2014)Computational Pseudomonas syringae interactome 11,000 Sahu et al. (2014)Split-ubiquitin system Membrane-linked Interactome

Database (MIND1)12,102 Jones et al. (2014b)

Computational Global interactome 3,432 Yang et al. (2013a)Yeast two-hybrid TOPLESS interactome 655 Causier et al. (2012)Computational PAIR (predicted Arabidopsis

interactome resource)149,900 Lin et al. (2011)

Yeast two-hybrid PPIN-1 (plant-pathogen immunenetwork)

3,148 Mukhtar et al. (2011)

Affinity purification-MS 14-3-3 family 129 Swatek et al. (2011)Yeast two-hybrid G-protein interactome 1,058 Klopffleisch et al. (2011)Affinity purification-MS Cell cycle 857 Van Leene et al. (2010)Yeast two-hybrid + BiFC Cell cycle 357 Boruc et al. (2010)Protein arrays MAPK target networks 1,280 Popescu et al. (2009)Affinity purification-MS 14-3-3 family 130 Chang et al. (2009)Computational AtPID (A. thaliana protein

interactome database)28,062 Cui et al. (2008)

Protein arrays Calmodulins 716 Popescu et al. (2007)Yeast two-hybrid MADS box 272 de Folter et al. (2005)

Brassica rapa Computational Global interactome 723,310 Yang et al. (2013b)Solanum lycopersicum Yeast two-hybrid MADS box 119 Leseberg et al. (2008)Coffea canephora Computational Global interactome 4,587 Geisler (2011)Oryza sativa Computational Global interactome 37,112 Ho et al. (2012)

Computational PRIN (predicted riceinteractome network)

76,585 Gu et al. (2011)

Hordeum vulgare Yeast two-hybrid +TAP-MS 14-3-3 family 500 Schoonheim et al. (2007)Zea mays Computational PPIM (PPI database for maize) 2,762,560 Zhu et al. (2016)

Computational PiZeaM (predicted interactomefor Zea mays)

49,026 Musungu et al. (2015)

Physcomitrella patens Computational Predicated interactome 67,740 Schuette et al. (2015)


Xing et al.



Box II: Structure and Physics of PPIs

The heterologous cellular environment can become challengingfor protein function. Difficulties arise related to the presence ofunspecific aggregation while retaining a high degree of specificityfor second-order homomeric or heteromeric PPIs at low cellularconcentrations (Reichmann et al., 2007; Marsh and Teichmann,2015). Van der Waals forces, ionic interactions, hydrogen bonds,interactions with water molecules, and hydrophobic effectsform the physicochemical foundations of PPIs (Keskin et al.,2008). Taking a more macromolecular view, the combinationof interface shape, its amino acid composition, hydrogen bonds,the solvation free energy gain from interface formation, andthe eventual binding energy all play important roles in PPIs(Sowmya et al., 2015). Single changes at the amino acid leveloften do not influence an interaction substantially due to thelarge surface areas that contribute to most interaction sites(Reichmann et al., 2007; Aakre et al., 2015). Nevertheless, thereare indeed cases where single point mutations can disrupt PPIs(Karnik et al., 2013a; Zhang et al., 2015).

The affinity between two interacting proteins, A and B, can bequantified through the Kd. The Kd describes the concentration ofprotein A at which it is bound to half of all available sites onprotein B at equilibrium (Kastritis and Bonvin, 2013). In otherwords, the smaller the Kd (femtomolar or picomolar range) of aprotein pair AB, the higher the affinity of A to B, as fewermolecules of A are needed for 50% of all A molecules to bind toB (Perkins et al., 2010). Kd values can be determined experimentallyor predicted, but they are influenced by exogenous factors such astemperature, pH, and salt concentration in the solvent (Kastritiset al., 2011).

PPIs also can be classified as obligate or non-obligate PPIs(Nooren and Thornton, 2003). Obligate PPIs are permanent,usually show low Kd values, and form strong heterotypic orhomotypic interactions, as is the case for potassium channeltetramers (Ward et al., 2009) and transporter oligomers suchas proton pumps (Schumacher and Krebs, 2010). Subunits ormonomers of these complexes are individually unstable andnon-functional; hence, the term obligate for structural stabilityis dictated by their interaction. While there are also examples ofpermanent, non-obligate interactions such as enzyme-inhibitorcomplexes (avidin-biotin is a classic example), the vast majorityof non-obligatory interactions are of a transient nature and mayor may not be weak (Nooren and Thornton, 2003). PPIs can betransient in terms of their cellular longevity but still might bevery strong in terms of binding affinity. For example, the ternarySNARE complexes, which facilitate vesicle fusion through atight but reversible interaction, show high binding affinities(Kd , 50 nM; Rice et al., 1997). Examples of weaker, transientinteractions in the micromolar range include enzyme-substrateand receptor-ligand interactions (Chinchilla et al., 2007; Jianget al., 2013).

Kd values can be used to describe the detection limit of a PPItechnique or the range in which it shows the highest sensitivityand stringency. The table in this box lists ranges of Kd values fromthe literature for the methods described in this review. For two-hybrid techniques such as yeast two-hybrid and FRET, in which theprobes (trans-activating and DNA-binding domains or monomericfluorescent proteins) do not carry intrinsic binding affinity towardeach other, Kd values of individual POI couples are sufficient todescribe the affinity range of the technique. The more the valuesare experimentally determined for individual PPI couples, the morereliably the PPI technique can be assessed for its suitability todetect a certain PPI.

Box II Figure: Kinetics and affinities of PPI techniques. TheKd is inverselycorrelated to the binding affinity. PPIs can exist as permanent or transientinteractions. While transient interactions can be described as weak dueto the higher probability of complex dissociation, strong transient orpermanent interactions bear higher intrinsic binding affinity and greatertemporal stability. More details are given in the text and references(Banaszynski et al., 2005; Perkins et al., 2010; Kastritis et al., 2011).

Box II Table: Kd values of two-hybrid and PCA techniques featured inthis review

Method Kd of Probesl Kd of PPI Reference

Yeast two-hybrid

n.a.a ,50 mM Estojak et al.(1995)e

Split-ubiquitin 7 mMb ,100 mM

c Jourdan andSearle (2000)f

1 mMc ,1 mM Muller and

Johnsson(2008)g

Split-luciferase(NanoLuc)

190 mM ,1 mM Dixon et al.(2016)h

BiFC 0 mMd ,1 mM Magliery et al.

(2005)i

FRET 110 mM (YFP) ,10 mM Piehler(2005)j

74 mM

(mYFP[A206K]),10 mM Zacharias et al.

(2002)

aDNA-binding and transactivation domains do not show intrinsicassociation; hence, a Kd value is not applicable (it would be veryhigh). bThe Kd of the probes was determined for the reassociationof wild-type versions of Nub and Cub. cIt is estimated that theNubI13A mutation increases the Kd to 1 mM and that the NubI13Gmutation value might be even higher, allowing the split-ubiquitinsystem to detect transient interactions up to Kd values of 100 mM

(Muller and Johnsson, 2008). dThe split fluorescent protein fragmentsare known to spontaneously and irreversibly reassociate; hence,the Kd is negligibly small (Magliery et al., 2005). e, 50 mM.f, 100 mM. g, 1 mM

h, 1 mM. i, 1 mM. j, 10 mM.k, 10 mM. lEach of these values is an approximation.





used in its full-length (ADH1; 21,500 bp) or truncated(ADH1*; 2700 bp) version. Notably, expression fromthe full-length ADH1 promoter decreases during thelate exponential, ethanol-producing growth phase,whereas the truncated version, while slightly weaker inexpression, is not repressed in the presence of ethanol(Ruohonen et al., 1995). This observation should beconsidered when growing yeast for too long in liquidmedium for immunoblot analysis. The ability of yeastto switch promoters on or off depending on the carbonsource available in the medium or by application ofexogenous factors also can be exploited for temporalcontrol (Weinhandl et al., 2014).

Inducible promoters that are commonly used includethe Gal-inducible GAL1 promoter (Johnston, 1987) andthe copper-inducible CUPI promoter (Butt et al., 1984).However, growth is considerably slower onGal than onGlc, and copper addition must be carefully balanced, asthis metal is toxic to yeast. Also, the leakiness of apromoter can preclude genetic analysis, as basal tracesof protein expression can lead to an interaction in asensitive detection assay. A potential alternative in-volves the use of a repressible promoter. MET25, forexample, is a strong promoter in the absence ofMet, butit shuts down in the presence of Met (Sangsoda et al.,1985). Notably, complete supplement mixture usuallycontains 134 mM Met (20 mg L21), which significantlyreduces gene expression (Grefen, 2014).

For many years, the strong cauliflower mosaic virus35S promoter (Ow et al., 1987) has been incorporated innumerous plant expression vectors and used for stableas well as transient transformation (Blatt and Grefen,2014). Unfortunately, particularly in stable transform-ants, the expression of a transgene driven by this pro-moter is often silenced and, even if successful, thestrong overexpression can cause artifacts in PPI analy-sis, as discussed above. However, if native expressionof a protein is low due to restricted spatial or temporalexpression, constitutive promoters such as RPS5a orUBQ10 can be used to allow sufficient expression and,in turn, detection of an interaction (Weijers et al., 2001;Grefen et al., 2010b). An alternative to constitutive ex-pression would be to regulate expression throughchemical treatment (estradiol, ethanol, or dexametha-sone), stimulus induction (heat shock or light), or theuse of tissue-specific expression, all of which allow aninteraction to be restricted spatiotemporally and, hence,putting a PPI closer to its native context (Borghi, 2010;Siligato et al., 2016).

The origin of replication is another important factorthat can be used to regulate the strength of expression,albeit primarily when using yeast plasmids (James,2001). Yeast maintains three different types of origins ofreplication: autosomal replication sequence, 2mm, andintegrating. While the autosomal replication sequenceoften is used in combination with a centromere sequenceto guarantee plasmid stability over multiple generations,these plasmids are maintained in low abundance of oneto two copies per cell. To enhance the expression of a POIeven further, a 2mmplasmid can be used, allowing yeast

cells to keep up to 50 copies. The most native-likemaintenance and expression would be achieved usingan integrating plasmid, which stably inserts into theyeast chromosome. Such plasmids, however, are rarelyused for the expression of heterologous plant proteinsand the analysis of their interactions.

Linkers

When contemplating the assembly of custom-builtcloning cassettes or vectors for a PPI technique, it iscrucial to choose the correct linker between the pro-tein and the tag (Chen et al., 2013). If the linker istoo short or rigid, an interaction of the POIs mightnot enable reassembly or might limit the area in which theprobes canmeet, preventing detection (Wehr and Rossner,2016). The short linker sequences left by Gateway cloningafter recombination may not be sufficiently flexible to al-low for probe interaction (Grefen and Blatt, 2012a).

The orientation of the linker and tag is an additionaland important factor to consider, as tagging of eitherthe N or C terminus of a protein can potentially masksignal peptides, targeting sequences, or even interac-tion interfaces, precluding the identification of bindingpartners (Grefen and Blatt, 2012b). While masking of asignal peptide or targeting sequence might prevent in-teraction due to artificial mislocalization, occlusion of alarge, often hydrophobic targeting sequence can resultin the aggregation of proteins and their subsequentdegradation. Before a PPI is tested, sequence analysis ofthe protein using TargetP, PSORT, or other in silico toolscould be used to obtain information about targeting se-quences and to determine tag orientation accordingly.

The topology of the POIs also can be an importantfactor, as some PPIs discussed below demand specificorientation of the termini. The classical split-ubiquitinanalysis, for example, requires the bait protein to bemembrane bound to prevent leakage of the fusion pro-tein into the nucleus, where the attached transcriptionfactor would cause reporter gene activity; however, atthe same time, the C terminus must face the cytosol (seebelow). In addition, it is important to note that, whenverifying interactions using different methods, the ab-sence of an interaction in onemethod could easily be dueto the tag orientation, which could be circumvented bytesting all possible tag combinations of a binary inter-action couple (Stellberger et al., 2010).

Subcellular localization is a point of concern whenperforming meaningful analysis of a detected PPI;however, deliberate targeting can be an advantage, andsome techniques are easier to perform in certain com-partments. For example, FRET acceptor photobleachingcan become unreliable when detecting a PPI in a highlymotile organelle or compartment (see below).

Experimental Factors: Controls

A final but central point before commencing any PPIanalysis is to select meaningful negative and positive


Xing et al.



controls. These controls should be chosen according tothe method used, and the location in which the inter-action is supposed to take place in the heterologoussystem versus its native localization must be taken intoaccount. In other words, for a protein couple that in-teracts in the vicinity of the plasma membrane, thenegative control should not be an obligatory nuclearprotein. The best negative control is usually a pointmutation of one of the proteins under investigation,provided that the mutation does not lead to its degra-dation or increased turnover (Grefen et al., 2010a;Kodama and Hu, 2012). The gold standard for negativecontrols should be that the expression of the fusionprotein is verified via immunoblot analysis.

THE YEAST TWO-HYBRID TECHNIQUE:WHERE IT ALL BEGAN

It is interesting to observe the impact that a techno-logical article can have on the community. The inge-nious yet simple method devised by Stanley Fields andOk-Kyu Song in 1989 is certainly an example of such aarticle, which revolutionized the study of PPIs (Fieldsand Song, 1989; Fields, 2009).The principle is straightforward (Figs. 1A and 3A).

The transcriptional activator GAL4 contains anN-terminal DNA-binding domain and a C-terminaltransactivation domain interspaced by a large struc-tural region. Both domains, the DNA-binding (aminoacids 1–147) and transactivation (amino acids 768–881)domains, are independently stable and functional anddo not require the central body of the protein. In otherwords, each one is able to fulfill its nuclear functiondespite being displaced from the rest of the GAL4protein: the DNA-binding domain still binds the GAL1upstream activating sequence, and the transactivationdomain, if brought into the vicinity of the TATA box, issufficient to activate transcription. When two POIsthat interact are fused to the DNA-binding and trans-activation domains, respectively, their association createsa chimeric transcription factor, turning on gene expres-sion downstream of the GAL1 upstream activating se-quence (Fields and Song, 1989).This system has also been successfully established

using different split probes. The DNA-binding domainof the bacterial repressor LexA can be fused to the GAL4,bacterial B42, or Herpes simplex VP16 transactivation do-main. The different domains feature properties that allowyeast two-hybrid analysis to be modulated. While theGAL4-DNA-binding domain contains a nuclear lo-calization signal, LexA does not, which results in lessbait protein entering the nucleus (Brent and Finley,1997). This could potentially reduce the number offalse positives but also might prevent the detection ofa weak interaction. Other approaches to manipulat-ing the system include adjusting the strengths of vari-ous activation domains. For example, while VP16 is avery strong transcriptional activator (Uhlmann et al.,2007), B42 is rather weak and can counteract toxic effects

associated with strong activation via the GAL4 trans-activation domain (Gill and Ptashne, 1988).

The choice of DNA-binding and transactivation do-mains also has implications for the choice of the re-porter gene cassettes incorporated in the yeast strain.While the GAL4-DNA-binding domain binds upstreamof the GAL1, GAL2, or GAL7 promoter, LexA requiresthe presence of the lexA operon of E. coli upstream ofthe reporter genes. These can code for chromogenicenzymes such as b-galactosidase (lacZ, for selection viablue/white coloring of colonies), prototrophic markers(e.g. HIS3 or ADE2, enabling survival on depletedgrowth medium), or both (James, 2001). For example,the yeast strain PJ69-4A (James et al., 1996) carries theHIS3 reporter driven by the GAL1 promoter, ADE2driven by the GAL2 promoter, and lacZ driven by theGAL7 promoter. Since the GAL4-DNA-binding domaincan bind to the upstream activating sequences of all threepromoters but with different affinities, the combination ofmultiple reporters provides higher stringency, signifi-cantly reducing the number of false positives. While theHIS3 and ADE2 reporters enable qualitative analysis viasurvival on selective growth medium, the lacZ gene al-lows for semiquantitative readout of interactions throughenzyme assays using substrates such as 5-bromo-4-chloro-3-indolyl-b-D-galactopyranoside acid (X-Gal)and 2-Nitrophenyl b-D-galactopyranoside (oNPG)(Serebriiskii and Golemis, 2000; Möckli and Auerbach,2004). Additionally, the haploid yeast strain PJ69-4Aalso exists as the isogenic partner strain PJ69-4a (James,2001). Rather than performing numerous, tedious dualtransformations (two plasmids at once), it is possibleto first transform a set of different bait constructs inone mating type and a set of preys in the other matingtype; then, by mixing the two haploid mating types,large numbers of binary interactions can be evaluated.

Screening and False Positives

Both the yeast two-hybrid and split-ubiquitin systemsare exceptionally well suited for high-throughputapproaches and complementary DNA (cDNA) libraryscreening (Fig. 3B; Table 1, Box I). Important points toconsider are the choice of library (polyA-cDNA, normal-ized, tissue or development specific) and whether se-quential transformation or a mating-based approach willbe used. One of the biggest limitations of the yeast two-hybrid system is the obligatory nuclear localization of theinteraction partners.Most of the problems that complicatethe analysis of the results of yeast two-hybrid assays are adirect consequence of this prerequisite. For example,membrane proteins must be truncated to allow for theirforced relocation to the nucleus. Large cytoplasmic pro-teins may be prevented from entering the nuclear pores(Kabachinski and Schwartz, 2015), and their truncationcan often render them non-functional, leaving the signifi-cance of any interaction result in question.

The false positives that usually come up in a cDNAlibrary screen arise from three different types of pro-teins and their interactions (Fig. 3C; Parchaliuk et al.,





Figure 3. The yeast two-hybrid system and its modifications. A, The original yeast two-hybrid analysis detects an interactionbetween two proteins fused to either the DNA-binding domain (BD; bait) or the DNA transactivation domain (AD; prey) of GAL4,respectively. An interaction reconstitutes the DNA-binding and transactivation domains to activate reporter genes such as ADE2,HIS3, and lacZ. B, A cDNA library screen allows the identification of novel binding partners (green crayon-like shape), whereasnon-interacting proteins (opaque square and sphere) fail to activate the reporter genes. C, Examples of false-positive results thatmight be detected in yeast two-hybrid cDNA library screens that should be excluded by appropriate testing (see “The Yeast Two-Hybrid Technique:Where It All Began” and Box I): (1) unrelated prey fusions that bind to the bait; (2) transcriptional activators thatare sufficient to trigger reporter gene activity; and (3) enzymes that overcome the selection pressure on depleted medium byrestoring prototrophy. D, An extension of the classic yeast two-hybrid system is yeast three-hybrid analysis, in which two non-interacting or weakly interacting proteins (bait I and bait II) are bridged or stabilized by a third protein (cyan; prey/bridge). E,Reverse yeast two-hybrid detects the interference of a known interaction couple and can be used to screen for inhibiting proteinsor molecules (red cones). Transcript activation of a selectionmarker such as URA3 renders yeast sensitive to 5-fluoroorotic acid. Ifthe interaction is prevented, the reporter genes are not activated and the yeast can survive on 5-fluoroorotic acid. UAS, Upstreamactivating sequence.


Xing et al.



1999): (1) proteins that bind unspecifically to the bait; (2)random transcriptional activators; and (3) enzymes thatrestore the prototrophy of the yeast. (1) Unrelatedproteins, such as spatiotemporally misexpressed pro-teins, may bind unspecifically to the bait and cause re-porter gene activity [Fig. 3C, (1)]. Such false positivescan be countered through the appropriate choice of a li-brary. Rather thanusing a library from thewhole organism,the use of tissue- or developmental stage-specific librariescan help reduce the number of false positives. Green tissue-derived librarieswill contain a large number of chloroplast-relatedprotein sequences,which can lead to numerous falsepositives when using a non-chloroplast-localized bait POI.(2) Intrinsic transcriptional activity independent of the in-teraction also can generate false positives through interac-tion with the bait protein [Fig. 3C, (2)]. As these effectsmight be less efficient than that of a bona fide interactor,adjusting the expression levels of bait and/or library preycan yieldmore stringent conditions and reduce the numberof false positives in this category. (3) In some cases, enzy-matic gene products from a cDNA library have the poten-tial to restore prototrophy in yeast [Fig. 3C, (3)]. Using atleast two auxotrophic markers such as ADE2 and HIS3provides a check against this possibility, as two differentenzymeswould need to be replaced to facilitate the steps inthebiochemical pathwaysofHis andadenineproduction. Ifonly HIS3 is used as a marker, the addition of 3-AT, aninhibitor of HIS3, can help titrate screening conditions andreduce the likelihood of false positives.

Ternary Interaction Analysis

The simplicity of the yeast two-hybrid system al-most immediately sparked the development of severalmodifications and improvements. Most importantly,the system can be used to detect trimeric interactionsby simply coexpressing a third protein partner, leadingto the development of the yeast three-hybrid screeningapproach. Here, an interaction can be (1) outcompeted, (2)enhanced, or (3) facilitated/bridged (Fig. 3D; Zhang andLautar, 1996). The system has been used successfully in theplantfield; examples of each of these bindingmodes can befound in the literature (Suzuki et al., 2009; Maruta et al.,2016). The yeast three-hybrid system is particularly elegantwhen using a controllable promoter for the thirdprotein partner, allowing the trimeric interaction to beswitched on or off through the addition of an inducer orinhibitor (Tirode et al., 1997).An elegant, alternative solution to screening for in-

hibitors is reverse yeast two-hybrid analysis (Fig. 3E;Leanna andHannink, 1996). This system is based on theexpression of a reporter gene that can be used as a neg-ative selection marker. By reintroducing CYH2 (LeannaandHannink, 1996) orURA3 (Vidal et al., 1996) into yeastcells, the cells are rendered sensitive to cycloheximide, aninhibitor of protein biosynthesis, or to 5-fluoroorotic acid,a precursor of the toxic uracil analog 59-fluorouracil, re-spectively. If an interaction is prevented through a mu-tation in the bait, prey, or third protein partner, the yeastcells can grow on cycloheximide or 5-fluoroorotic acid.

When assaying for third binding partners, it is im-portant to keep in mind that an ortholog of the thirdprotein might actually be present in yeast and facili-tate bridging in lieu or absence of the plant candidate.A solution to this problem would be to knock out thecorresponding gene in the reporter strain used forthe yeast three-hybrid analysis, provided that it is notessential for survival and/or does not cause severephenotypes.

There are a range of further modifications and im-provements to the yeast two-hybrid system. A com-prehensive list of references for these can be found in anexcellent review by Patrick Van Dijck and coworkers(Stynen et al., 2012). One notable example is the appli-cation of a two-hybrid approach in plant cells. The planttwo-hybrid approach allows an interaction in proto-plasts to be detected using GUS expression as a readout(Ehlert et al., 2006).

THE SPLIT-UBIQUITIN SYSTEM:A FULL-LENGTH ALTERNATIVE

Despite its great popularity, the greatest disadvan-tage of the classical yeast two-hybrid system is theobligatory nuclear localization of the proteins and,hence, their site of interaction. Artifacts may arise fromthe truncation and mislocalization of otherwise non-nuclear proteins. Five years after the initial descriptionof the yeast two-hybrid system, an alternative with thepotential to overcome these limitations was described.The split-ubiquitin system never managed to gain asmuch popularity as the yeast two-hybrid system de-spite providing the same ease of application, yet it al-lows virtually all protein types to be tested withoutthe need to truncate or mislocalize these proteins andwithout introducing additional artifacts to those as-sociated with the yeast two-hybrid system.

Ubiquitin is a eukaryotic protein that is highly con-served among species, with an almost identical se-quence comprising just 76 amino acids. This protein isinvolved in a multitude of vital processes through itsinherent function (i.e. tagging proteins destined fordegradation via ubiquitin-specific proteases). In 1994,Nils Johnsson and Alexander Varshavsky split ubiquitininto two fragments, anN-terminalNub (amino acids 1–34)and a C-terminal Cub (amino acids 35–76), and fused aHA-tagged protein to its C terminus (Johnsson andVarshavsky, 1994). Coexpression of both split fragmentsleads to the reassembly of ubiquitin and subsequentcleavage of the hemagglutinin-tagged protein, which canbe visualized through immunoblot analysis. Mutatingthe Ile at position 13 of Nub to an Ala or Gly preventedthis reassembly, leading to the establishment of the firstPCA system. Later work identified 10 additional singleand double point mutations in the Nub peptide that offercontinuous levels of affinity between Cub and Nub, anexcellent tool set of varying stringency that facilitates theestimation of the dissociation constant (Kd) range of twointeracting POIs (Box II; Raquet et al., 2001).





A few years later, Igor Stagljar and coworkers usedthe chimeric transcription factor PLV as a reporter. PLVconsists of Protein A from Staphylococcus aureus (con-taining two IgG-binding sites) followed by the LexADNA-binding and H. simplex VP16 DNA activationdomains (Figs. 1D and 4A; Stagljar et al., 1998). TheProtein A tag was originally incorporated for immu-nochemical detection of the fusion construct and cleavedreporter; however, comparedwith a bait vector depletedof the Protein A tag, its presence also appears to prolongthe lifetime of the transcription factor, hence increasingreporter gene activity (C. Grefen, unpublished data).While the incorporation of a transcriptional readoutprovides a highly stringent detection system, it imposesthe use of membrane-anchored or -attached proteins asCub fusions; otherwise, diffusion of the entire fusionprotein into the nucleus would result in reporter geneactivation. A later development termed the cyto-split-ubiquitin system circumvents this problem by attachingan N-terminal membrane anchor (the transmembranedomain of Oligosaccharyltransferase 4, OST4) to theCub fusion, thereby allowing transcription factors orcytosolic proteins to be used as bait (Fig. 4B; Möckliet al., 2007; Karnik et al., 2015).

In addition to the transcriptional readout reportersystem, another approach was developed using thecomplementation of ubiquitin. Here, the Cub moiety isfused to orotidine-59-phosphate decarboxylase (URA3),implementing an alternative reporting mechanism tonuclear transcription (Fig. 4D). URA3 is preceded by anArg residue that destabilizes it after cleavage accordingto the N-end rule of degradation (Fig. 4D; Wittke et al.,1999). Rapid turnover of URA3 renders cells auxotro-phic to uracil but resistant to 5-fluoroorotic acid (Müllerand Johnsson, 2008). Hence, the URA3-based split-ubiquitin system can be used for either the detection ofbinary PPIs (when counterselection with 5-fluorooroticacid is used) or the screening of inhibitors or point mu-tations that abolish an interaction (when using uracil-depleted medium for selection). While the cost forthe chemical compound 5-fluoroorotic acid and thenecessary replica plating represent downsides, thisreporter system facilitates screening for inhibitors ofan interaction at the membrane or in the cytosol ornucleus (Dirnberger et al., 2006).

The split-ubiquitin system has been used extensivelyin plant research to identify interactions of membraneproteins such as G-proteins (Aranda-Sicilia et al., 2015),phosphate transporters (Fontenot et al., 2015), andaquaporins (Besserer et al., 2012; Hachez et al., 2014),potassium channels (Obrdlik et al., 2004), and their in-teractions with SNARE proteins (Honsbein et al., 2009;Grefen et al., 2015; Zhang et al., 2015), specifically usingthe vectors modified by Petr Obrdlik and coworkers(Obrdlik et al., 2004). Improvements to the originalvector system included increasing the signal-to-noiseratio through introducing a Met-repressible promoterdriving the bait Cub fusion (Obrdlik et al., 2004), in-corporating cloning sites for in vivo cloning as well asGateway recombination technology (Obrdlik et al., 2004;

Grefen et al., 2007, 2009), and altering linkers and tags(Grefen and Blatt, 2012b), all of which, together with thedevelopment of strains that allow a mating-based ap-proach to be taken (Ludewig et al., 2003), have made thesystem suitable for high-throughput analysis (Box I;Lalonde et al., 2010; Chen et al., 2012b; Jones et al., 2014).

The split-ubiquitin system also suffers from false-positive results similar to those detected using theyeast two-hybrid system. In particular, in cDNA libraryscreens, typical false positives include ubiquitin-relatedorthologs and proteolytic enzymes, which probablylead to the recognition of ubiquitin-specific proteases ordirect cleavage of the PLV reporter, even in the absenceof an interaction or prey (C. Grefen, unpublished data).This type of background activation of ubiquitin-specificproteases also might account for the observation thatusing the split-ubiquitin system in planta with a desta-bilized GFP as a reporter, whose cleavage can be detec-ted via immunoblot, yields high background ratioscompared with its application in yeast (Rahim et al.,2009). Nevertheless, optimizing the signal-to-noiseratio of this in planta split-ubiquitin system wouldmake this in vivo PPI technique highly valuable.

Other Applications

The split-ubiquitin system also is suited for trimericanalysis, a method known as the split-ubiquitin systembridge assay (Fig. 4C; Honsbein et al., 2009, Grefen andBlatt, 2012b; Grefen, 2014). The possibility of analyzingmembrane proteins as potential enhancer, bridging, orinhibiting factors becomes increasingly interesting, forexample, in the analysis of (ligand-induced) receptoroligomerization. Apoplastic proteins, peptides, smallmolecules, or drugs can be screened for their potentialto inhibit or increase the interactions of membrane-bound receptors (Dirnberger et al., 2008; Grefen, 2014;Wehr and Rossner, 2016).

A novel modification, termed SPLIFF, enables spa-tiotemporal examination of PPIs (Fig. 4E; Moreno et al.,2013). The bait POI is C-terminally fused with mCherryfollowed by the Cub peptide. A destabilized GFP isfused to the C terminus of the Cub moiety. Bait andprey fusions are expressed in yeast strains of differentmating types. Upon mating, the GFP is cleaved off andquickly degraded, increasing the mCherry-to-GFP ratioat the site of interaction. Time-lapse microscopy of theyeast enables the PPI to be monitored with high spatialand temporal resolution (Dünkler et al., 2015). If thehigh background reported for plant cells when usingthe split-ubiquitin system could be reduced and if theexpression of the bait fusion is controlled by induciblepromoters, SPLIFF would represent an interesting inplanta tool for monitoring interactions in real-time.

FRET: LIFETIME VERSUS INTENSITY

The discovery and subsequent engineering of fluo-rescent proteins revolutionized cell biological detectionsystems 20 years ago (Heim and Tsien, 1996). For the


Xing et al.



first time, protein tagging with fluorescent proteinsallowed the subcellular and spatiotemporal distribu-tion of proteins to be resolved and their dynamics to bestudied in vivo. Expanding the fluorescent palette alsocontributed to further advancements in developing bio-sensors, subcellular markers, and optimized fluorescentproteins for interaction studies (Day andDavidson, 2009).

However, as much as the development of high-endconfocal microscopes and fluorescent proteins hasboosted the analysis of proteins on a subcellular scale,assessing colocalization and PPIs remain two separateissues (Shaw and Ehrhardt, 2013). Colocalization be-tween two POIs using fluorescent proteins with differentspectroscopic properties might suggest their involvement

Figure 4. The split-ubiquitin system and its modifications. A and B, N-terminal (NubG) and C-terminal (Cub) ubiquitin moieties(blue half-spheres) are fused to two POIs, prey and bait, whereby the bait protein (red) needs to be membrane anchored eitherthrough an intrinsic transmembrane domain (red helix; A) or an artificial N-terminally attached OST4 (Oligosaccharyltransferase4) transmembrane domain (blue helix; B). In both split-ubiquitin versions, the reassembly of Cub and NubG through the inter-action of bait and prey leads to the release of a chimeric transcription factor, LexA-VP16, which activates the reporter genesADE2,HIS3, and lacZ. C, The split-ubiquitin system also can be expanded to a split-ubiquitin bridge assay. D, An alternative split-ubiquitin approach that does not require the use of membrane-anchored baits uses the N-end rule of degradation. The auxotrophymarker URA3 (which can be counterselected using 5-fluoroorotic acid) is preceded by a destabilizing Arg residue. Cleavage ofURA3 upon the reassembly of Cub and Nub leads to its rapid degradation. E, A recent improvement of the split-ubiquitin systemcalled SPLIFF enables the time-resolved analysis of interactions. The bait protein is tagged with an mCherry-Cub-GFP protein.Upon interaction, Cub andNub reassemble, leading to the cleavage and degradation of GFP. The ratio betweenmCherry andGFPcan be measured over time, providing information about the kinetics of an interaction.





in the same pathway or protein complexes. However,spatial resolution is dependent on the diffraction limit oflight microscopy, which is in the range of 200 nm. Con-sidering this distance and (for simplicity) just one di-mensionally, the size of a GFPmolecule, which is roughly3 3 4 nm (Yang et al., 1996), would allow 48 fluorescentproteinmolecules tofit between twoPOIs at either end. Inother words, while image analysis would reveal thepresence of both fusion proteins in the same spot, dozensof other protein molecules could be situated betweenthem. Hence, the colocalization of fluorescence deter-mined by standard light microscopy alone cannot pro-vide information about protein complex formation.

A solution to this problem is provided by the physicalphenomenon of FRET (Förster, 1946), a process that,for example, enables the light-harvesting complex totransfer the energy of absorbed photons to the reactioncenter of photosynthesis. This non-radiative energytransfer is based on long-range dipole-dipole couplingbetween two molecules. Hence, using “fluorescence”instead of “Förster” is not only a misnomer by con-vention but also is misleading (Figs. 1B and 5A). FREToccurs when the energy relaxation of an excited donorfluorescent protein is not emitted via its own fluores-cence but instead is emitted through non-radiativetransfer of this energy to a nearby acceptor molecule;incidentally, the acceptor does not have to be a fluo-rescent protein but can simply be an energy sink for thedonor (Lakowicz, 2006).

For FRET to occur, three prerequisites must bemet: (1)the spectral overlap of donor emission and acceptor ab-sorption must be sufficient for effective energy transfer;(2) the distance between fluorophores must be below 10nm (see below and Fig. 5A); and (3) the dipole orientationof the fluorophores must be aligned, ideally in parallelfashion. Additional factors that influence energy transferefficiency include the quantum yield of the donor (theratio of photons emitted per photons absorbed) and theextinction coefficient of the acceptor, which describesthe capacity of amolecule to attenuate energy. Thedistanceat which 50% of the energy is transferred is called theFörster radius (R0). In general, the energy transfer is de-pendent on the inverse sixthpower of thedistance betweendonor and acceptor, and as a consequence, FRET is limitedto distances of 1 to 10 nm (Sun et al., 2011). This distance isconsiderably lower than the diffraction limit of light, yetthe efficiency of energy transfer allows optical analysis ofPPI dynamics to be performed in vivo with high spatio-temporal resolution.

While FRET is often used to infer protein interactions,its detection actually informs on the vicinity and/ororientation, which can be visualized using anothertheoretical calculation (Lalonde et al., 2008). Thecylindrical GFP has a volume (V ) of roughly 30 nm3

(V = pr2h; where r = 1.5 nm and h = 4 nm). The 10-nmradius of a hypothetical sphere in which FRET occursequals a volume of approximately 4,200 nm3 (V = 4/3pr3; where r = 10 nm), indicating that almost 140 GFPmolecules could be considered within that area. ForPOIs taggedwithGFP, this numberwould be far smaller,

which demonstrates that even when FRET is detected, itdoes not necessarily prove a direct physical interactionbut instead could result from the proximity of proteins,such as in a multiprotein complex. However, in a two-dimensional space, such as a membrane domain, farfewer molecules can fit into the FRET radius. As a re-sult, the observance of FRET with membrane proteinsis more likely due to direct interaction, whereas FRETwith soluble proteins has a probability, albeit small, tobe caused by vicinity.

The use of FRET-based methods to detect PPIs ap-pears to have outstripped the yeast two-hybrid system(Fig. 2B; Table II). One reason for the popularity ofFRET is that, in addition to its use for detecting PPIs, italso can be used with genetically encoded, unimolec-ular sensors, and it can be used in native species in vivo(Frommer et al., 2009; Uslu and Grossmann, 2016). Arange of sensor molecules have been developed or op-timized for use in plants, enabling the detection of cal-cium (Krebs et al., 2012; Wagner et al., 2015), phosphate(Mukherjee et al., 2015), zinc (Lanquar et al., 2014), andabscisic acid (Jones et al., 2014a) aswell as physiochemicalstates of the cell such as pH (Michard et al., 2008) andmembrane voltage (Grefen et al., 2015). One great ad-vantage of using intramolecular FRET sensors lies in thetheoretically equal stoichiometry, rendering ratiometricreadouts independent of sensor concentration.

Detecting FRET through Sensitized Emission, AcceptorPhotobleaching, and Fluorescence Lifetime Microscopy

There are a number ofways to detect FRET, eachwithits own set of benefits and pitfalls. For brevity, we willfocus on the three most commonly used technologies:(1) sensitized emission; (2) acceptor photobleaching;and (3) fluorescence lifetime imaging (FLIM). Morespecialized methodologies are described in reviewsfocusing on FRET (Clegg, 2009; Sun et al., 2011;Ishikawa-Ankerhold et al., 2012).

Measuring the increase in intensity in acceptoremission when exciting the donor is the basic principlebehind sensitized emission FRET. In practice, this canbe done by exciting the donor in the range of its ab-sorption maximum while measuring emission in therange of the acceptor’s emission spectrum. This methodis straightforward, but it must be carefully executed, asunequal loading of the fluorescent proteins will givedramatically different results. Figure 5B shows the ab-sorption (dashed lines) and emission spectra (solidlines) of mTurquoise2 and mVenus (modified fromHecker et al., 2015). The l4 weighted spectral overlap,which is depicted in dark gray, is one of three impor-tant factors for efficient FRET. However, the overlapbetween absorption (magenta) and emission spectra(green) of the donor and acceptor defines what is re-ferred to as spectral bleed through or cross talk. Crosstalk arises from the fact that all fluorescent proteinsabsorb and emit photons across a range of wavelengthsrather than at a specific wavelength, like a laser. There-fore, even if detection is restricted to the peak emission of


Xing et al.



onefluorescent protein, it does not preclude the emissionof other wavelengths. The same applies to excitation:even wavelength regions that are very low in absolutepercentage of absorbance will lead to fluorophore exci-tation when high-excitation light intensity and/or largeamounts of fluorescent protein are used. Only whenwavelengths are chosen that lie completely outside of theemission and/or excitation region of a fluorescent pro-tein can spectral bleed through be avoided. A simple testfor bleed through is to observe the fluorescent proteinsalone, at different concentrations, and to record the fullspectral characteristics of each fluorophore.

Two types of spectral bleed through are common: (1)acceptor spectral bleed through (Fig. 5B, magenta),which occurs when excitation of the donor does not ex-clude excitation of the acceptor, causing an emission ofthe acceptor that is not due to FRET; and (2) donorspectral bleed through (Fig. 5B, green), where emissionof the donor leaks into the emission spectrum of the ac-ceptor. To distinguish FRET frombleed-through artifactsusing sensitized emission, great caremust be takenwhenchoosing the controls, including donor only, acceptoronly, and donor and acceptor together, as well as theanalysis of spectral data (Müller et al., 2013).

Figure 5. FRET. A, Simplified Jablonski diagram. A fluorescent protein (here depicted as the cyan-colored crystal structure of a CFPattached to a POI in gray) that absorbs light (blue wavy arrow) causes an electron to be raised to an excited singlet state, S1* (bluesolid arrow). Internal conversion (black wavy arrow) causes relaxation to the S1 ground state (cyan solid arrow). Crossing of theenergy barrier between S1 and S0 can cause the emission of a photon at a longer wavelength (red shifted, Stokes shift; i.e. fluo-rescence). If an acceptor molecule is in close enough proximity (here, YFPattached to a red-colored interacting protein), the energycan be transferred non-radiatively to the acceptormolecule. Relaxation of the acceptormolecule, in turn, can lead to evenmore red-shifted fluorescence. B, Absorbance/fluorescence spectrum of mTurquoise2 and mVenus. The dark gray surface depicts the l4

weighted overlap integral. Acceptor and donor spectral bleed through are shown in magenta and green, respectively (see “FRET:Lifetime Versus Intensity” for details; image modified from Hecker et al., 2015). C, Example of a FRET acceptor photobleachingexperiment. A nucleus expressing two interacting proteins attached to a donor and an acceptor fluorophore. After bleaching (right),the acceptor fluorescence is almost completely lost, but an increase in donor fluorescence can be detected (image from Heckeret al., 2015, modified for visualization purposes only). D, Exemplary fluorescence lifetime decay curve of a putative donormoleculewhen FRETis occurring (red solid curve; t1) or in the absence of FRET (cyan solid line; t2). Fluorescence lifetime t is the average timethat a fluorescent protein resides in the excited state and at which fluorescence intensity decreased to 1/e of its initial value.





Uncontrolled gene dosage can easily bemisinterpretedas a specific signal in sensitized emission assays. Forexample, donor emission in the acceptor peak channelwhen the acceptor is absent or in lower abundance couldbe interpreted as the presence of the acceptor. Conversely,when the abundance of the acceptor exceeds that of thedonor and is excited by the same light source, moreemission in the acceptor would be taken as an indicationof FRET when none exists. One solution in these cir-cumstances is to adjust the relative amount offluorescentproteins by controlling gene dosage (Hecker et al., 2015).

Another way to detect FRET is to determine the in-crease in donor intensity after photobleaching of theacceptor (Fig. 5C). The rationale behind this approachlies in the loss of energy transfer from the donor to theacceptor when bleached, resulting in an increase indonor fluorescence intensity and lifetime. Acceptorphotobleaching is a straightforward method that doesnot require high-end microscopes, is not affected byspectral bleed through, and allows the sample analyzedto serve as its own control.

However, a number of pitfalls should be consideredwhen applying acceptor photobleaching. Bleaching ofthe acceptor also can bleach the donor, causing itsphotoconversion or making it generally toxic to thespecimen (Ishikawa-Ankerhold et al., 2012). As bleach-ing occurs in the excited state of a fluorescent protein, thelaser line used to bleach the acceptor must not overlapwith the excitation spectrum of the donor. Another issueis the temporal delay in recording; in images betweenprebleaching and postbleaching, POIs or organelles canmove in or out of the focal plane, thereby affectingfluorescence collection and rendering a potential changein donor fluorescence intensity meaningless. Solutionsinclude fixing the tissue or using immunolocalization,but such handling also can influence FRET efficiency(Bücherl et al., 2013; Joosen et al., 2014). As mentionedabove, stoichiometric differences in the expression ofan acceptor and donor also can reduce confidence inthe detection of FRET itself. Again, guaranteeing atleast equal gene dosages can increase FRET efficiencyand confidence in acceptor photobleaching analysis(Hecker et al., 2015).

FLIM is a method that is unaffected by spectral bleedthrough, differences in donor or acceptor concentration,excitation intensity variance, and photobleaching(Bücherl et al., 2014). The fluorescence lifetime t of afluorescent protein represents the average time it re-sides in its excited S1 state before relaxation to the S0ground state through photon emission. A disadvantageof FLIM is that it requires sophisticated hardware add-ons,including a pulsed laser that can emit extremely short laserbursts betweenwhich the decay of donor fluorescence canbe measured pixel by pixel. Nevertheless, having this in-formation avoids many of the difficulties of spectral bleedthrough, as one needs only look at donor output.

A typical decay curve is depicted in Figure 5D. FRETresults in a decrease in t for the donor due to energytransfer (Fig. 5D, red curve), whereas without FRET, thevalue for t is greater (Fig. 5D, cyan curve). FLIM requires

control measurements of the donor without the acceptoras well as a control of the donor with a fluorescentlylabeled, unrelated but colocalizing protein.

Of all the methods used to detect FRET, FLIM isprobably the most reliable, allowing highly dynamicspatiotemporal resolution of PPIs. Even so, FLIMmeasurements can be affected by environmental factorssuch as pH, temperature, and the refractive index of themedium. These properties of fluorophores can beexploited when synthesizing biosensors for FLIManalysis (Frommer et al., 2009).

Optimal FRET Couples

Table III summarizes FRET couples and examples oftheir use in the plant field since Theodorus Gadella andcoworkers first demonstrated that the cyan fluorescentprotein (CFP)/yellow fluorescent protein (YFP) coupleis suitable for detecting FRET in planta (Gadella et al.,1999). A detailed discussion of individual point muta-tions, resulting spectral changes, and their implica-tions for FRET can be found elsewhere (Kremers andGoedhart, 2009; Müller et al., 2013).

Despite being outdated, the classic FRET pair CFP/YFP still enjoys great popularity (Tramier et al., 2006;Bhat, 2009). Continuous improvements and optimiza-tion of other fluorescent proteins and the inherent flawsof CFP and YFP suggest their replacement with moreadvanced fluorescent protein couples (Table III; Heckeret al., 2015). Like all GFP-based fluorescent proteins,the original CFP (and YFP) has a weak tendency todimerize, a feature that can complicate detecting FRETwhen cluster formation of an acceptor or donor resultsin the quenching of fluorescence (Zacharias et al., 2002).In addition, CFP has a low quantum yield, biexponen-tial fluorescence decay, and long fluorescence lifetimethat renders it extremely sensitive to bleaching (Tramieret al., 2006). YFP is much brighter than CFP but has anintrinsic sensitivity to environmental conditions such aspH and halide concentration that is responsible for arti-facts in biosensors (Schwarzländer et al., 2014). Together,the resulting FRET couple has a small R0 and a large dif-ferential in brightness, which can mask differences in theconcentrations of the donor and acceptor and lead to ar-tifactual FRET detection, depending on the method used.

What would characterize an optimal FRET couple?The ideal FRET pair consists of a donor that is mono-meric, has a high quantum yield, and is relativelyinsensitive to photobleaching yet features a long fluo-rescence lifetime and a large Stokes shift (the distancebetween absorption and emissionmaxima). An optimalacceptor is also monomeric, has a high absorption co-efficient, and, if acceptor photobleaching is planned,should be photolabile to guarantee at least 90% bleach-ing. The FRET pair together should have a large spectraloverlap, yet ideally with little bleed through, nearlyequal, short maturation times, and a high R0 to allowdetection of an interaction between large proteins orcomplexes (Table III).


Xing et al.



Future Applications of FRET

FRET analysis is an excellent approach to studying inplanta PPI dynamics in real-time, as complex associationis not hindered through irreversible complementation.FRET can be expanded to detect ternary interactions byusing more than two fluorescently labeled proteins inthree-color spectral FRET (Sun et al., 2010), two-stepFRET (Seidel et al., 2010), three-chromophore FRET(Galperin et al., 2004), or triple FRET (Peter et al.,2014). In addition to the donor and acceptor, thesetechniques take advantage of a red-shifted third fluo-rescent protein, acceptor 2, for analyzing larger pro-tein complexes that introduce a significant increasein R0. A persistent problem, however, is the need forspectral separation of the donor and acceptor 2, be-cause overlap between the two short-circuits the ac-ceptor (Seidel et al., 2010). The combination of FRETand BiFC is not a three-color FRET approach, but itrepresents a ternary interaction analysis nonetheless.This approach has been applied successfully in plantato detect ternary SNARE complex formation usingacceptor photobleaching (Kwaaitaal et al., 2010).In many ways, FRET is still in its infancy in plant

research (Fig. 2; Table II). With the development ofFRET pairs that use excitation wavelengths in the rangeof the green gap of chlorophyll, problems derived fromtissue autofluorescence can be overcome. Confocal laserscanning microscopy can be used for detection in FRETanalysis, as well as BiFC, a technique currently favoredbymost plant scientists that we discuss in the followingsection.

BIMOLECULAR FLUORESCENCECOMPLEMENTATION:FROM CHERRY PICKING TO QUANTIFICATION

A considerable number of biological studies, espe-cially in the plant sciences, feature the PCA techniqueBiFC as themethod of choice when it comes to detectingPPIs (Fig. 2C; Table II). BiFC was originally described

in E. coli when Ghosh et al. (2000) demonstrated thata fluorophore such as GFP could be split into N- andC-terminal fragments, which on their own are non-fluorescent (Fig. 1E). Upon interaction, the fused frag-ments (now in close proximity) can refold, mature, andfluoresce (Ghosh et al., 2000). The method was lateradapted for use in mammalian and plant systems (Huet al., 2002; Bracha-Drori et al., 2004; Walter et al., 2004).

Despite its popularity, BiFC suffers from two majorlimitations: (1) the irreversible complementation of thesplit fluorescent protein fragments with estimated half-life times of 10 years precludes the analysis of interac-tion dynamics (but not necessarily their frequency);and (2) the tendency of most split fragments tospontaneously reassemble autonomously if unhin-dered (Box II;Magliery et al., 2005; Berendzen et al., 2012;Ohashi et al., 2012). Both flaws have the additive effect ofincreasing the likelihood of false positives, as the splitfragments will eventually reassemble on their own; be-cause this self-assembly is effectively irreversible, false-positive signals will accumulate over time. The challengehas been and remains how to determine when this is thecase.

While a range of publications simply show exem-plary images of black negative controls and brightpositive samples without biochemical validation of ex-pression (as exemplified in Fig. 1E), a growingnumber ofresearchers have come to view BiFCwith greater cautiondue to these inherent complications (Lalonde et al., 2008;Grefen and Blatt, 2012a; Horstman et al., 2014). Here,rather than enumerating the different (individual ex-pression) vector systems that have been reported sincethe original development of BiFC, which unfortunatelyhave done little to tackle the issues associated with thespontaneous reassembly of split fluorescent proteinfragments and its irreversibility, this section addressesthe issues around the use of BiFC and their potentialsolutions.

Most in planta BiFC systems rely on overexpressionof the split fluorescent protein fusions, conditions

Table III. Physicochemical properties of FRET pairs used in plant research

Values are from Mas et al. (2000), Ando et al. (2002), Shaner et al. (2005, 2008), Kremers et al. (2006), Akrap et al. (2010), Goedhart et al. (2012),and Wolf et al. (2013).

FRET Pairs Donor

Excitation lmax

Acceptor

Emission lmax

Donor Quantum

Yield

Acceptor Extinction

CoefficientR0

Examples of Fluorescent Protein Pairs

Used in Plant Research

Donor Acceptor Technique Reference

nm M21 cm21 nm

TSapphire mOrange2 399 565 0.6 58,000 4.66 FLIM, AB Bayle et al. (2008)ECFP EYFP 434 527 0.36 72,000 4.72 FLIM Sheerin et al. (2015)mTRQ1 mVenus 434 527 0.84 105,000 5.73 FLIM Laptenok et al. (2014)mTRQ2 mVenus 434 527 0.93 105,000 5.84 FLIM Hecker et al. (2015)mEGFP mCherry 488 610 0.65 72,000 5.31 FLIM, AB Hecker et al. (2015)mEGFP mCherry 488 610 0.65 72,000 5.31 FLIM Kriechbaumer et al. (2015)GFP DsRed1 488 586 0.60 22,500 5.0 SE, AB Mas et al. (2000)mVenus tagRFP 515 584 0.64 100,000 6.15 FLIM Hecker et al. (2015)EYFP mCherry 513 610 0.76 72,000 5.66 SE Seidel et al. (2010)EYFP (Q96K) mStrawberry 514 596 0.76 90,000 6.42 FLIM Adjobo-Hermans et al. (2006)Kaede (green) Kaede (red) 504 578 0.80 60,400 5.74 SE Wolf et al. (2013)





under which irreversible reassembly occurs even in theabsence of an interaction. Hence, complex formation, itsdetection, and its timing are influenced and overlaid byfactors such as protein expression levels, folding effi-ciency, turnover rates, and the steric accessibility of thefluorescent protein, the POI, or both (Box III). In otherwords, it becomes difficult to determine whether fluo-rescence complementation truly informs on an inter-action of the POIs or indicates the presence of anotherfactor that may or may not occur in addition to a trueinteraction. In addition, the irreversible nature of BiFCmay trap the complex where the POIs first met. Theseeming disadvantage of the spontaneous reassemblyof fluorescent protein fragments and its irreversibility,however, can offer other readouts beyond the detectionof PPIs: the folding and topology of membrane proteinsare two parameters that have been assessed successfullyby exploiting the characteristics of BiFC (Cabantous et al.,2005; Zamyatnin et al., 2006; Stefano et al., 2015).

Designing BiFC Experiments

In the following section, we discuss the precautionsthat should be taken when designing a meaningfulBiFC analysis, particularly in terms of those related to(1) construct design, (2) expression system, (3) detectionsystem, and (4) suitable controls.

Construct Design

An important choice to be made is between using anavailable vector system (preferentially a system thatallows cloning of both fusion proteins and a transfor-mation marker from one strand of DNA; Grefen andBlatt, 2012a; Gookin and Assmann, 2014) and buildingone’s own using available cloning techniques such asGreenGate (Lampropoulos et al., 2013), GoldenGate(Engler et al., 2008), Gibson Assembly (Gibson et al.,2009), Multisite Gateway (Sasaki et al., 2004), andthe 2in1 conversion system (Grefen and Blatt, 2012a).For example, expressing fusion proteins from a singleT-DNA causes less variability in their expression levelsin individual cells (Fig. 6A). By contrast, when twoA. tumefaciens strains were mixed with individualplasmids and infiltration into Nicotiana benthamianaleaves, an overall high level of variability of coex-pression was observed, with only approximately 80%of cells cotransfected with both constructs (Hecker et al.,2015). This problem can be overcome through statisticalanalysis, but the situation is further complicated because,like the FRET assay, unequal gene dosage will yieldmisleading results, impeding meaningful analysis whenusing independent T-DNA transformations.

Recently, it was shown that splitting the YFP-derivedVenus at amino acid 210 (instead of the classical aminoacid 155/156 split) causes a significant reduction inbackground fluorescence (Ohashi et al., 2012). This splitsystem has been implemented in a plant vector systemusing the expression of multiple proteins from a singleT-DNA, including a Golgi-localized CFP marker as the

transformation control (Gookin and Assmann, 2014).However, the binary vector system uses classical cloningsites, requiring tedious, sequential cloning. Furthermore,the initial publication only includes exemplary images,with neither quantification nor immunoblot analysis toconfirm protein expression in any of the experiments(Gookin and Assmann, 2014); quantitative confirmationof this split site in planta is still pending.

The other currently availablemulticistronic expressionsystem features the Gateway-compatible 2in1 system,which allows simultaneous transformation of two inde-pendent genes into a single T-DNA. Additionally, thissystem contains a constitutively expressed, soluble redfluorescent protein (RFP) to be used as both transfor-mation control and ratiometric marker (Fig. 6B; Grefenand Blatt, 2012a). This ratiometric BiFC system has beenutilized successfully by a number of different laborato-ries with a diverse range of PPIs (Karnik et al., 2013b;Lipka et al., 2014; Albert et al., 2015; Kumar et al., 2015;Le et al., 2016; Perrella et al., 2016).

Expression System

Should the experiment be performed in a heterolo-gous or homologous expression system, using transienttransfection of leaves or protoplasts, or in stably trans-formed plants? If stable transformation is attempted, theirreversibility of the interaction at hand should be ac-knowledged. Many processes in signaling or traffickingare facilitated by short-lived PPIs, where interaction is asimportant as the eventual release; however, BiFC willliterally cross-link the interaction partners in vivo, pre-venting their release, which might be paramount fortheir physiological function. A possible exception iswhen the actual interaction targets the protein com-plex for degradation, which is not hindered by thecomplemented fluorescent protein. In general, we pre-sume that most interaction couples will not yield viabletransformants independent of the strength or spatialdistribution of the expression. In our own experience, theheterodimerization of the Arabidopsis ethylene recep-tors cannot be detected using BiFC, perhaps due todominant-negative effects of the irreversibly bound re-ceptor dimers (Grefen et al., 2008). Methods allowingreversible and dynamic interactions, such as FRET-basedtechniques, are more appropriate for detecting PPIs instably transformed Arabidopsis lines.

Detection System

The detection offluorophore complementation shouldbe quantitative instead of being based on a biased choiceof exemplary images. As Richard Immink points out inhis excellent critique of the current use of BiFC in plantresearch, “[T]he goal of a BiFC experiment is to obtainstrong support of a protein-protein interaction andnot just to obtain an image of a fluorescent cell”(Horstman et al., 2014). The goal should be to quan-tify the complemented fluorescence (or its absence)


Xing et al.



against a coexpressed fluorophore that serves as both atransformation control of the cell under study and areference marker for ratiometric analysis (Fig. 6B; Gre-fen and Blatt, 2012a). The use of a large sample size alsomay aid in the quantification of BiFC signals, takingadvantage of methods such as flow cytometry, as dis-cussed below.

Suitable Controls

Validating the PPI with a suitable negative control iskey to guarding against false positives. Validation alsois essential for instances in which no interaction is ap-parent. An absence of fluorescence can occur (1) if thecell is untransformed, (2) if the cell is transformed in-completely (with only one of two plasmids), (3) if thereis a lack of expression of either or both fusion proteins,or (4) if the interaction is absent. Indeed, some of themain issues with many published BiFC results are theincorrect use of controls (irrelevant or unsuitable), alack of quantification (e.g. showing images but no effortat statistical analysis), and the failure to verify proteinexpression, such as by immunoblot analysis (Horstmanet al., 2014). It has been suggested that one of thebest negative controls is to include an untagged versionof one of the two interacting POIs as a competitor(Kodama andHu, 2012). The rationale is that one shouldobserve a decrease in complemented YFP in the presenceof the competitor (Fig. 6C). However, we tested thisapproach using the 2in1 system (Grefen and Blatt, 2012a)and a competitor tagged with CFP to verify its expres-sion while quantifying the ratio of its intensity to that of

the complemented YFP. The use of a competing proteindid not lead to a decrease in BiFC (C. Grefen, unpub-lished data). This result is not surprising given theinherent feature of irreversible YFP complementation.Similarly, Morell et al. (2008) found that BiFCwas onlyblocked if a specific competitor was added before thePOIs were induced, and similarly, it could not be re-versed after complex formation. An equilibrium gener-ally can be influenced by the presence of a competitor,but irreversible product formation shifts the equilibriumto the product side (to the right in Fig. 6C). Hence, if thisanalysis is to be of any use, competition must be moni-tored in a tight time frame, as differences in the kineticsof only YFP formation might be detectable. Steady-statecompetition experiments only make sense using a trulyreversible detection system. Recently, a method waspublished featuring a reversible BiFC approach basedon a phytochromic fluorophore (To et al., 2016). It re-mains to be seenwhether this system can overcome thelimitations associated with classical BiFC.

Taken together, when selecting a BiFC system (pro-duced in house or previously reported), the strength ofthe promoter, tag orientation, type of split fluorophore(physicochemical properties, split sites), and, most of all,careful choice of controls are important factors to con-sider. To gain credible results from BiFC analysis, futurework needs to incorporate what was learned from pastexperience: it is essential to ensure the expression of allfusion proteins from the same stretch of DNA, to includea fluorescent marker allowing ratiometric analysis of aninteraction (Grefen and Blatt, 2012a), or to use a combi-nation of these approaches with split fluorescent protein

Box III: The Special Case of PCAs

Split protein fragments often carry inherent affinity toward each other. Hence, determining the suitability of a PCA-based technique for a givenPPI depends on the Kd values of the split probes and needs to take account of the threshold of a Kd where the association of these probes becomesthe driving force of an interaction rather than the PPI in question (Box II; Muller and Johnsson, 2008). The balance between PPI affinities and theaffinity of the marker or tag is an inherent issue of the heavily used BiFC technique, where the spontaneous reassembly of the unhindered non-fluorescent fragments and the irreversibility of their reassociation result in a negligibly small Kd (see “BiFC: From Cherry Picking toQuantification”; Do and Boxer, 2011). In fact, an excess of more than 100 M peptide competition cannot disrupt a reconstituted YFPmolecule, which remains intact even when incubated in 2 M urea for several days (Morell et al., 2007). This clearly highlights the notion thata PPI method does not allow conclusions to be drawn on the dynamics of a transient PPI if the Kd of the split probes is orders of magnitude smallerthan that of the PPI couple of interest. We stress that useful biological information still can be obtained with BiFC, but the user should be awarethat the BiFC readout is related to the KON and is normally not related to the KOFF of a POI and that other factors likely play a large role in BiFC(Kd is the ratio of KOFF to KON).

Many additional interdependent factors obscure the results of PCA-based techniques. For example, the localization and concentration of splitprobes fused to POIs have strong effects on reassembly: PPIs between membrane proteins can report artifactual (positive) results through anincrease in the concentration of bait and prey in the two-dimensional membrane space and the resulting increased collision probability. Also,proteostatic effects, including maturation time, stability, and turnover rates of both the POI and split probes, contribute to the signal strength of aPCA method. Not all proteins are equally stable in a cell, which presents issues, especially for in vivo experiments in which POIs may beregulated natively and independently. Furthermore, the split probes fused to the POIs may influence the turnover of the whole fusion protein,resulting in the presence of suboptimal (or at least unequal) amounts of bait and prey fusions. A delay in the maturation of one POI as opposed tothe predicted partner affects kinetics during reassembly. Equally important is the steric accessibility of split probes in fusions for reconstitution: aninteraction between bait and prey can result in occlusion of the interaction surfaces of the split probes, resulting in a false-negative readout. Suchproblems might be resolved through the addition of flexible linkers, and they can provide information about protein structure. All of these factorsmust be considered when choosing the optimal PCA technique.





fragments that might have less intrinsic affinity towardeach other (Ohashi et al., 2012).

In any event, fluorescence should be detected asearly as possible and not at a time point whenall possible interactions are complete (or saturated),as recommended by some. With many proteins ex-pressed using the 35S promoter, the highest expres-sion levels occur after 3 d (Grefen et al., 2010b), anobservation that underpins the importance of mea-suring BiFC fluorescence after 1 d (or a maximum of2 d) post infiltration; weak expression levels often canbe informative in these circumstances. Time-lapseanalysis over a time frame of hours could also distin-guish true from false interactions and, depending onthe split sites of the fluorescent protein fragments,might even allow weak and strong interactions to be

distinguished. Using a large sample size also mighthelp sift the chaff from the wheat, for example, bytaking advantage of protoplast expression systemsand flow cytometry.

Flow Cytometry Lights the Way for Quantitative BiFC

Over the past decade, various laboratories havecoupled BiFC analysis with flow cytometry for analysis,sorting (fluorescence-activated cell sorting), or both(Miller et al., 2015). Flow cytometry is a technique usedto examine a population of particles or cells in freesuspension, one by one, typically by collecting andanalyzing fluorescence and scattered light. This tech-nique allows individualfluorescence emission intensities

Figure 6. Ratiometric BiFC (rBiFC). A, Cotransformation ratios of transiently transformed N. benthamiana. Mixing of A. tumefacienswith two individual plasmids each resulted in approximately 77.6% cotransformed cells (n = 1,221 cells), whereas transformation withone plasmid containing both expression constructs yielded 99.3%cotransformation (n=1,099 cells). Cotransformation is characterizedby a ratio of mCherry to mEGFP intensity of between 0.2 and 5, with ratios below and above considered to indicate non-cotransformedcells. Sample confocal images, from left to right, are as follows: mCherry fluorescence, mEGFP fluorescence, and merged channels(image modified fromHecker et al., 2015). B, Sample ratiometric BiFC experiment with a T-DNA cartoon at the top (data fromHeckeret al., 2015). Cytosolic nYFP-SNAP33 and either plasma membrane-localized cYFP-SYP121 or pre-vacuolar compartment/tonoplastmembrane-localized cYFP-SYP21 were coexpressed from a single contiguous T-DNA including a reference RFP marker (Grefen andBlatt, 2012a). The ratio of complemented YFP to RFP was calculated from 20 complete individual images, each recorded at the in-terfaces of two to three cells to avoid bias arising from choosing single cells. As the interaction is at the cell periphery, imaging of thenucleus was prevented to avoid a disproportionate increase in the detection of soluble RFP, whichmigrates into the nucleus andwouldbias this assay. Data from 20 independent YFP-RFP ratios are shown as box-plot diagrams (the light gray box represents the 25% to 50%range and the dark gray box represents the 50% to 75% range of where all values lay).Whiskers mark the 1.53 interquartile range, andblack dots indicate outliers. C, Potential interactions in a BiFC experimentwhen a competitor is coexpressed. From left to right, dynamicbinding of a potential transient interaction causes alternation in binding of the prey protein (gray) to either the tagged or untagged bait(red). However, if the fluorescent protein fragments are allowed to interact long enough, the irreversibility of their complementationdepletes them from the equation. The signal accumulates and reports an interaction despite the presence of a competitor.


Xing et al.



to be obtained for large populations (n . 10,000) in avery short time (5 min or less). Flow cytometry turns thefact that BiFC is irreversible from a problem into an at-tribute, as trapped, weak interactions can be detectedand quantified using this method (Morell et al., 2007;Berendzen et al., 2012).Ozalp et al. (2005) found that, due to its irreversibility,

BiFC gives an output representing a summation of pro-tein interactions over time. Looking at a large enoughpopulation of cells, one can observe variance in signalstrength. In order for a BiFC signal to be scored, enoughreconstituted fluorescent protein must accumulate toexceed the intrinsic autofluorescence of an individualcell. Articles have been published showing only thepercentage of cells with BiFC signals (Llorca et al., 2015),only their intensity (Kimet al., 2012;Wang andCarnegie,2013), or a combination thereof (Li et al., 2010; Lee et al.,2011; Berendzen et al., 2012). Variance in interactionstrengths, bothweak and strong,were observed in cross-check glutathione S-transferase pull-down assays (Wangand Carnegie, 2013), lending further support to the no-tion that BiFC flow cytometry indeed captures varyingdegrees of interaction strengths and are not artifacts oftransfection variance.Where can we expect to observe weak interactions

and how do we know they are genuine? According toHu et al. (2002), the split fluorescent protein fragmentsare estimated to assemble with a half-time of 60 s (Huet al., 2002). Due to the irreversibility of BiFC, it hasbeen proposed that BiFC might be able to capture anenzyme interacting with its substrate (Morell et al.,2007;Miller et al., 2015). In this case, one can expect veryweak (infrequent) signals if we presume that the PCA isdominated by the POIs. For example, in one study, arandom interaction screen using BiFC flow cytometrywas followed by FLIM-FRET analysis (Berendzen et al.,2012). The screen captured eight interactions by BiFC,including one thatwas very strong, three thatwereweak,and four that were extremely weak but identifiable. Allinteractions were shown to be positive via FLIM-FRET inheterologous N. benthamiana leaf assays. Most of theweak BiFC outputs also gave weak (but significant)FLIM-FRET signals, and likewise, strong BiFC signalsgave strong FLIM-FRET. These results suggest that theweak interactions detected in vivo by BiFC flow cy-tometry were not artifacts and that the implementationof an arbitrary threshold would have eliminated themfrom the analysis (Berendzen et al., 2012).A weak BiFC signal may occur if a protein is de-

graded faster than other proteins and, hence, reportsturnover in addition to interaction, as observed for theRUM1 IAA protein (von Behrens et al., 2011). Thus,although not thoroughly demonstrated, BiFC flow cy-tometry appears to be able to capture informationpertaining to interaction frequency and binding strengthby providing a quantitative measure of BiFC. The com-bination of flow cytometry and ratiometric BiFC (Grefenand Blatt, 2012a) should be an even more powerful ap-proach, as the intrinsic RFP helps to exclude untrans-fected protoplasts.

SPLIT-LUCIFERASE COMPLEMENTATION ASSAY:LET THERE BE LIGHT

Originally developed for use in mammalian cells(Ozawa et al., 2001), the split-luciferase assay and itssubsequent complementation have since been estab-lished successfully in plants to identify a variety of PPIs(Fujikawa and Kato, 2007; Li et al., 2011; Lin et al., 2015;Lund et al., 2015). The split-luciferase complementationassay is a PCA that relies on the complementation oftwo split fragments of a luciferase protein and its sub-sequent detection through substrate conversion (Fig.1F). Light-producing species can be found in almost allphyla of life. From bacteria and fungi to vertebrates, fromland-dwelling hexapods to marine mollusks and fishes,bioluminescence has evolved convergently in as manyas 40 different species (Haddock et al., 2010). The termsluciferase and luciferin are generic and are used forall classes of light-producing enzymes; their respectivesubstrates are usually preceded by the species name.Photon emission emerges as a side product of oxidativedecarboxylation of a luciferin by its corresponding lucif-erase (Woo et al., 2008; Marques and Esteves da Silva,2009). Similarities in substrate conversion aside, lucifer-ases differ substantially from each other in terms ofstructure, size, substrate specificity, and cofactors. TableIV provides an overview of the different systems used inplanta, where the luciferases from the North Americanfirefly (Photinus pyralis) and the sea pansy (Renilla reni-formis) are utilized exclusively, while inmammalian cells,other luciferases derived from click beetle, Gaussia spp.and nanoLUC also have been used (Remy andMichnick,2006; Kim et al., 2007; Dixon et al., 2016).

Firefly Versus Renilla: Comparison of Properties

Firefly luciferase is a 62-kD protein comprising asingle polypeptide chain folded into a large N-terminaldomain and a small C-terminal domain (Conti et al.,1996; Thorne et al., 2010). This enzyme uses D-luciferinas a substrate. In the presence of Mg2+, ATP, andoxygen, firefly luciferase catalyzes a two-step reaction,converting D-luciferin into oxyluciferin, AMP, andCO2, thereby releasing a photon (Marques and Estevesda Silva, 2009). The emission is visualized in the green-yellow spectrum of light (540–570 nm). The reactioncreates inhibitory products, leading to an inactivationof firefly luciferase that can be counteracted by CoA(de Ruijter et al., 2003; Fraga et al., 2005; da Silva andda Silva, 2011). The reaction is pH and temperaturesensitive and will undergo a red shift if firefly lucif-erase is exposed to acidic pH or higher temperatures(Seliger and McElroy, 1964; Fraga, 2008).

Renilla luciferase is 37 kD in size and also consists ofa single polypeptide. Therefore, the split forms of thisenzyme result in moieties that are considerably smallerthan the split firefly PCAs. The size of the split probescan become an important factor when analyzing PPIs ofsmall proteins. Here, large probes could potentiallymask interaction sites and prevent the detection of a





PPI. The substrate of Renilla luciferase is coelenterazine,which is catalyzed into coelenteramide and CO2 underthe consumption of oxygen, leading to the emission ofblue light (480–510 nm; Matthews et al., 1977; Wooet al., 2008).

Firefly luciferase emits green-yellow to red light,which, however, is less bright than the blue light thatRenilla luciferase produces. Also, firefly luciferase re-quires ATP as a cofactor, which might not be sufficientlyaccessible in all cellular compartments or might affectendogenous pools, thereby interfering with cellularphysiology. Renilla luciferase does not require ATPto catalyze its reaction; however, coelenterazine, thesubstrate of Renilla luciferase, is unstable and can un-dergo spontaneous oxidation (Zhao et al., 2004),whereasD-luciferin is sufficiently stable to be used successfully tomonitor circadian rhythms (Millar et al., 1992). Interest-ingly, coelenterazine can be transported in human cellsby a P-glycoprotein (Pichler et al., 2004). This family ofproteins also exists in plants and plays a role in polarauxin transport (Geisler and Murphy, 2006).

While split-luciferase complementation is a useful invivo technique, it is equally suitable for in vitro analysis(Ohmuro-Matsuyama et al., 2013). Reconstitution ofluciferase is in many cases reversible (Stefan et al., 2007;Li et al., 2011; Lund et al., 2015), and, at least for fireflysplit-luciferase complementation, the emission of lightis dependent on the interaction of the fused partners(Dale et al., 2016). In addition, the short half-life of theluciferases (firefly luciferase, approximately 3 h; Renillaluciferase, approximately 4.5 h; Thorne et al., 2010)makes split-luciferase complementation an ideal sys-tem for monitoring dynamic interactions in real-timeand makes it suitable for high-throughput analysis (Liet al., 2011; Lund et al., 2015). With regard to freefragments, any possible background from the fireflyluciferase moiety is orders of magnitude lower thanthat of the reconstituted complex or the full-lengthprotein (Dale et al., 2016).

As plants are not bioluminescent organisms, split-luciferase complementation is a very sensitive detectionsystem, giving high signal-to-noise ratios. The intrinsiclight production via luciferase avoids the use of exog-enous light sources, preventing plants or cells fromphotodamage or a decline in signal intensity due tobleaching. However, luciferase activity must be mea-sured in the dark, which should be considered, forexample, when studying light-dependent processes.Finally, the substrates of luciferase, D-luciferin and

coelenterazine, must be exogenously applied. In planta,this can be easily achieved via the incubation of proto-plasts or watering (i.e. spraying the plants with dis-solved luciferin; Chen et al., 2008).

The advantages of the split-luciferase complementa-tion assay include flexible temporal control with fastturnover and/or reversibility of probe assembly, therebyallowing for highly dynamic PPI analysis in planta.Nevertheless, the technique appears to be neglectedamong PPI techniques used by plant scientists (Fig. 2C).Implementation of other luciferases (Gaussia spp., clickbeetle, or nanoLUC), which may be more suitable forthe demands of plant systems, will likely pull the split-luciferase complementation assay out of its nicheexistence. In fact, the nanoLuc split-luciferase comple-mentation systemNanoBiT, which has been engineeredto be ideal for most PPIs (Dixon et al., 2016), willhopefully be available for plant scientists in the nearfuture.

IN PLANTA COIP: ONE TO BIND AND ONE TO FIND

Given the range of techniques featured in this review,CoIP appears to be the odd one out, a chimera betweenan in vivo and in vitro system. Correctly labeled an exvivo technique, CoIP is often used as an alternativemethod to one of the above systems, allowing thedetection of an interaction in specific tissues or de-velopmental stages, under experimentally determinedconditions, and/or in different genetic backgrounds.

Historically speaking, CoIP was adapted from col-umn affinity chromatography (or affinity purification;Cuatrecasas et al., 1968), which is based on the fact thatany given biomolecule usually has an inherent recog-nition site that can be bound by another natural or ar-tificial molecule (e.g. an enzyme binding its substrate orinhibitor; Cuatrecasas et al., 1968; Magdeldin andMoser, 2012). Thus, affinity purification is principallybased on molecular recognition between a matrix-bound molecule and the target molecule in an extractthat is presented to it (Fig. 1C). This approach wasoriginally used to purify biomolecules of interest from asolution or cell lysate. The technique is simple, involv-ing the setting up of a column and binding of the targetmolecule, followed by washing and elution, with allsteps processed consecutively in the column (Magdeldinand Moser, 2012).

CoIP can be categorized as a specialized or small-scaleaffinity purification that exploits the antibody-antigen

Table IV. Available split-luciferase fragments and their first use in planta

LuciferaseAmino Acids

Expression System ReferencesNLuc CLuc

Renilla 1–229 230–311 A. thaliana protoplasts Fujikawa and Kato (2007)Renilla 1–110 111–310 N. benthamiana leaf extracts Lund et al. (2015)Firefly 2–416 398–550 A. thaliana protoplasts Chen et al. (2008)

N. benthamiana leavesFirefly 1–398 394–550 A. thaliana protoplasts Li et al. (2011)


Xing et al.



immune reaction (Kaboord and Perr, 2008). The prin-ciple of CoIP is that the POI or antigen forms an im-mune complex in a cell lysate with its specific antibody.As the antibody is bound to beads (agarose or mag-netic), which are usually coated with proteins (e.g.Protein A) to capture antibodies, the immune complexcan be precipitated easily. Unspecifically bound pro-teins can be washed off using appropriate detergentsand salt conditions, allowing purification of the POI inits complex. After elution, the interacting candidatescan be analyzed by various assays such as westernblotting, ELISA, or MS (Butler et al., 1978; Masters,2004; Monti et al., 2005). The combination of CoIP andMS analysis is a particularly powerful screening tool fordiscovering novel interactors or establishing interactionmaps of a POI (Miernyk and Thelen, 2008).CoIP often is performed using native antibodies

raised against the POIs or using epitope tags (Qoronflehet al., 2003). Notably, the discovery that antibodiesof the camelid family require only part of the H chainfor antigen recognition, the so-called nanobody,has greatly enhanced binding specificity while re-ducing incubation times (Hamers-Casterman et al.,1993; Harmsen and De Haard, 2007; Muyldermans,2013).Nevertheless, weak or transient interactions are dif-

ficult to detect using CoIP (Lee et al., 2013; Avila et al.,2015), although cross-linking can be used to help sta-bilize the protein complexes (Garcia-Molina et al., 2014;Pertl-Obermeyer et al., 2014). Additionally, an interac-tion in CoIP could be due to the presence of a proteincomplex of multiple interactors, with an apparent bi-nary interaction facilitated by known or unknowncofactors (Hall, 2005). This comes as a blessing indisguise, as it allows higher order protein complexes orligand-induced dimerization to be dissected, a phe-nomenon that is neatly exploited when an interactionbetween two POIs is either induced by an exogenouslyapplied factor (ligand, pathogen, and so on) or dis-rupted in a knockout line of the endogenous bridgingfactor (Chinchilla et al., 2007; Albert et al., 2015; Fuchset al., 2016)Recently, two novel techniques have been developed

that circumvent the limitations of traditional CoIP. Bothdevelopments, single-molecule pull down (Jain et al.,2011) and real-time single-molecule CoIP (Lee et al.,2013), use total internal reflection microscopy (Axelrod,2001) to detect interactions on a coverslip. Bothmethodsinvolve immobilizing the bait POI on the surface of acoverslip while the prey protein is labeled with a fluo-rescent tag. Interaction can be visualized and quantifiedthrough the emergence of a fluorescent spot. Usingmultiple prey proteins with different fluorophores al-lows for the exploration and permutation of higher orderprotein complexes. Additionally, combining total inter-nal reflection microscopy with photobleaching or life-time imaging allows for stoichiometric analysis (Gustet al., 2014). These methods have not yet been applied inthe plant field but will surely provide a powerful plat-form in the near future.

CONCLUSION

Even if all points outlined in the introduction aretaken into account, one is spoiled for choice in selectingthe right PPI technique to detect a POI couple. Everymethod comes with advantages and drawbacks, un-derpinning the importance of using combinationsof alternative detection methods. Our analysis alsohighlights how vital it is to have a basic understand-ing of the biology behind each method. An example(one that is far too often overlooked) is the issue ofprobe affinity in PCAs (Table in Box II). While it mightbe beneficial to use a method that detects weak,transient interactions by literally cross-linking them,this irreversible fusion will undoubtedly have an impacton the detection system (organism) and, therefore, re-quires the use of appropriate negative controls.

Certainly, the best negative controls in all PPI ex-periments are point-mutated POIs using site-directedmutagenesis approaches (Karnik et al., 2013a), as longas the stability, turnover, and localization are notinfluenced by the mutation. Conversely, if such amutation does not affect the interaction signal, theuser may be justified in his/her suspicions that thePCA detects other factors (see Box III). These factorsmight provide biologically important insights, butthey may not necessarily inform on any interaction.

Some of the challenges of PCA techniques, as men-tioned above, need to be assessed when designing newprobes for improved PPI detection methods. For ex-ample, although not yet tested in plants, the design ofthe NanoBiT system has involved determining the Kdof the free fragments, characterizing their turnoverrates, and documenting differences in steric accessi-bility (Dixon et al., 2016).

Reversibility in probe interactions is an importantfeature, as it reduces the rate of false positives and al-lows kinetic analysis of interactions (Michnick et al.,2007). While two-hybrid methods such as yeast two-hybrid analysis and FRET can be used to study dy-namic interactions, PCA-based methods usually sufferfromprobe affinity and, in the case of BiFC, irreversibility

CHALLENGES

Future developments will need to extend PPI methods toaccommodate the detection of multiple interaction partners or

competitors to help dissecting protein complexes.

For the detection of transient interactions, probe affinity hasto be adjusted to allow for kinetic and dynamic analysis of –

ideally – single molecule interactions.

Protein expression under native, cell-type-specific or inducibleconditions should allow spatiotemporal resolution of interactionsthereby distinguishing biologically meaningful interactions

from artifacts potentially caused through aberrant(mislocalized) overexpression.

Side effects of unequal gene dosage causing unbalancedstoichiometry of interactions partners need to be avoided

and intrinsic controls – not least to aid in quantification – shouldbe implemented.





of assembly. However, even in a reversible PCAmethodsuch as the split-ubiquitin system, the finite cleavage ofthe reporter and its resulting signal amplification pre-clude dynamic PPI analysis. Fluorescent reporters are thepreferred choice for such kinetic studies, as the signal isgenerally lost upon dissipation of the interaction, at leastfor FRET or light emission using the PCA split-luciferasecomplementation techniques.

While onemethod alone is never sufficient to validatea PPI, even when three different techniques are used, thefailure of one of these techniques to show an interactiondoes not dismiss the other two positive results. Varioustechnical features of themethodmight be the underlyingreason for its failure. For example, if a binary interactionis in reality trimeric and the third method lacks the ad-ditional factor, such information would represent abiological explanation worthy of further attention.

OUTLOOK

In this omics age of ours, a gap remains between thesheer scale of the data accumulated and our under-standing of basic fundamental processes at the molec-ular level. As informative as large-scale PPI approachesmay be, deciphering binary interactions and elucidatingcomponents of small protein complexes, their dynamics,stoichiometry, and structure-function relationships willultimately provide the molecular and functional detailsthat are necessary to decode cellular mechanisms.

Newly developed, sophisticated, superresolutionmicroscopy techniques such as structured illuminationmicroscopy, stimulated emission depletion micros-copy, photoactivated localization microscopy, andstochastic optical reconstruction microscopy make itfeasible to detect PPIs at the single-molecule level(Gutierrez et al., 2010; Komis et al., 2015; Sydor et al.,2015). In addition, spectromicroscopy techniques(Harter et al., 2012) such asfluorescence cross-correlationspectroscopy (Macháň and Wohland, 2014) representtechnological improvements that greatly improve thedetection of the spatiotemporal resolution of interactions,particularly in (but not limited to) membrane environ-ments. While these novel techniques allow dynamic PPIstudies to be performed in an in vivo environment, thegreatest limitation to their application at present remainsthe expensive equipment needed for their execution.

ACKNOWLEDGMENTS

We thank Mike Blatt, Theodorus Gadella, Klaus Harter, Nils Johnsson,Thorsten Seidel, Marja Timmermans, and Sven zur Oven-Krockhaus for criticalcomments and discussions during article and figure preparation, our reviewers,and Jennifer Lockhart, science editor at the American Society of Plant Biologists,whose comments helped to greatly improve the article. Our apologies toresearchers whose work we have not cited due to space limitations.

Received March 24, 2016; accepted April 19, 2016; published April 25, 2016.

LITERATURE CITED

Aakre CD, Herrou J, Phung TN, Perchuk BS, Crosson S, Laub MT (2015)Evolving new protein-protein interaction specificity through promiscu-ous intermediates. Cell 163: 594–606

Adjobo-Hermans MJ, Goedhart J, Gadella TW Jr (2006) Plant G proteinheterotrimers require dual lipidation motifs of Galpha and Ggamma anddo not dissociate upon activation. J Cell Sci 119: 5087–5097

Akrap N, Seidel T, Barisas BG (2010) Förster distances for fluorescenceresonant energy transfer between mCherry and other visible fluorescentproteins. Anal Biochem 402: 105–106

Albert I, BohmH,AlbertM, Feiler CE, Imkampe J,WallmerothN, BrancatoC, Raaymakers TM, Oome S, Zhang HQ, et al (2015) An RLP23-SOBIR1-BAK1 complex mediates NLP-triggered immunity. Nat Plants 1: 15140

Ando R, Hama H, Yamamoto-Hino M, Mizuno H, Miyawaki A (2002) Anoptical marker based on the UV-induced green-to-red photoconversionof a fluorescent protein. Proc Natl Acad Sci USA 99: 12651–12656

Aranda-Sicilia MN, Trusov Y, Maruta N, Chakravorty D, Zhang Y, Botella JR(2015) Heterotrimeric G proteins interact with defense-related receptor-likekinases in Arabidopsis. J Plant Physiol 188: 44–48

Avila JR, Lee JS, Torii KU (2015) Co-immunoprecipitation of membrane-bound receptors. The Arabidopsis Book 13: e0180. doi: 10.1199/tab.0180

Axelrod D (2001) Total internal reflection fluorescence microscopy in cellbiology. Traffic 2: 764–774

Banaszynski LA, Liu CW, Wandless TJ (2005) Characterization of theFKBP.rapamycin.FRB ternary complex. J Am Chem Soc 127: 4715–4721

Bayle V, Nussaume L, Bhat RA (2008) Combination of novel green fluo-rescent protein mutant TSapphire and DsRed variant mOrange to set upa versatile in planta FRET-FLIM assay. Plant Physiol 148: 51–60

Berendzen KW, Böhmer M, Wallmeroth N, Peter S, Vesi�c M, Zhou Y,Tiesler FK, Schleifenbaum F, Harter K (2012) Screening for in plantaprotein-protein interactions combining bimolecular fluorescence com-plementation with flow cytometry. Plant Methods 8: 25

Besserer A, Burnotte E, Bienert GP, Chevalier AS, Errachid A, Grefen C,Blatt MR, Chaumont F (2012) Selective regulation of maize plasmamembrane aquaporin trafficking and activity by the SNARE SYP121.Plant Cell 24: 3463–3481

Bhat RA (2009) FRET and FLIM applications in plants. Lab Tech BiochemMol Biol 33: 413–445

Blatt MR, Grefen C (2014) Applications of fluorescent marker proteins inplant cell biology. Methods Mol Biol 1062: 487–507

Borghi L (2010) Inducible gene expression systems for plants. Methods MolBiol 655: 65–75

Boruc J, InzéD,Russinova E (2010)A high-throughput bimolecularfluorescencecomplementation protein-protein interaction screen identifies functionalArabidopsis CDKA/B-CYCD4/5 complexes. Plant Signal Behav 5: 1276–1281

Boute N, Jockers R, Issad T (2002) The use of resonance energy transfer in high-throughput screening: BRET versus FRET. Trends Pharmacol Sci 23: 351–354

Bracha-Drori K, Shichrur K, Katz A, Oliva M, Angelovici R, Yalovsky S,Ohad N (2004) Detection of protein-protein interactions in plants usingbimolecular fluorescence complementation. Plant J 40: 419–427

Braun P, Aubourg S, Van Leene J, De Jaeger G, Lurin C (2013) Plantprotein interactomes. Annu Rev Plant Biol 64: 161–187

Braun P, Gingras AC (2012) History of protein-protein interactions: fromegg-white to complex networks. Proteomics 12: 1478–1498

Brent R, Finley RL Jr (1997) Understanding gene and allele function withtwo-hybrid methods. Annu Rev Genet 31: 663–704

Bücherl CA, Bader A, Westphal AH, Laptenok SP, Borst JW (2014) FRET-FLIM applications in plant systems. Protoplasma 251: 383–394

Bücherl CA, van Esse GW, Kruis A, Luchtenberg J, Westphal AH, Aker J,van Hoek A, Albrecht C, Borst JW, de Vries SC (2013) Visualization ofBRI1 and BAK1(SERK3) membrane receptor heterooligomers duringbrassinosteroid signaling. Plant Physiol 162: 1911–1925

Buhr F, Jha S, Thommen M, Mittelstaet J, Kutz F, Schwalbe H, RodninaMV, Komar AA (2016) Synonymous codons direct cotranslational fold-ing toward different protein conformations. Mol Cell 61: 341–351

Butler JE, Feldbush TL, McGivern PL, Stewart N (1978) The enzyme-linked immunosorbent assay (ELISA): a measure of antibody concen-tration or affinity. Immunochemistry 15: 131–136

Butt TR, Sternberg EJ, Gorman JA, Clark P, Hamer D, Rosenberg M,Crooke ST (1984) Copper metallothionein of yeast, structure of the gene,and regulation of expression. Proc Natl Acad Sci USA 81: 3332–3336

Cabantous S, Terwilliger TC, Waldo GS (2005) Protein tagging and de-tection with engineered self-assembling fragments of green fluorescentprotein. Nat Biotechnol 23: 102–107

Causier B, Ashworth M, Guo W, Davies B (2012) The TOPLESS inter-actome: a framework for gene repression in Arabidopsis. Plant Physiol158: 423–438


Xing et al.


http://dx.doi.org/10.1199/tab.0180


Chang IF, Curran A, Woolsey R, Quilici D, Cushman JC, Mittler R, Harmon A,Harper JF (2009) Proteomic profiling of tandem affinity purified 14-3-3 proteincomplexes in Arabidopsis thaliana. Proteomics 9: 2967–2985

Chartier M, Gaudreault F, Najmanovich R (2012) Large-scale analysis ofconserved rare codon clusters suggests an involvement in co-translational molecular recognition events. Bioinformatics 28: 1438–1445

Chen H, Zou Y, Shang Y, Lin H, Wang Y, Cai R, Tang X, Zhou JM (2008)Firefly luciferase complementation imaging assay for protein-proteininteractions in plants. Plant Physiol 146: 368–376

Chen J, Lalonde S, Obrdlik P, Noorani Vatani A, Parsa SA, Vilarino C,Revuelta JL, Frommer WB, Rhee SY (2012) Uncovering Arabidopsismembrane protein interactome enriched in transporters using mating-basedsplit ubiquitin assays and classification models. Front Plant Sci 3: 124

Chen X, Zaro JL, Shen WC (2013) Fusion protein linkers: property, designand functionality. Adv Drug Deliv Rev 65: 1357–1369

Chinchilla D, Zipfel C, Robatzek S, Kemmerling B, Nürnberger T, JonesJD, Felix G, Boller T (2007) A flagellin-induced complex of the receptorFLS2 and BAK1 initiates plant defence. Nature 448: 497–500

Clegg RM (2009) Förster resonance energy transfer: FRET what is it, whydo it, and how it’s done. Lab Tech Biochem Mol Biol 33: 1–57

Conti E, Franks NP, Brick P (1996) Crystal structure of firefly luciferase throwslight on a superfamily of adenylate-forming enzymes. Structure 4: 287–298

Cuatrecasas P, Wilchek M, Anfinsen CB (1968) Selective enzyme purifi-cation by affinity chromatography. Proc Natl Acad Sci USA 61: 636–643

Cui J, Li P, Li G, Xu F, Zhao C, Li Y, Yang Z, Wang G, Yu Q, Li Y, et al(2008) AtPID: Arabidopsis thaliana Protein Interactome Database—anintegrative platform for plant systems biology. Nucleic Acids Res 36:D999–D1008

Dale R, Ohmuro-Matsuyama Y, Ueda H, Kato N (2016) Mathematicalmodel of the firefly luciferase complementation assay reveals a non-linear relationship between the detected luminescence and the affinityof the protein pair being analyzed. PLoS ONE 11: e0148256

da Silva LP, da Silva JC (2011) Kinetics of inhibition of firefly luciferase bydehydroluciferyl-coenzyme A, dehydroluciferin and L-luciferin. Photo-chem Photobiol Sci 10: 1039–1045

Day RN, Davidson MW (2009) The fluorescent protein palette: tools forcellular imaging. Chem Soc Rev 38: 2887–2921

de Folter S, Immink RG, Kieffer M, Parenicová L, Henz SR, Weigel D,Busscher M, Kooiker M, Colombo L, Kater MM, et al (2005) Com-prehensive interaction map of the Arabidopsis MADS box transcriptionfactors. Plant Cell 17: 1424–1433

Dell A, Galadari A, Sastre F, Hitchen P (2010) Similarities and differencesin the glycosylation mechanisms in prokaryotes and eukaryotes. Int JMicrobiol 2010: 148178

de Ruijter NCA, Verhees J, van Leeuwen W, van der Krol AR (2003)Evaluation and comparison of the GUS, LUC and GFP reporter systemfor gene expression studies in plants. Plant Biol 5: 103–115

Dirnberger D, Messerschmid M, Baumeister R (2008) An optimized split-ubiquitin cDNA-library screening system to identify novel interactors ofthe human Frizzled 1 receptor. Nucleic Acids Res 36: e37

Dirnberger D, Unsin G, Schlenker S, Reichel C (2006) A small-molecule-protein interaction system with split-ubiquitin as sensor. ChemBioChem7: 936–942

Dixon AS, Schwinn MK, Hall MP, Zimmerman K, Otto P, Lubben TH,Butler BL, Binkowski BF, Machleidt T, Kirkland TA, et al (2016)NanoLuc complementation reporter optimized for accurate measure-ment of protein interactions in cells. ACS Chem Biol 11: 400–408

Do K, Boxer SG (2011) Thermodynamics, kinetics, and photochemistry ofb-strand association and dissociation in a split-GFP system. J Am ChemSoc 133: 18078–18081

Duan G, Walther D (2015) The roles of post-translational modifications in thecontext of protein interaction networks. PLOS Comput Biol 11: e1004049

Dünkler A, Rösler R, Kestler HA, Moreno-Andrés D, Johnsson N (2015)SPLIFF: a single-cell method to map protein-protein interactions in timeand space. Methods Mol Biol 1346: 151–168

Ehlert A, Weltmeier F, Wang X, Mayer CS, Smeekens S, Vicente-Carbajosa J,Dröge-Laser W (2006) Two-hybrid protein-protein interaction analysis inArabidopsis protoplasts: establishment of a heterodimerization map ofgroup C and group S bZIP transcription factors. Plant J 46: 890–900

Ehrhardt D (2003) GFP technology for live cell imaging. Curr Opin PlantBiol 6: 622–628

Engler C, Kandzia R, Marillonnet S (2008) A one pot, one step, precisioncloning method with high throughput capability. PLoS ONE 3: e3647

Estojak J, Brent R, Golemis EA (1995) Correlation of two-hybrid affinitydata with in vitro measurements. Mol Cell Biol 15: 5820–5829

Fields S (2009) Interactive learning: lessons from two hybrids over twodecades. Proteomics 9: 5209–5213

Fields S, Song O (1989) A novel genetic system to detect protein-proteininteractions. Nature 340: 245–246

Fontenot EB, Ditusa SF, Kato N, Olivier DM, Dale R, Lin WY, Chiou TJ,Macnaughtan MA, Smith AP (2015) Increased phosphate transport ofArabidopsis thaliana Pht1;1 by site-directed mutagenesis of tyrosine 312may be attributed to the disruption of homomeric interactions. PlantCell Environ 38: 2012–2022

Förster T (1946) Energiewanderung und Fluoreszenz. Naturwissenschaften33: 166–175

Fraga H (2008) Firefly luminescence: a historical perspective and recentdevelopments. Photochem Photobiol Sci 7: 146–158

Fraga H, Fernandes D, Fontes R, Esteves da Silva JC (2005) Coenzyme Aaffects firefly luciferase luminescence because it acts as a substrate andnot as an allosteric effector. FEBS J 272: 5206–5216

Fricker M, Runions J, Moore I (2006) Quantitative fluorescence micros-copy: from art to science. Annu Rev Plant Biol 57: 79–107

Frommer WB, Davidson MW, Campbell RE (2009) Genetically encodedbiosensors based on engineered fluorescent proteins. Chem Soc Rev 38:2833–2841

Fuchs R, Kopischke M, Klapprodt C, Hause G, Meyer AJ, SchwarzlanderM, Fricker MD, Lipka V (2016) Immobilized subpopulations of leafepidermal mitochondria mediate PEN2-dependent pathogen entrycontrol in Arabidopsis. Plant Cell 28: 130–145

Fujikawa Y, Kato N (2007) Split luciferase complementation assay to studyprotein-protein interactions in Arabidopsis protoplasts. Plant J 52: 185–195

Fujiwara M, Uemura T, Ebine K, Nishimori Y, Ueda T, Nakano A, SatoMH, Fukao Y (2014) Interactomics of Qa-SNARE in Arabidopsis thali-ana. Plant Cell Physiol 55: 781–789

Gadella TW Jr, van der Krogt GN, Bisseling T (1999) GFP-based FRETmicroscopy in living plant cells. Trends Plant Sci 4: 287–291

Galperin E, Verkhusha VV, Sorkin A (2004) Three-chromophore FRETmicroscopy to analyze multiprotein interactions in living cells. NatMethods 1: 209–217

Garcia-Molina A, Xing S, Huijser P (2014) A conserved KIN17 curved DNA-binding domain protein assembles with SQUAMOSA PROMOTER-BINDING PROTEIN-LIKE7 to adapt Arabidopsis growth and developmentto limiting copper availability. Plant Physiol 164: 828–840

Geisler M, Murphy AS (2006) The ABC of auxin transport: the role ofp-glycoproteins in plant development. FEBS Lett 580: 1094–1102

Geisler MFE (2011) A predicted interactome for coffee (Coffea canephoravar Robusta). J Plant Mol Biol Biotechnol 2: 34–46

Geisler-Lee J, O’Toole N, Ammar R, Provart NJ, Millar AH, Geisler M (2007)A predicted interactome for Arabidopsis. Plant Physiol 145: 317–329

Ghosh I, Hamilton AD, Regan L (2000) Antiparallel leucine zipper-directed protein reassembly: application to the green fluorescent pro-tein. J Am Chem Soc 122: 5658–5659

Gibson DG, Young L, Chuang RY, Venter JC, Hutchison CA III, SmithHO (2009) Enzymatic assembly of DNA molecules up to several hun-dred kilobases. Nat Methods 6: 343–345

Gietz RD (2006) Yeast two-hybrid system screening. Methods Mol Biol 313:345–371

Gill G, Ptashne M (1988) Negative effect of the transcriptional activatorGAL4. Nature 334: 721–724

Goedhart J, von Stetten D, Noirclerc-Savoye M, Lelimousin M, Joosen L,Hink MA, van Weeren L, Gadella TW Jr, Royant A (2012) Structure-guided evolution of cyan fluorescent proteins towards a quantum yieldof 93%. Nat Commun 3: 751

Gookin TE, Assmann SM (2014) Significant reduction of BiFC non-specificassembly facilitates in planta assessment of heterotrimeric G-proteininteractors. Plant J 80: 553–567

Grefen C (2014) The split-ubiquitin system for the analysis of three-component interactions. Methods Mol Biol 1062: 659–678

Grefen C, Blatt MR (2012a) A 2in1 cloning system enables ratiometric bi-molecular fluorescence complementation (rBiFC). Biotechniques 53: 311–314

Grefen C, Blatt MR (2012b) Do calcineurin B-like proteins interact inde-pendently of the serine threonine kinase CIPK23 with the K+ channelAKT1? Lessons learned from a ménage à trois. Plant Physiol 159: 915–919





Grefen C, Chen Z, Honsbein A, Donald N, Hills A, Blatt MR (2010a) Anovel motif essential for SNARE interaction with the K+ channel KC1and channel gating in Arabidopsis. Plant Cell 22: 3076–3092

Grefen C, Donald N, Hashimoto K, Kudla J, Schumacher K, Blatt MR(2010b) A ubiquitin-10 promoter-based vector set for fluorescent pro-teinbtagging facilitates temporal stability and native protein distributionin transient and stable expression studies. Plant J 64: 355–365

Grefen C, Karnik R, Larson E, Lefoulon C, Wang YZ, Waghmare S, Zhang B,Hills A, Blatt MR (2015) A vesicle-trafficking protein commandeers Kvchannel voltage sensors for voltage-dependent secretion. Nat Plants 1:15108

Grefen C, Lalonde S, Obrdlik P (2007) Split-ubiquitin system for identi-fying protein-protein interactions in membrane and full-length proteins.Curr Protoc Neurosci Chapter 5: Unit 5.27

Grefen C, Obrdlik P, Harter K (2009) The determination of protein-proteininteractions by the mating-based split-ubiquitin system (mbSUS). MethodsMol Biol 479: 217–233

Grefen C, Städele K, R�uzicka K, Obrdlik P, Harter K, Horák J (2008)Subcellular localization and in vivo interactions of the Arabidopsisthaliana ethylene receptor family members. Mol Plant 1: 308–320

Gstaiger M, Knoepfel L, Georgiev O, Schaffner W, Hovens CM (1995) AB-cell coactivator of octamer-binding transcription factors. Nature 373:360–362

Gu H, Zhu P, Jiao Y, Meng Y, Chen M (2011) PRIN: a predicted rice in-teractome network. BMC Bioinformatics 12: 161

Gust A, Zander A, Gietl A, Holzmeister P, Schulz S, Lalkens B, TinnefeldP, Grohmann D (2014) A starting point for fluorescence-based single-molecule measurements in biomolecular research. Molecules 19: 15824–15865

Gutierrez R, Grossmann G, Frommer WB, Ehrhardt DW (2010) Oppor-tunities to explore plant membrane organization with super-resolutionmicroscopy. Plant Physiol 154: 463–466

Hachez C, Laloux T, Reinhardt H, Cavez D, Degand H, Grefen C, DeRycke R, Inzé D, Blatt MR, Russinova E, et al (2014) ArabidopsisSNAREs SYP61 and SYP121 coordinate the trafficking of plasma mem-brane aquaporin PIP2;7 to modulate the cell membrane water permea-bility. Plant Cell 26: 3132–3147

Haddock SH, Moline MA, Case JF (2010) Bioluminescence in the sea. AnnuRev Mar Sci 2: 443–493

Hall RA (2005) Co-immunoprecipitation as a strategy to evaluate receptor-receptor or receptor-protein interactions. In SR George, BF Dowd, eds, GProtein-Coupled Receptor-Protein Interactions. John Wiley & Sons, NewYork, pp 165–178

Hamers-Casterman C, Atarhouch T, Muyldermans S, Robinson G,Hamers C, Songa EB, Bendahman N, Hamers R (1993) Naturally oc-curring antibodies devoid of light chains. Nature 363: 446–448

Harmsen MM, De Haard HJ (2007) Properties, production, and applica-tions of camelid single-domain antibody fragments. Appl MicrobiolBiotechnol 77: 13–22

Hart GT, Ramani AK, Marcotte EM (2006) How complete are current yeastand human protein-interaction networks? Genome Biol 7: 120

Harter K, Meixner AJ, Schleifenbaum F (2012) Spectro-microscopy ofliving plant cells. Mol Plant 5: 14–26

Hecker A, Wallmeroth N, Peter S, Blatt MR, Harter K, Grefen C (2015)Binary 2in1 vectors improve in planta (co)localization and dynamicprotein interaction studies. Plant Physiol 168: 776–787

Heim R, Prasher DC, Tsien RY (1994) Wavelength mutations and post-translational autoxidation of green fluorescent protein. Proc Natl AcadSci USA 91: 12501–12504

Heim R, Tsien RY (1996) Engineering green fluorescent protein for im-proved brightness, longer wavelengths and fluorescence resonance en-ergy transfer. Curr Biol 6: 178–182

Ho CL, Wu Y, Shen HB, Provart NJ, Geisler M (2012) A predicted proteininteractome for rice. Rice (N Y) 5: 15

Honsbein A, Sokolovski S, Grefen C, Campanoni P, Pratelli R, PanequeM, Chen Z, Johansson I, Blatt MR (2009) A tripartite SNARE-K+

channel complex mediates in channel-dependent K+ nutrition in Arabi-dopsis. Plant Cell 21: 2859–2877

Horák J, Grefen C, Berendzen KW, Hahn A, Stierhof YD, StadelhoferB, Stahl M, Koncz C, Harter K (2008) The Arabidopsis thaliana re-sponse regulator ARR22 is a putative AHP phospho-histidine phos-phatase expressed in the chalaza of developing seeds. BMC Plant Biol8: 77

Horstman A, Tonaco IA, Boutilier K, Immink RG (2014) A cautionary noteon the use of split-YFP/BiFC in plant protein-protein interaction studies.Int J Mol Sci 15: 9628–9643

Hu CD, Chinenov Y, Kerppola TK (2002) Visualization of interactionsamong bZIP and Rel family proteins in living cells using bimolecularfluorescence complementation. Mol Cell 9: 789–798

Hurst CH, Hemsley PA (2015) Current perspective on protein S-acylationin plants: more than just a fatty anchor? J Exp Bot 66: 1599–1606

Ishikawa-Ankerhold HC, Ankerhold R, Drummen GP (2012) Advancedfluorescence microscopy techniques: FRAP, FLIP, FLAP, FRET andFLIM. Molecules 17: 4047–4132

Ito T, Chiba T, Ozawa R, Yoshida M, Hattori M, Sakaki Y (2001) Acomprehensive two-hybrid analysis to explore the yeast protein inter-actome. Proc Natl Acad Sci USA 98: 4569–4574

Jain A, Liu R, Ramani B, Arauz E, Ishitsuka Y, Ragunathan K, Park J,Chen J, Xiang YK, Ha T (2011) Probing cellular protein complexes usingsingle-molecule pull-down. Nature 473: 484–488

James P (2001) Yeast two-hybrid vectors and strains. Methods Mol Biol 177:41–84

James P, Halladay J, Craig EA (1996) Genomic libraries and a host straindesigned for highly efficient two-hybrid selection in yeast. Genetics 144:1425–1436

Jiang J, Zhang C, Wang X (2013) Ligand perception, activation, and earlysignaling of plant steroid receptor brassinosteroid insensitive 1. J IntegrPlant Biol 55: 1198–1211

Johnsson N, Varshavsky A (1994) Split ubiquitin as a sensor of proteininteractions in vivo. Proc Natl Acad Sci USA 91: 10340–10344

Johnston M (1987) A model fungal gene regulatory mechanism: the GALgenes of Saccharomyces cerevisiae. Microbiol Rev 51: 458–476

Jones AM, Danielson JAH, Manojkumar SN, Lanquar V, Grossmann G,Frommer WB (2014a) Abscisic acid dynamics in roots detected withgenetically encoded FRET sensors. eLife 3: e01741

Jones AM, Xuan Y, Xu M, Wang RS, Ho CH, Lalonde S, You CH, SardiMI, Parsa SA, Smith-Valle E, et al (2014b) Border control: a membrane-linked interactome of Arabidopsis. Science 344: 711–716

Joosen L, Hink MA, Gadella TW Jr, Goedhart J (2014) Effect of fixationprocedures on the fluorescence lifetimes of Aequorea victoria derivedfluorescent proteins. J Microsc 256: 166–176

Joska TM, Mashruwala A, Boyd JM, Belden WJ (2014) A universal cloningmethod based on yeast homologous recombination that is simple, effi-cient, and versatile. J Microbiol Methods 100: 46–51

Jourdan M, Searle MS (2000) Cooperative assembly of a nativelike ubiq-uitin structure through peptide fragment complexation: energetics ofpeptide association and folding. Biochemistry 39: 12355–12364

Kabachinski G, Schwartz TU (2015) The nuclear pore complex: structureand function at a glance. J Cell Sci 128: 423–429

Kaboord B, Perr M (2008) Isolation of proteins and protein complexes byimmunoprecipitation. Methods Mol Biol 424: 349–364

Karnik A, Karnik R, Grefen C (2013a) SDM-Assist software to design site-directed mutagenesis primers introducing “silent” restriction sites. BMCBioinformatics 14: 105

Karnik R, Grefen C, Bayne R, Honsbein A, Köhler T, KioumourtzoglouD, Williams M, Bryant NJ, Blatt MR (2013b) Arabidopsis Sec1/Munc18 protein SEC11 is a competitive and dynamic modulator ofSNARE binding and SYP121-dependent vesicle traffic. Plant Cell 25:1368–1382

Karnik R, Zhang B, Waghmare S, Aderhold C, Grefen C, Blatt MR (2015)Binding of SEC11 indicates its role in SNARE recycling after vesiclefusion and identifies two pathways for vesicular traffic to the plasmamembrane. Plant Cell 27: 675–694

Kastritis PL, Bonvin AM (2013) On the binding affinity of macromolecularinteractions: daring to ask why proteins interact. J R Soc Interface 10:20120835

Kastritis PL, Moal IH, Hwang H, Weng Z, Bates PA, Bonvin AM, Janin J(2011) A structure-based benchmark for protein-protein binding affinity.Protein Sci 20: 482–491

Keskin O, Gursoy A, Ma B, Nussinov R (2008) Principles of protein-protein interactions: what are the preferred ways for proteins to inter-act? Chem Rev 108: 1225–1244

Kiemer L, Cesareni G (2007) Comparative interactomics: comparing applesand pears? Trends Biotechnol 25: 448–454

Kim SB, Otani Y, Umezawa Y, Tao H (2007) Bioluminescent indi-cator for determining protein-protein interactions using intramolecular


Xing et al.



complementation of split click beetle luciferase. Anal Chem 79: 4820–4826

Kim YG, Park B, Ahn JO, Jung JK, Lee HW, Lee EG (2012) New cell linedevelopment for antibody-producing Chinese hamster ovary cells usingsplit green fluorescent protein. BMC Biotechnol 12: 24

Kittanakom S, Chuk M, Wong V, Snyder J, Edmonds D, Lydakis A,Zhang Z, Auerbach D, Stagljar I (2009) Analysis of membrane proteincomplexes using the split-ubiquitin membrane yeast two-hybrid(MYTH) system. Methods Mol Biol 548: 247–271

Klopffleisch K, Phan N, Augustin K, Bayne RS, Booker KS, Botella JR,Carpita NC, Carr T, Chen JG, Cooke TR, et al (2011) Arabidopsis G-protein interactome reveals connections to cell wall carbohydrates andmorphogenesis. Mol Syst Biol 7: 532

Kodama Y, Hu CD (2012) Bimolecular fluorescence complementation(BiFC): a 5-year update and future perspectives. Biotechniques 53:285–298

Komis G, Šamajová O, Ove�cka M, Šamaj J (2015) Super-resolution mi-croscopy in plant cell imaging. Trends Plant Sci 20: 834–843

Krebs M, Held K, Binder A, Hashimoto K, Den Herder G, Parniske M,Kudla J, Schumacher K (2012) FRET-based genetically encoded sensorsallow high-resolution live cell imaging of Ca2⁺ dynamics. Plant J 69: 181–192

Kremers GJ, Goedhart J (2009) Visible fluorescent proteins for FRET. LabTech Biochem Mol Biol 33: 171–223

Kremers GJ, Goedhart J, van Munster EB, Gadella TW Jr (2006) Cyan andyellow super fluorescent proteins with improved brightness, proteinfolding, and FRET Förster radius. Biochemistry 45: 6570–6580

Kriechbaumer V, Botchway SW, Slade SE, Knox K, Frigerio L, Oparka K,Hawes C (2015) Reticulomics: protein-protein interaction studies withtwo plasmodesmata-localized reticulon family proteins identify bindingpartners enriched at plasmodesmata, endoplasmic reticulum, and theplasma membrane. Plant Physiol 169: 1933–1945

Kumar MN, Hsieh YF, Verslues PE (2015) At14a-Like1 participates inmembrane-associated mechanisms promoting growth during drought inArabidopsis thaliana. Proc Natl Acad Sci USA 112: 10545–10550

Kwaaitaal M, Keinath NF, Pajonk S, Biskup C, Panstruga R (2010)Combined bimolecular fluorescence complementation and Forster res-onance energy transfer reveals ternary SNARE complex formation inliving plant cells. Plant Physiol 152: 1135–1147

Lakowicz JR (2006) Plasmonics in biology and plasmon-controlled fluo-rescence. Plasmonics 1: 5–33

Lalonde S, Ehrhardt DW, Loqué D, Chen J, Rhee SY, Frommer WB (2008)Molecular and cellular approaches for the detection of protein-proteininteractions: latest techniques and current limitations. Plant J 53: 610–635

Lalonde S, Sero A, Pratelli R, Pilot G, Chen J, Sardi MI, Parsa SA, KimDY, Acharya BR, Stein EV, et al (2010) A membrane protein/signalingprotein interaction network for Arabidopsis version AMPv2. FrontPhysiol 1: 24

Lampropoulos A, Sutikovic Z, Wenzl C, Maegele I, Lohmann JU, Forner J(2013) GreenGate: a novel, versatile, and efficient cloning system forplant transgenesis. PLoS ONE 8: e83043

Lanquar V, Grossmann G, Vinkenborg JL, Merkx M, Thomine S, From-mer WB (2014) Dynamic imaging of cytosolic zinc in Arabidopsis rootscombining FRET sensors and RootChip technology. New Phytol 202:198–208

Laptenok SP, Snellenburg JJ, Bücherl CA, Konrad KR, Borst JW (2014)Global analysis of FRET-FLIM data in live plant cells. Methods Mol Biol1076: 481–502

Le CT, Brumbarova T, Ivanov R, Stoof C, Weber E, Mohrbacher J, Fink-Straube C, Bauer P (2016) ZINC FINGER OF ARABIDOPSIS THALIANA12(ZAT12) interacts with FER-LIKE IRON DEFICIENCY-INDUCEDTRANSCRIPTION FACTOR (FIT) linking iron deficiency and oxida-tive stress responses. Plant Physiol 170: 540–557

Leanna CA, Hannink M (1996) The reverse two-hybrid system: a geneticscheme for selection against specific protein/protein interactions. Nu-cleic Acids Res 24: 3341–3347

Lee HW, Kyung T, Yoo J, Kim T, ChungC, Ryu JY, Lee H, Park K, Lee S, JonesWD, et al (2013a) Real-time single-molecule co-immunoprecipitation analysesreveal cancer-specific Ras signalling dynamics. Nat Commun 4: 1505

Lee OH, Kim H, He Q, Baek HJ, Yang D, Chen LY, Liang J, Chae HK,Safari A, Liu D, et al (2011) Genome-wide YFP fluorescence comple-mentation screen identifies new regulators for telomere signaling inhuman cells. Mol Cell Proteomics 10: M110.001628

Lee T, Yang S, Kim E, Ko Y, Hwang S, Shin J, Shim JE, Shim H, Kim H,Kim C, et al (2015) AraNet v2: an improved database of co-functionalgene networks for the study of Arabidopsis thaliana and 27 other non-model plant species. Nucleic Acids Res 43: D996–D1002

Leseberg CH, Eissler CL, Wang X, Johns MA, Duvall MR, Mao L (2008)Interaction study of MADS-domain proteins in tomato. J Exp Bot 59:2253–2265

Li JF, Bush J, Xiong Y, Li L, McCormack M (2011) Large-scale protein-protein interaction analysis in Arabidopsis mesophyll protoplasts bysplit firefly luciferase complementation. PLoS ONE 6: e27364

Li M, Doll J, Weckermann K, Oecking C, Berendzen KW, Schöffl F (2010)Detection of in vivo interactions between Arabidopsis class A-HSFs,using a novel BiFC fragment, and identification of novel class B-HSFinteracting proteins. Eur J Cell Biol 89: 126–132

Lin M, Shen X, Chen X (2011) PAIR: the predicted Arabidopsis interactomeresource. Nucleic Acids Res 39: D1134–D1140

Lin ZJD, Liebrand TWH, Yadeta KA, Coaker G (2015) PBL13 is a serine/threonine protein kinase that negatively regulates Arabidopsis immuneresponses. Plant Physiol 169: 2950–2962

Lipka E, Gadeyne A, Stöckle D, Zimmermann S, De Jaeger G, EhrhardtDW, Kirik V, Van Damme D, Müller S (2014) The phragmoplast-orienting kinesin-12 class proteins translate the positional informationof the preprophase band to establish the cortical division zone in Ara-bidopsis thaliana. Plant Cell 26: 2617–2632

Llorca CM, Berendzen KW, Malik WA, Mahn S, Piepho HP, Zentgraf U(2015) The elucidation of the interactome of 16 Arabidopsis bZIP fac-tors reveals three independent functional networks. PLoS ONE 10:e0139884

Ludewig U, Wilken S, Wu B, Jost W, Obrdlik P, El Bakkoury M, MariniAM, André B, Hamacher T, Boles E, et al (2003) Homo- and hetero-oligomerization of ammonium transporter-1 NH4 uniporters. J BiolChem 278: 45603–45610

Lumba S, Toh S, Handfield LF, Swan M, Liu R, Youn JY, Cutler SR,Subramaniam R, Provart N, Moses A, et al (2014) A mesoscale abscisicacid hormone interactome reveals a dynamic signaling landscape inArabidopsis. Dev Cell 29: 360–372

Lund CH, Bromley JR, Stenbæk A, Rasmussen RE, Scheller HV, SakuragiY (2015) A reversible Renilla luciferase protein complementation assayfor rapid identification of protein-protein interactions reveals the exis-tence of an interaction network involved in xyloglucan biosynthesis inthe plant Golgi apparatus. J Exp Bot 66: 85–97

Macháň R, Wohland T (2014) Recent applications of fluorescence correla-tion spectroscopy in live systems. FEBS Lett 588: 3571–3584

Magdeldin S, Moser A (2012) Affinity Chromatography: Principles andApplications. INTECH Open Access Publisher

Magliery TJ, Wilson CG, Pan W, Mishler D, Ghosh I, Hamilton AD,Regan L (2005) Detecting protein-protein interactions with a green flu-orescent protein fragment reassembly trap: scope and mechanism. J AmChem Soc 127: 146–157

Marques SM, Esteves da Silva JC (2009) Firefly bioluminescence: amechanistic approach of luciferase catalyzed reactions. IUBMB Life 61:6–17

Marsh JA, Teichmann SA (2015) Structure, dynamics, assembly, and ev-olution of protein complexes. Annu Rev Biochem 84: 551–575

Maruta N, Trusov Y, Botella JR (2016) Yeast three-hybrid system forthe detection of protein-protein interactions. Methods Mol Biol 1363:145–154

Más P, Devlin PF, Panda S, Kay SA (2000) Functional interaction of phy-tochrome B and cryptochrome 2. Nature 408: 207–211

Masters SC (2004) Co-immunoprecipitation from transfected cells.Methods Mol Biol 261: 337–350

Matthews JC, Hori K, Cormier MJ (1977) Substrate and substrate analoguebinding properties of Renilla luciferase. Biochemistry 16: 5217–5220

Michard E, Dias P, Feijo JA (2008) Tobacco pollen tubes as cellularmodels for ion dynamics: improved spatial and temporal resolutionof extracellular flux and free cytosolic concentration of calcium andprotons using pHluorin and YC3.1 CaMeleon. Sex Plant Reprod 21:169–181

Michnick SW, Ear PH, Manderson EN, Remy I, Stefan E (2007) Universalstrategies in research and drug discovery based on protein-fragmentcomplementation assays. Nat Rev Drug Discov 6: 569–582

Miernyk JA, Thelen JJ (2008) Biochemical approaches for discoveringprotein-protein interactions. Plant J 53: 597–609





Millar AJ, Short SR, Chua NH, Kay SA (1992) A novel circadian pheno-type based on firefly luciferase expression in transgenic plants. PlantCell 4: 1075–1087

Miller JP, Lo RS, Ben-Hur A, Desmarais C, Stagljar I, Noble WS, Fields S(2005) Large-scale identification of yeast integral membrane protein in-teractions. Proc Natl Acad Sci USA 102: 12123–12128

Miller KE, Kim Y, Huh WK, Park HO (2015) Bimolecular fluorescencecomplementation (BiFC) analysis: advances and recent applications forgenome-wide interaction studies. J Mol Biol 427: 2039–2055

Mira-Rodado V, Veerabagu M, Witthöft J, Teply J, Harter K, Desikan R(2012) Identification of two-component system elements downstream ofAHK5 in the stomatal closure response of Arabidopsis thaliana. PlantSignal Behav 7: 1467–1476

Möckli N, Auerbach D (2004) Quantitative beta-galactosidase assay suit-able for high-throughput applications in the yeast two-hybrid system.Biotechniques 36: 872–876

Möckli N, Deplazes A, Hassa PO, Zhang Z, Peter M, Hottiger MO,Stagljar I, Auerbach D (2007) Yeast split-ubiquitin-based cytosolicscreening system to detect interactions between transcriptionally activeproteins. Biotechniques 42: 725–730

Monti M, Orrù S, Pagnozzi D, Pucci P (2005) Interaction proteomics. BiosciRep 25: 45–56

Morell M, Czihal P, Hoffmann R, Otvos L, Aviles FX, Ventura S (2008)Monitoring the interference of protein-protein interactions in vivo bybimolecular fluorescence complementation: the DnaK case. Proteomics8: 3433–3342

Morell M, Espargaró A, Avilés FX, Ventura S (2007) Detection of transientprotein-protein interactions by bimolecular fluorescence complementa-tion: the Abl-SH3 case. Proteomics 7: 1023–1036

Moreno D, Neller J, Kestler HA, Kraus J, Dünkler A, Johnsson N (2013) Afluorescent reporter for mapping cellular protein-protein interactions intime and space. Mol Syst Biol 9: 647

Morise H, Shimomura O, Johnson FH, Winant J (1974) Intermolecularenergy transfer in the bioluminescent system of Aequorea. Biochemistry13: 2656–2662

Morsy M, Gouthu S, Orchard S, Thorneycroft D, Harper JF, Mittler R,Cushman JC (2008) Charting plant interactomes: possibilities andchallenges. Trends Plant Sci 13: 183–191

Mukherjee P, Banerjee S, Wheeler A, Ratliff LA, Irigoyen S, Garcia LR,Lockless SW, Versaw WK (2015) Live imaging of inorganic phosphate inplants with cellular and subcellular resolution. Plant Physiol 167: 628–638

Mukhtar MS, Carvunis AR, Dreze M, Epple P, Steinbrenner J, Moore J,Tasan M, Galli M, Hao T, Nishimura MT, et al (2011) Independentlyevolved virulence effectors converge onto hubs in a plant immune sys-tem network. Science 333: 596–601

Müller J, Johnsson N (2008) Split-ubiquitin and the split-protein sensors:chessman for the endgame. ChemBioChem 9: 2029–2038

Müller SM, Galliardt H, Schneider J, Barisas BG, Seidel T (2013) Quan-tification of Förster resonance energy transfer by monitoring sensitizedemission in living plant cells. Front Plant Sci 4: 413

Musungu B, Bhatnagar D, Brown RL, Fakhoury AM, Geisler M (2015) Apredicted protein interactome identifies conserved global networks anddisease resistance subnetworks in maize. Front Genet 6: 201

Muyldermans S (2013) Nanobodies: natural single-domain antibodies.Annu Rev Biochem 82: 775–797

Nooren IM, Thornton JM (2003) Diversity of protein-protein interactions.EMBO J 22: 3486–3492

Obrdlik P, El-Bakkoury M, Hamacher T, Cappellaro C, Vilarino C,Fleischer C, Ellerbrok H, Kamuzinzi R, Ledent V, Blaudez D, et al(2004) K+ channel interactions detected by a genetic system optimizedfor systematic studies of membrane protein interactions. Proc Natl AcadSci USA 101: 12242–12247

Ohashi K, Kiuchi T, Shoji K, Sampei K, Mizuno K (2012) Visualization ofcofilin-actin and Ras-Raf interactions by bimolecular fluorescence com-plementation assays using a new pair of split Venus fragments. Bio-techniques 52: 45–50

Ohmuro-Matsuyama Y, Chung CI, Ueda H (2013) Demonstration ofprotein-fragment complementation assay using purified firefly lucifer-ase fragments. BMC Biotechnol 13: 31

Ow DW, Jacobs JD, Howell SH (1987) Functional regions of the cauli-flower mosaic virus 35S RNA promoter determined by use of the fireflyluciferase gene as a reporter of promoter activity. Proc Natl Acad SciUSA 84: 4870–4874

Ozalp C, Szczesna-Skorupa E, Kemper B (2005) Bimolecular fluorescencecomplementation analysis of cytochrome p450 2c2, 2e1, and NADPH-cytochrome p450 reductase molecular interactions in living cells. DrugMetab Dispos 33: 1382–1390

Ozawa T, Kaihara A, Sato M, Tachihara K, Umezawa Y (2001) Splitluciferase as an optical probe for detecting protein-protein interactionsin mammalian cells based on protein splicing. Anal Chem 73: 2516–2521

Parchaliuk DL, Kirkpatrick D, Simon R, Agatep SL, Daniel R, Gietz R(1999) Yeast two-hybrid system. Part C. Characterizing positives. TechTips Online 4: 35–44

Pechmann S, Chartron JW, Frydman J (2014) Local slowdown of transla-tion by nonoptimal codons promotes nascent-chain recognition by SRPin vivo. Nat Struct Mol Biol 21: 1100–1105

Perkins JR, Diboun I, Dessailly BH, Lees JG, Orengo C (2010) Transientprotein-protein interactions: structural, functional, and network prop-erties. Structure 18: 1233–1243

Perrella G, Carr C, Asensi-Fabado MA, Donald NA, Páldi K, Hannah MA,Amtmann A (2016) The Histone Deacetylase Complex (HDC) 1 proteinof Arabidopsis thaliana has the capacity to interact with multiple pro-teins including histone 3-binding proteins and histone 1 variants. PlantPhysiol 171: 62–70

Pertl-Obermeyer H, Schulze WX, Obermeyer G (2014) In vivo cross-linking combined with mass spectrometry analysis reveals receptor-like kinases and Ca2+ signalling proteins as putative interactionpartners of pollen plasma membrane H+ ATPases. J Proteomics 108:17–29

Peter S, Harter K, Schleifenbaum F (2014) Fluorescence microscopy.Methods Mol Biol 1062: 429–452

Pichler A, Prior JL, Piwnica-Worms D (2004) Imaging reversal of multidrugresistance in living mice with bioluminescence: MDR1 P-glycoproteintransports coelenterazine. Proc Natl Acad Sci USA 101: 1702–1707

Piehler J (2005) New methodologies for measuring protein interactions invivo and in vitro. Curr Opin Struct Biol 15: 4–14

Piya S, Shrestha SK, Binder B, Stewart CN Jr, Hewezi T (2014) Protein-protein interaction and gene co-expression maps of ARFs and Aux/IAAs in Arabidopsis. Front Plant Sci 5: 744

Popescu SC, Popescu GV, Bachan S, Zhang Z, Gerstein M, Snyder M,Dinesh-Kumar SP (2009) MAPK target networks in Arabidopsisthaliana revealed using functional protein microarrays. Genes Dev 23:80–92

Popescu SC, Popescu GV, Bachan S, Zhang Z, Seay M, Gerstein M,Snyder M, Dinesh-Kumar SP (2007) Differential binding of calmodulin-related proteins to their targets revealed through high-density Arabi-dopsis protein microarrays. Proc Natl Acad Sci USA 104: 4730–4735

Prasher D, McCann RO, Cormier MJ (1985) Cloning and expression of thecDNA coding for aequorin, a bioluminescent calcium-binding protein.Biochem Biophys Res Commun 126: 1259–1268

Prasher DC, Eckenrode VK, Ward WW, Prendergast FG, Cormier MJ(1992) Primary structure of the Aequorea victoria green-fluorescentprotein. Gene 111: 229–233

Qoronfleh MW, Ren L, Emery D, Perr M, Kaboord B (2003) Use of im-munomatrix methods to improve protein-protein interaction detection.J Biomed Biotechnol 2003: 291–298

Rahim G, Bischof S, Kessler F, Agne B (2009) In vivo interaction betweenatToc33 and atToc159 GTP-binding domains demonstrated in a plantsplit-ubiquitin system. J Exp Bot 60: 257–267

Raquet X, Eckert JH, Müller S, Johnsson N (2001) Detection of alteredprotein conformations in living cells. J Mol Biol 305: 927–938

Reichmann D, Rahat O, Cohen M, Neuvirth H, Schreiber G (2007) Themolecular architecture of protein-protein binding sites. Curr Opin StructBiol 17: 67–76

Remy I, Michnick SW (2006) A highly sensitive protein-protein interactionassay based on Gaussia luciferase. Nat Methods 3: 977–979

Rice LM, Brennwald P, Brünger AT (1997) Formation of a yeast SNAREcomplex is accompanied by significant structural changes. FEBS Lett415: 49–55

Ruohonen L, Aalto MK, Keränen S (1995) Modifications to the ADH1promoter of Saccharomyces cerevisiae for efficient production of heter-ologous proteins. J Biotechnol 39: 193–203

Sahu SS, Weirick T, Kaundal R (2014) Predicting genome-scale Arabidopsis-Pseudomonas syringae interactome using domain and interolog-basedapproaches. BMC Bioinformatics (Suppl 11) 15: S13


Xing et al.



Sanchez C, Lachaize C, Janody F, Bellon B, Röder L, Euzenat J, RechenmannF, Jacq B (1999) Grasping at molecular interactions and genetic networks inDrosophila melanogaster using FlyNets, an Internet database. Nucleic AcidsRes 27: 89–94

Sangsoda S, Cherest H, Surdin-Kerjan Y (1985) The expression of theMET25 gene of Saccharomyces cerevisiae is regulated transcriptionally.Mol Gen Genet 200: 407–414

Sapkota A, Liu X, Zhao XM, Cao Y, Liu J, Liu ZP, Chen L (2011) DIPOS:Database of Interacting Proteins in Oryza sativa. Mol Biosyst 7: 2615–2621

Sasaki Y, Sone T, Yoshida S, Yahata K, Hotta J, Chesnut JD, Honda T,Imamoto F (2004) Evidence for high specificity and efficiency of multiplerecombination signals in mixed DNA cloning by the Multisite Gatewaysystem. J Biotechnol 107: 233–243

Schoonheim PJ, Veiga H, Pereira DdaC, Friso G, van Wijk KJ, de BoerAH (2007) A comprehensive analysis of the 14-3-3 interactome in barleyleaves using a complementary proteomics and two-hybrid approach.Plant Physiol 143: 670–683

Schuette S, Piatkowski B, Corley A, Lang D, Geisler M (2015) Predictedprotein-protein interactions in the moss Physcomitrella patens: a newbioinformatic resource. BMC Bioinformatics 16: 89

Schumacher K, Krebs M (2010) The V-ATPase: small cargo, large effects.Curr Opin Plant Biol 13: 724–730

Schwarzländer M, Wagner S, Ermakova YG, Belousov VV, Radi R,Beckman JS, Buettner GR, Demaurex N, Duchen MR, Forman HJ, et al(2014) The ‘mitoflash’ probe cpYFP does not respond to superoxide.Nature 514: E12–E14

Seidel T, Seefeldt B, Sauer M, Dietz KJ (2010) In vivo analysis of the 2-Cysperoxiredoxin oligomeric state by two-step FRET. J Biotechnol 149: 272–279

Seliger HH, McElroy WD (1964) The colors of firefly bioluminescence:enzyme configuration and species specificity. Proc Natl Acad Sci USA52: 75–81

Serebriiskii IG, Golemis EA (2000) Uses of lacZ to study gene function:evaluation of beta-galactosidase assays employed in the yeast two-hybrid system. Anal Biochem 285: 1–15

Shaner NC, Lin MZ, McKeown MR, Steinbach PA, Hazelwood KL,Davidson MW, Tsien RY (2008) Improving the photostability ofbright monomeric orange and red fluorescent proteins. Nat Methods 5:545–551

Shaner NC, Steinbach PA, Tsien RY (2005) A guide to choosing fluores-cent proteins. Nat Methods 2: 905–909

Shaw SL, Ehrhardt DW (2013) Smaller, faster, brighter: advances in opticalimaging of living plant cells. Annu Rev Plant Biol 64: 351–375

Sheerin DJ, Menon C, zur Oven-Krockhaus S, Enderle B, Zhu L, JohnenP, Schleifenbaum F, Stierhof YD, Huq E, Hiltbrunner A (2015) Light-activated phytochrome A and B interact with members of the SPAfamily to promote photomorphogenesis in Arabidopsis by reorganizingthe COP1/SPA complex. Plant Cell 27: 189–201

Shimomura O, Johnson FH, Saiga Y (1962) Extraction, purification andproperties of aequorin, a bioluminescent protein from the luminoushydromedusan, Aequorea. J Cell Comp Physiol 59: 223–239

Siligato R, Wang X, Yadav SR, Lehesranta S, Ma G, Ursache R, Sevilem I,Zhang J, Gorte M, Prasad K, et al (2016) MultiSite Gateway-compatiblecell type-specific gene-inducible system for plants. Plant Physiol 170:627–641

Sowmya G, Breen EJ, Ranganathan S (2015) Linking structural featuresof protein complexes and biological function. Protein Sci 24: 1486–1494

Stagljar I, Korostensky C, Johnsson N, te Heesen S (1998) A geneticsystem based on split-ubiquitin for the analysis of interactions be-tween membrane proteins in vivo. Proc Natl Acad Sci USA 95: 5187–5192

Stefan E, Aquin S, Berger N, Landry CR, Nyfeler B, Bouvier M, MichnickSW (2007) Quantification of dynamic protein complexes using Renillaluciferase fragment complementation applied to protein kinase A ac-tivities in vivo. Proc Natl Acad Sci USA 104: 16916–16921

Stefano G, Renna L, Brandizzi F (2015) BiFC for protein-protein interac-tions and protein topology: discussing an integrative approach for anold technique. Methods Mol Biol 1242: 173–182

Stellberger T, Häuser R, Baiker A, Pothineni VR, Haas J, Uetz P (2010)Improving the yeast two-hybrid system with permutated fusions pro-teins: the Varicella Zoster Virus interactome. Proteome Sci 8: 8

Stynen B, Tournu H, Tavernier J, Van Dijck P (2012) Diversity in geneticin vivo methods for protein-protein interaction studies: from the yeasttwo-hybrid system to the mammalian split-luciferase system. MicrobiolMol Biol Rev 76: 331–382

Sun Y, Wallrabe H, Booker CF, Day RN, Periasamy A (2010) Three-colorspectral FRET microscopy localizes three interacting proteins in livingcells. Biophys J 99: 1274–1283

Sun Y, Wallrabe H, Seo SA, Periasamy A (2011) FRET microscopy in 2010:the legacy of Theodor Förster on the 100th anniversary of his birth.ChemPhysChem 12: 462–474

Suzuki H, Park SH, Okubo K, Kitamura J, Ueguchi-Tanaka M, Iuchi S,Katoh E, Kobayashi M, Yamaguchi I, Matsuoka M, et al (2009) Dif-ferential expression and affinities of Arabidopsis gibberellin receptorscan explain variation in phenotypes of multiple knock-out mutants.Plant J 60: 48–55

Swatek KN, Graham K, Agrawal GK, Thelen JJ (2011) The 14-3-3 isoformschi and epsilon differentially bind client proteins from developingArabidopsis seed. J Proteome Res 10: 4076–4087

Sydor AM, Czymmek KJ, Puchner EM, Mennella V (2015) Super-resolution microscopy: from single molecules to supramolecular as-semblies. Trends Cell Biol 25: 730–748

Szklarczyk D, Jensen LJ (2015) Protein-protein interaction databases.Methods Mol Biol 1278: 39–56

Thorne N, Inglese J, Auld DS (2010) Illuminating insights into firefly lu-ciferase and other bioluminescent reporters used in chemical biology.Chem Biol 17: 646–657

Tirode F, Malaguti C, Romero F, Attar R, Camonis J, Egly JM (1997) Aconditionally expressed third partner stabilizes or prevents the forma-tion of a transcriptional activator in a three-hybrid system. J Biol Chem272: 22995–22999

To TL, Zhang Q, Shu X (2016) Structure-guided design of a reversiblefluorogenic reporter of protein-protein interactions. Protein Sci 25: 748–753

Tramier M, Zahid M, Mevel JC, Masse MJ, Coppey-Moisan M (2006)Sensitivity of CFP/YFP and GFP/mCherry pairs to donor photo-bleaching on FRET determination by fluorescence lifetime imaging mi-croscopy in living cells. Microsc Res Tech 69: 933–939

Uetz P, Giot L, Cagney G, Mansfield TA, Judson RS, Knight JR, LockshonD, Narayan V, Srinivasan M, Pochart P, et al (2000) A comprehensiveanalysis of protein-protein interactions in Saccharomyces cerevisiae.Nature 403: 623–627

Uhlmann T, Boeing S, Lehmbacher M, Meisterernst M (2007) The VP16activation domain establishes an active mediator lacking CDK8 in vivo. JBiol Chem 282: 2163–2173

Uslu VV, Grossmann G (2016) The biosensor toolbox for plant develop-mental biology. Curr Opin Plant Biol 29: 138–147

Van Leene J, Hollunder J, Eeckhout D, Persiau G, Van De Slijke E, StalsH, Van Isterdael G, Verkest A, Neirynck S, Buffel Y, et al (2010)Targeted interactomics reveals a complex core cell cycle machinery inArabidopsis thaliana. Mol Syst Biol 6: 397

Vidal M, Braun P, Chen E, Boeke JD, Harlow E (1996) Genetic charac-terization of a mammalian protein-protein interaction domain by usinga yeast reverse two-hybrid system. Proc Natl Acad Sci USA 93: 10321–10326

von Behrens I, Komatsu M, Zhang Y, Berendzen KW, Niu X, Sakai H,Taramino G, Hochholdinger F (2011) Rootless with undetectable mer-istem 1 encodes a monocot-specific AUX/IAA protein that controlsembryonic seminal and post-embryonic lateral root initiation in maize.Plant J 66: 341–353

von Mering C, Krause R, Snel B, Cornell M, Oliver SG, Fields S, Bork P(2002) Comparative assessment of large-scale data sets of protein-protein interactions. Nature 417: 399–403

Wagner S, Behera S, De Bortoli S, Logan DC, Fuchs P, Carraretto L,Teardo E, Cendron L, Nietzel T, Füßl M, et al (2015) The EF-Hand Ca2+

binding protein MICU choreographs mitochondrial Ca2+ dynamics inArabidopsis. Plant Cell 27: 3190–3212

Walter M, Chaban C, Schütze K, Batistic O, Weckermann K, Näke C,Blazevic D, Grefen C, Schumacher K, Oecking C, et al (2004) Visual-ization of protein interactions in living plant cells using bimolecularfluorescence complementation. Plant J 40: 428–438

Wang L, Carnegie GK (2013) Flow cytometric analysis of bimolecularfluorescence complementation: a high throughput quantitative methodto study protein-protein interaction. J Vis Exp 78: doi/10.3791/50529





Ward JM, Mäser P, Schroeder JI (2009) Plant ion channels: gene families,physiology, and functional genomics analyses. Annu Rev Physiol 71: 59–82

Wehr MC, Rossner MJ (2016) Split protein biosensor assays in molecularpharmacological studies. Drug Discov Today 21: 415–429

Weijers D, Franke-van Dijk M, Vencken RJ, Quint A, Hooykaas P, OffringaR (2001) An Arabidopsis Minute-like phenotype caused by a semi-dominantmutation in a RIBOSOMAL PROTEIN S5 gene. Development 128: 4289–4299

Weinhandl K, Winkler M, Glieder A, Camattari A (2014) Carbon sourcedependent promoters in yeasts. Microb Cell Fact 13: 5

Wittke S, Lewke N, Müller S, Johnsson N (1999) Probing the molecularenvironment of membrane proteins in vivo. Mol Biol Cell 10: 2519–2530

Wolf A, Akrap N, Marg B, Galliardt H, Heiligentag M, Humpert F, SauerM, Kaltschmidt B, Kaltschmidt C, Seidel T (2013) Elements of tran-scriptional machinery are compatible among plants and mammals. PLoSONE 8: e53737

Woo J, Howell MH, von Arnim AG (2008) Structure-function studies onthe active site of the coelenterazine-dependent luciferase from Renilla.Protein Sci 17: 725–735

Yang B, Ruan R, Cantu D, Wang X, Ji W, Ronald PC, Dubcovsky J (2013a)A comparative approach expands the protein-protein interaction nodeof the immune receptor XA21 in wheat and rice. Genome 56: 315–326

Yang F, Moss LG, Phillips GN Jr (1996) The molecular structure of greenfluorescent protein. Nat Biotechnol 14: 1246–1251

Yang J, Osman K, Iqbal M, Stekel DJ, Luo Z, Armstrong SJ, Franklin FC(2013b) Inferring the Brassica rapa interactome using protein-proteininteraction data from Arabidopsis thaliana. Front Plant Sci 3: 297

Yu CH, Dang Y, Zhou Z, Wu C, Zhao F, Sachs MS, Liu Y (2015) Codonusage influences the local rate of translation elongation to regulate co-translational protein folding. Mol Cell 59: 744–754

Yu H, Braun P, Yildirim MA, Lemmens I, Venkatesan K, Sahalie J, Hirozane-Kishikawa T, Gebreab F, Li N, Simonis N, et al (2008) High-quality binaryprotein interaction map of the yeast interactome network. Science 322: 104–110

Zacharias DA, Violin JD, Newton AC, Tsien RY (2002) Partitioning oflipid-modified monomeric GFPs into membrane microdomains of livecells. Science 296: 913–916

Zamyatnin AA Jr, Solovyev AG, Bozhkov PV, Valkonen JP, Morozov SY,Savenkov EI (2006) Assessment of the integral membrane protein to-pology in living cells. Plant J 46: 145–154

Zhang B, Karnik R, Wang Y, Wallmeroth N, Blatt MR, Grefen C (2015)The Arabidopsis R-SNARE VAMP721 interacts with KAT1 and KC1 K+

channels to moderate K+ current at the plasma membrane. Plant Cell 27:1697–1717

Zhang J, Lautar S (1996) A yeast three-hybrid method to clone ternaryprotein complex components. Anal Biochem 242: 68–72

Zhao H, Doyle TC, Wong RJ, Cao Y, Stevenson DK, Piwnica-Worms D,Contag CH (2004) Characterization of coelenterazine analogs for mea-surements of Renilla luciferase activity in live cells and living animals.Mol Imaging 3: 43–54

Zhu G, Wu A, Xu XJ, Xiao PP, Lu L, Liu J, Cao Y, Chen L, Wu J, Zhao XM(2016) PPIM: a protein-protein interaction database for maize. PlantPhysiol 170: 618–626


Xing et al.



full text (pdf) - plant · pdf filetopical review on protein-protein interaction techniques...

Documents