Identification of kinase substrates by bimolecular complementation assays


  • Stefan Pusch,

    1. Unigruppe am Max-Planck-Institut für Züchtungsforschung, Max-Delbrück-Laboratorium, Lehrstuhl für Botanik III, Universität Köln, Carl-von-Linné-Weg 10, D-50829 Köln, Germany
    Search for more papers by this author
    • Both authors contributed equally to this work.

    • Present address: Deutsches Krebsforschungszentrum (DKFZ), Im Neuenheimer Feld 280, D-69120 Heidelberg, Germany.

  • Hirofumi Harashima,

    1. Institut de Biologie Moléculaire des Plantes (IBMP), UPR2357 du CNRS, 12, rue du Général Zimmer, 67084 Strasbourg, France
    Search for more papers by this author
    • Both authors contributed equally to this work.

  • Arp Schnittger

    Corresponding author
    1. Unigruppe am Max-Planck-Institut für Züchtungsforschung, Max-Delbrück-Laboratorium, Lehrstuhl für Botanik III, Universität Köln, Carl-von-Linné-Weg 10, D-50829 Köln, Germany
    2. Institut de Biologie Moléculaire des Plantes (IBMP), UPR2357 du CNRS, 12, rue du Général Zimmer, 67084 Strasbourg, France
    Search for more papers by this author

(fax +33 (0) 3 88 61 44 42; e-mail


As a consequence of the transient nature of kinase–substrate interactions, the detection of kinase targets, although central for understanding many biological processes, has remained challenging. Here we present a straightforward procedure that relies on the comparison of wild type with activation-loop mutants in the kinase of interest by bimolecular complementation assays. As a proof of functionality, we present the identification and in vivo confirmation of substrates of the major cell-cycle kinase in Arabidopsis, revealing a direct link between cell proliferation and the control of the redox state.


Protein kinases (PKs) play important roles in the control of many – if not all – aspects of cellular life, from humans to plants. It has been estimated that 2–4% of eukaryote genomes encode PKs (Manning et al., 2002), which may phosphorylate 30% of the proteome (Ahn and Resing, 2001). Matching this prediction, approximately 3% of all genes (894 genes) of the Arabidopsis genome encode for PKs (TAIR 10 release, (Dissmeyer and Schnittger, 2011a). PKs build a superfamily that shows a similar structural organization, known as the PK fold (Hunter, 1995; Johnson and O’Reilly, 1996). All harbor a conserved catalytic domain, which consists of approximately 200–300 residues, and is typically divided into a small N-terminal lobe, primarily involved in ATP binding, and a large C-terminal lobe that is largely responsible for substrate binding and catalysis, with an additional role in orienting ATP.

An important feature of the C-terminal lobe is the activation loop, which gates access to the ATP and/or substrate binding pockets. In addition, the T-loop participates in substrate recognition and orientation during the actual kinase reaction (Nolen et al., 2004). Modulating the conformation of the activation loop is the most common mechanism for regulating kinase activity. Activation loop movement can be induced through the interaction of the protein kinase with regulatory subunits and/or by phosphorylation of one or more of its residues. An example for both types of regulation can be found in cyclin-dependent kinases (CDKs). Binding of the cyclin partner positions the T-loop of the kinase subunit, as shown by crystallographic analyses of human Cdk2 with cyclin A (Russo et al., 1996). However, this conformational change is not sufficient for high levels of kinase activity, and only upon phosphorylation by another kinase does the T-loop flatten and move closer to the cyclin partner, thereby allowing an effective interaction with kinase substrates. This mechanism of T-loop action is highly conserved and has also been found to be mandatory for the activation of plant Cdks (Dissmeyer et al., 2007; Harashima et al., 2007).

Whereas the function of many kinases is understood in structural depth, it is still a major challenge to identify their substrates. One of the major reasons for our lack of knowledge of substrates is the intrinsically transient ‘kiss-and-run’ nature of the enzyme–substrate interaction. This transient contact typically precludes or at least strongly reduces the efficiency of the detection of substrates with commonly used methods of detecting protein–protein interactions, such as the yeast two-hybrid system (Y2H), and elaborate biochemical or chemical biological techniques are required to identify kinase targets.

One of the many variants of the Y2H system is the split-ubiquitin system (SUS) (Johnsson and Varshavsky, 1994). In SUS, the proteins to be tested for interaction are combined with the N- or C-terminal half of ubiquitin (Nub and Cub), and upon successful interaction, a complete ubiquitin moiety is reconstituted that is coupled to a metabolic and/or visual read out. The bimolecular fluorescence complementation (BiFC) assay – also called split-YFP assay – is based on the same principle as the SUS, i.e. the reconstitution of the reporter system. In the BiFC assay, the reporter is a fluorescent protein that is separated into two halves, with each half fused to the proteins to be tested for interaction (Hu et al., 2002). The BiFC assay was adapted for use in plants by Bracha-Drori et al. (2004) and Walter et al. (2004), and since then has been intensively used to study protein–protein interactions in Arabidopsis and other plant species.

The SUS and the BiFC approaches are known to stabilize interactions, as both Ub and YFP are stable after reconstitution by interacting proteins. Here we have exploited this property and have used BiFC to systematically detect kinase–substrate interactions. Following a SUS screen, we have employed a second selection criterion and re-tested potential substrates with T-loop mutants in our kinase of interest, selecting for proteins that showed a reduction in their interaction strength in comparison with the wild-type kinase version. Subsequent kinase assays with CDK activity purified from plant extracts confirmed that the proteins identified here are indeed phosphorylated by CDKA;1, leading to the discovery of the first CDKA;1 substrates that were found in an experimentally unbiased manner.


CDK substrates can be detected by SUS and BiFC assays

To test whether enzyme–substrate interactions can be faithfully detected by SUS, we analyzed the interaction between CDKA;1, the Arabidopsis homolog of the yeast cdc2/CDC28 kinase, and the Arabidopsis homologs of the two pre-replication factors, CDC6 and CDT1. Both pre-replication factors are well-known targets of Cdk action in yeast and animals, and their phosphorylation by Cdks plays a key role in DNA replication control (Bell and Dutta, 2002). Both factors have also been found to play a key role for DNA replication in plants (Castellano et al., 2001; del Mar Castellano et al., 2004). As a third known Cdk substrate, we chose bovine histone H1, which is often used as a generic substrate in CDK kinase reactions (Dissmeyer and Schnittger, 2011b; Harashima and Sekine, 2011). As positive controls for our interaction assays, we employed well-known CDK binding partners, i.e. CKS1, a B-type cyclin (CYCB1;2), a D-type cyclin (CYCD3;1) and the CDK inhibitor ICK1/KRP1 (van Leene et al., 2011).

In the SUS assay applied here, yeast colonies with interacting proteins grow on 5-fluoroorotic acid (5-FOA)-containing plates, as interaction causes degradation of the orotidine 5-phosphate decarboxylase (URA3), and thus prevents the conversion of 5-FOA to the toxic compound 5-fluorouracil that causes cell death (Johnsson and Varshavsky, 1994). Degradation of URA3 results from the fact that the reconstituted ubiquitin will be recognized and cleaved by endogenous ubiquitinases immediately after the C terminus of Ub, resulting in the removal of the URA3 protein that is fused to the C-terminal half of Ub. Conversely, cells with non-interacting pairs do not grow on FOA plates, and grow instead on plates without supplemented uracil because they accumulate URA3.

No growth was observed when the empty bait vector (pMet) was transformed. In addition, no interaction between CDKA;1 in the bait vector and the negative control (GUS) in the prey vector was observed. In contrast, yeast colonies co-expressing CDKA;1 and all interactors, as well as all three CDK substrates, grew on FOA plates, whereas no or very little growth was observed on synthetic drop out medium (SD) plates without uracil (Figure 1, data not shown).

Figure 1.

 Enzyme–substrate interactions can be revealed by the split-ubiquitin system (SUS). A dilution series of the yeast strains, starting with an OD600 of 1 (approximately 105 cells) were plated onto plates without uracil and without 5-fluoroorotic acid (5-FOA) (SD – Uracil), and on plates supplemented with 1 mg ml−1 5-FOA and uracil (FOA). Row 1: yeast cells do not grow with an empty bait vector (pMet). As yeast cells grow with an empty prey (pNuI) vector in the SUS set-up, this experiment is not conclusive and was not performed. Row 2: negative control, i.e. no interaction between CDKA;1 and the bacterial protein GUS is found. Rows 3–5: positive control, SUS reveals the strong binding of CDKA;1 in the bait vector (pMet) to its interaction partner CKS1 (Suc1 homolog), and the CDK inhibitor ICK1/KRP1, as well as to a lesser degree with CYCLIN D3;1, all in the prey vector (pNuI). Rows 6–8 substrate interactions: positive SUS assay with CDKA;1 as a bait, and the preys CDT1, CDC6 and histone H1 as well-known substrates of Cdk1, Cdc2 and CDC28, the metazoan and yeast homologs of CDKA;1.

Next, we analyzed the interaction between CDKA;1 and the aforementioned interactors CKS, CYCB1;2 and CYCD3;1, as well as substrates CDC6, CDT1 and Histone H1, by BiFC assays. Interaction in the BiFC assay can be visualized and quantified by measuring the epifluorescence of the reconstituted YFP (Pusch et al., 2011). Co-infiltration of tobacco cells (Nicotiana benthamiana) with expression vectors for CDKA;1, as well interactors and substrates, resulted in a strong YFP signal, whereas no fluorescence with CDKA;1 and GUS or with CDKA;1 and an empty vector was observed (Experimental procedures; Figures 2 and S1). Thus, BiFC and SUS showed similar results, demonstrating that both bimolecular complementation assays can faithfully detect kinase substrate interactions.

Figure 2.

 Enzyme–substrate interactions can be revealed by bimolecular fluorescence complementation (BiFC).
(a–c) YFP signal and bright-field overlays of BiFC interactions of CDKA;1 with CKS1 (a), with CDT1 (b) and as negative control with GUS (c).
(d–f) GUS staining of the leaves shown in (a–c), respectively.
(g–l) BiFC signals of CDKA;1 variants wild type (wt) (g and j), T161D (h and k) and T161V (i and l) interacting with histone H1 (g–i, close-ups of nuclei; j–l, overviews of leaves).
(m) Relative fluorescent units (RFU) quantification of BiFC interactions of known Cdk1, Cdc2 and CDC28 substrates (CDT1, CDC6 and Histone H1) with CDKA;1 variants. RFU quantification of BiFC interactions of known interactors (CYCB1;2, CYCD3;1 and CKS1) with CDKA;1 variants. The interaction strength was not altered when using different CDKA;1 versions in which the T-loop was mutated. The CDKA;1–wild type (wt) interaction was set to 100%. (n) In contrast to CDKA;1-binding proteins, known substrates of animal and human Cdks, showed a gradual decline in florescence strength with the different CDK variants, the decline correlated with the previously published level of kinase activity observed in these mutants (Dissmeyer et al., 2007, 2009).

SUS screen and BiFC re-screen

The faithful detection of CDK substrates in bimolecular interaction assays prompted us to perform a SUS screen to search for CDKA;1 substrates (Experimental procedures). Out of the initially 4 × 106 transformants, 144 different proteins could be identified to bind to CDKA;1 in our SUS screen (Figure 3). Notably, the detected proteins could link CDK action to many biological processes, for instance response to wounding, pyrovate metabolism, cellulose biosynthesis and redox regulation (Table S1). Many of these proteins harbored an (S/T)P or (S/T)PX(K/R) motif, the consensus site of the CDKA;1 homolog Cdk1, and CDC28 from animals and yeast. In addition, many proteins also contained an RXL, also called Cy motif, that has often been found to mediate contact between a kinase substrate and the cyclin partner of CDK-cyclin complexes (Table S1) (Morgan, 2007). However, both the Cdk1 consensus site and the Cy motif are short sequences that appear in several thousand proteins in the Arabidopsis proteome, and thus their presence does not unequivocally identify a CDK substrate. Screening the Phosphat 3.0 database ( that lists known phosphoproteins in Arabidopsis (Heazlewood et al., 2007; Durek et al., 2009) revealed that only a few of the proteins identified here are currently known to be phosphoryatled at S or T in vivo: 13 proteins contained a phosphorylated (S/T)P site, of which four contained a phosphorylated (S/T)PX(K/R) site (Table S1). Thus, the existing possibilities in data mining are restricted in their power to pinpoint CDKA;1 substrates among the proteins identified here. Moreover, several proteins without the minimal Cdk1 consensus site were phosphorylated by Cdk1 in budding yeast (Ubersax et al., 2003).

Figure 3.

 Flow chart of the methodology used here to identify kinase substrates.
After a split-ubiquitin system (SUS) screen (main screen and rescreen), the possible candidates were tested by BiFC with different kinase variants. Those with a substrate signature were tested in direct kinase assays, and four proteins could be confirmed as CDKA;1 targets.

To experimentally identify substrates among the SUS interactors of CDKA;1, we implemented a second step in which we re-screened the identified proteins in a BiFC assay using CDKA;1 variants that show compromised substrate binding (Dissmeyer et al., 2007). The allele cdka;1T161D is an Asp substitution of Thr161 in the T-loop of CDKA;1, and was found to result in the reduced substrate binding of CDKA;1; the substitution of Thr161 with the neutral amino acid Val (called cdka;1T161V hereafter) affects substrate binding even more severely (Dissmeyer et al., 2007). As an additional control, besides the non-mutated wild-type kinase, we used substitutions of Thr14 and Tyr15 with Val and Phe (cdka;1T14V;Y15F) that were found to completely rescue cdka;1 mutants and have wild-type levels of kinase activity (Dissmeyer et al., 2009, 2010). The cdka;1T14V;Y15F allele was originally designed to test the importance of negative regulatory phosphorylation of the amino acids T14 and Y15. However, in contrast to animals and fission yeast, this CDKA;1 variant was found to completely rescue cdka;1 mutants, and the corresponding plants resemble wild-type plants.

We reasoned that putative substrates should bind equally well to the wild type and cdka;1T14V;Y15F, whereas their interaction with cdka;1T161D and in particular with cdka;1T161V should be reduced. This interaction pattern was confirmed, consistent with previously published data, by quantifying the fluorescence intensity as well as determining the number of fluorescent cells in BiFC assays when comparing the three binding partners of CDKA;1, i.e. CYCB1;2, CYCD3;1 and CKS, with the three substrates CDC6, CDT1 and histone H1 (Figures 2 and S1) (Dissmeyer et al., 2007, 2009).

Next, we tested a random selection of 25 proteins identified in the SUS screen with the four different kinase versions in BiFC assays (Figure S2; Table S1). For nine proteins no interaction was found (including four that localized to chloroplasts). Eight out of the remaining 16 proteins showed binding as predicted for stable interaction partners, but the other half matched our criteria for substrates. Interestingly, these putative substrates included a set of proteins involved in redox regulation in plants: GLUTHATIONE PEROXIDASE 2 (GPX2), GLUTHATIONE S-TRANSFERASE PHI 2 (GSTF2) and METALLOTHIONIN 2A (MT2A) (Figure 4a).

Figure 4.

 Identification and confirmation of CDKA;1 substrates.
(a) RFU quantification of bimolecular fluorescence complementation (BiFC) interactions. The interaction of the different proteins tested with the wild-type version of CDKA;1 was set to 100%. The four proteins MT2A, GPX2, GSTF2 and HSC70.1 delivered from the SUS screen show a substrate-like interaction profile with CDKA;1 in BiFC assays with different CDKA;1 versions (see Figure S2 and Table S1 for a complete list of BiFC assays).
(b) Kinase assays with p13suc1-associated kinases from Col-0 inflorescences.
(c, d) Kinase assays with purified StrepIII-CDKA;1 from ProCDKA;1::StrepIII-CDKA;1 seedlings. (b, c) Lane 1, GST-His6; lane 2, HisGST-MT2A; lane 3, HisGST-GPX2; lane 4, HisGST-GSTF2. (d) HisGST-HSC70.1. Top row, autoradiograph; bottom row, CBB staining. Purer CDKA;1 complexes could be obtained from StrepIII-CDKA;1 plant extracts than from extractions with p13suc1 beads. Therefore, there is less background in kinase assays with StrepIII-CDKA;1, even though the film was exposed for longer than for kinase assays with p13suc1 beads.

Biochemical confirmation of substrates

To test whether the above-identified proteins are indeed substrates of CDKA;1, we performed kinase assays with the purified proteins from Escherichia coli extracts. Because of the unexpected prediction of a link between CDK activity and redox state, we decided to focus on GPX2, GSTF2 and MT2A, in addition to the heat-shock protein HSC70.1.

Kinase assays performed with p13Suc1-bound CDKs from plant extracts revealed that all three proteins involved in redox regulation and HSC70 are phosphorylated by CDKs (Figure 4b). Next, we wanted to specifically test the ability of CDKA;1 to phosphorylate the above-identified proteins. For that, a construct harboring a Strep-tag III fused CDKA;1 (StrepIII-CDKA;1) variant driven from the previously characterized CDKA;1 promoter was generated and transformed into heterozygous cdka;1 mutants (Nowack et al., 2006) (Experimental procedures). Both the expression of the N-terminal and the C-terminal fusions of Strep-III with CDKA;1 could fully rescue cdka;1 mutants. However, CDKA;1 activity could only be effectively pulled down from plant extracts with the N-terminal fusion (Figure S3, data not shown). While it was not possible to detect whether GSTF2 was phosphorylated above the background level, GPX2, HSC70.1 and MT2A were efficiently phosphorylated by N-terminal Strep-III CDKA;1 activity, purified with Strep-Tactin, a derivative of streptavidin (Figure 4c,d). Conversely, two additional proteins that did not show a substrate BiFC signature (At1g10590 and UBC35) were expressed in E. coli, but could not be phosphorylated in our kinase assays (data not shown).


The identification of kinase substrates is a challenging task, and the overall progress in unraveling kinase targets has been slow. For instance, the first proteome-wide identification of substrates of Cdc28, the key regulator of cell cycle progression in Saccharomyces cerevisiae and homolog of the mammalian Cdk1, was only accomplished in 2003 (Ubersax et al., 2003). One reason for the delay in the identification of kinase targets is the transient nature of enzyme–substrate interactions, which typically prevents or at least strongly reduces the number of substrates identified by conventional protein–protein interaction assays. For instance, Van Leene et al. (2007) identified 20 CDKA;1 binding proteins by tandem affinity purification. Of these, at least seven are most likely to represent no substrate interactors, such as cyclins or CDK inhibitors; whether the other proteins identified are substrates was not assessed (Van Leene et al., 2007). Even in an enlarged tandem affinity purification screen, mostly cyclins and CDK inhibitors were found, and only very few potential substrates were identified; again, whether these proteins are indeed substrates was not tested further (Van Leene et al., 2010). Similarly, no obvious CDK substrates were reported from conventional Y2H screens with CDKA;1 (De Veylder et al., 1997; Wang et al., 1998).

Kinase substrates can now be isolated by biochemical or chemical genetic procedures, but both methods require a careful optimization of reaction conditions (Hodgson and Schroder, 2011). For instance, in a chemical genetic approach the typically large hydrophobic or polar residue in the ATP-binding pocket of the kinase of interest is mutated to a smaller amino acid, such as alanine (A) or glycine (G) (Elphick et al., 2009). The exchange of this ‘gatekeeper amino acid’ increases the ATP-binding pocket so that either enlarged (‘bulky’) ATP analogues or bulky kinase inhibitors can specifically target these gatekeeper mutant kinase versions (Shogren-Knaak et al., 2001; Elphick et al., 2007). This method has also been named the ‘bump-and-hole’ approach, and such engineered kinases have been a very powerful tool to study many biological problems, for instance for the above-mentioned identification of Cdc28 substrates or substrates for human Cdk1 (Ubersax et al., 2003; Blethrow et al., 2008). However, the introduction of a gatekeeper mutation often reduces kinase activity, and secondary mutations that stabilize the mutated kinase, restoring high activity levels, often have to be introduced (Zhang et al., 2005). Moreover, one of the most commonly used approaches relies on the usage of a bulky ATPγS, but the identification of thio-phosphorylated substrates requires exquisite expertise in biochemistry, and thus precludes the general application of this method (Allen et al., 2007).

An alternative and potentially very powerful method to identify enzyme substrates are protein arrays, which have also been successfully employed in plants for the identification of components in mitogen-activated protein (MAP) kinase networks (Popescu et al., 2009). However, protein arrays are currently still cost intensive, and represent only a small fraction of the proteome, for instance in the aforementioned analysis arrays with 2158 unique proteins were used.

Taking the example of CDKA;1, we have shown here that bimolecular complementation assays represent a straightforward method to address enzyme–substrate interactions. In particular, the combination of this screen with typically easy to generate T-loop mutants should allow the identification of substrates for many different kinases. To our knowledge, GPX2, MT2A and HSC70 are the first three bonafide substrates of Cdk1-type kinases that were identified in plants in an unbiased manner, i.e. not by comparison with substrates from other species, as is the case for the plant homolog of Rb, for example (Nakagami et al., 1999; M.K. Nowack, H. Harashima, N. Dissmeyer, X. Zhao, D. Bouyer, A.K. Weimer, F. De Winter, F. Yang and A. Schnittger, unpublished data).

In our screen we have only found a few proteins that were annotated to have a function in cell-cycle regulation, e.g. cyclins (although there is a link between HSC70, as well as some of the other identified proteins, with the cell cycle; see for instance Diehl et al., 2003). In addition, for only 10 genes encoding the proteins identified here, a cell-cycle phase-specific expression pattern was observed (Menges et al., 2005) (Table S1). One likely reason for this finding is that cell-cycle genes were under-represented in our Y2H library, which contained a large proportion of cDNAs made from mature plant tissues. In addition, we observed that the expression of cell-cycle genes, specifically cyclins such as CYCLIN B1;2 or CYCLIN D3;1, interfered with yeast growth, presumably because of the similarity with endogenous regulators (Figure 1, data not shown). In particular, under screening conditions yeast cells with just a moderate growth defect can easily be lost. For instance, we identified a strong interaction between CDKA;1 and CDT1 as well as CDC6 in our direct assay, but both proteins could not be recovered from the interactions found in our screen. However, this set-up has facilitated the detection of CDKA;1 substrates outside of core cell cycle regulation, an area about which little is currently known. In addition, among all of the proteins identified in our SUS screen, very few homologs were found in S. cerevisiae substrates, thereby pinpointing plant-specific regulatory interactions (Table S1) (Ubersax et al., 2003). Here, we could show that at least three proteins that control the redox state of a cell are phosphorylated by CDKA;1. Interestingly, the redox state has been recently found to influence proliferation and growth of Arabidopsis anthers and roots, and our data now indicates a direct regulation of the redox state by the cell cycle, hinting at a feedback control system (Xing et al., 2005; Tsukagoshi et al., 2010).

Experimental Procedures

Gateway-compatible vectors for Split-Ubiquitin (pMet, pNuI, pCKZ and pCup-CGK) were a kind gift from Laurent Deslandes and Imre E. Somssich. pNuA and pCKZ-A were produced by site-directed mutagenesis with primers P29 and P30. The Gateway-compatible BiFC plant expression vectors pSYN and pSYC were constructed as previously published (Jakoby et al., 2006). The open reading frames (ORFs) of CDKA;1, CDC6, CDT1, CKS1, MT2A, GPX2, GSTF2, CYCB1;2 and CYCD3;1, and Histone H1, were introduced without STOP codon into the Split-Ubiquitin assay vectors (pMet and pNuI) and into the BiFC assay vectors (pSYN and pSYC). cDNAs were amplified and manipulated via PCR (Table S2), and cloned into pDONR201 following the BP-reaction protocol (Invitrogen, After sequence verification they were cloned into the different destination vectors by using the LR-reaction protocol (Invitrogen).

Split ubiquitin Y2H screen

The JD53 S. cerevisiae strain was used. For the screen the bait vector pMet containing the CDKA;1 ORF was transformed into JD53. After transformation, yeast cells were streaked out on synthetic drop-out (SD) medium. Positive colonies were tested for expression, transformation efficiency and reliability. The largest colony was chosen to perform the cDNA-library screen. The cDNA library (a kind gift of Filipa Santos, Iris Ottenschläger and Klaus Palme) was generated from: (i) plate-grown seedlings and 4-week-old plants; (ii) soil-grown aerial parts of 4–6-week-old plants; and (iii) suspension culture of Arabidopsis thaliana ecotype Columbia. The cDNA-library transformation was plated onto SD supplemented with 1 mg ml−1 5-FOA and 100 mm copper sulphate. Yeast cells were grown at 30°C for 2 weeks. Positive colonies were harvested into liquid SD-media. Liquid cultures were grown at 30°C for 3 day and re-plated on SD without uracil, supplemented with 100 mm copper sulfate. All colonies that grew on these plates were used for colony PCR.

Split ubiquitin yeast two-hybrid assays

The S. cerevisiae strain used was JD53. The bait vector pMet and the prey vector pNuI were transformed into JD53. After transformation, yeast cells were streaked out on SD medium. A dilution series of the yeast strains, starting with an OD600 of 1 (approximately 105 cells), were plated onto SD plates for growth control, on SD without uracil supplemented with 100 mm copper sulfate, and on SD supplemented with 1 mg ml−1 5-FOA and 100 mm copper sulfate. Yeast cells were grown at 30°C for 3 days.

Colony PCR

Part of the colony was added to 50 μl of 20-mm NaOH and lysed for 5 min. The lysate was centrifuged at 10 000 rpm for 1 min, and 5 μl of the supernatant was used for PCR. The PCR was performed with 1 μl of each primer (SP40 and SP42), 5 μl of 10× Taq-Polymerase Buffer, 1 μl of deoxyribonucleotide triphosphates (dNTPs; 10mm), 1 μl Taq-Polymerase, 36 μl of sterile water. The PCR-protocol was 95°C for 5 min for denaturation, and 30 cycles of 95°C for 30 sec followed by 60°C 45 sec for primer binding, and 72°C for 2 min 30 sec as an elongation time.

Restriction digest

The digest of the PCR product was performed with 17.8 μl of PCR, 2 μl yellow buffer (Fermentas) and 0.2 μl HpaII (Fermentas, The sample was incubated for 1 h at 37°C, and was subsequently loaded onto a 2% agarose gel. PCR products with the same restriction pattern were assumed to be identical, and only one of them was used for further gap repair analysis.

Gap repair transformation

pCup-CGK was digested with EcoRI and XhoI (Fermentas) at 37°C for 3 h. The PCR product was digested with DpnI (Fermentas), so as to exclude template transformation. The digests were stopped by heating to 85°C for 20 min. Both digests were mixed (1 μl pCup-CGK with 5 μl PCR product), and transformed into JD53. After transformation yeast cells were streaked out onto SD for the control of transformation efficiency, and on SD supplemented with 1 mg ml−1 5-FOA and 100 mm copper sulphate for retesting.

BiFC assay

For the infiltration of N. benthamiana leaves, the A. tumefaciens strain GV3101 pMP90RK was used. The Agrobacterium strains containing the BiFC vectors were infiltrated as described by Walter et al. (2004). Infiltration was always performed on the abaxial leaf side of 2-month-old tobacco plants, and analyzed after 3 days via confocal microscopy.


Confocal laser-scanning microscopy was performed with a Leica TCS SP2 AOBS CLSM system, equipped with an argon-krypton laser and a 405-nm diode laser (Leica, Images were processed using Photoshop CS 8.0 and Illustrator CS 11.0 (Adobe,, and ImageJ 1.36 (

BiFC signal quantification

For signal quantification, pictures of at least 10 cells were taken. All pictures of one measurement had the same laser settings and magnification. Signals were then quantified using the measurement tool in ImageJ.

Strep-tag III fused CDKA;1

To purify CDKA;1 protein complexes, plants expressing a Strep-tag III fused CDKA;1 (StrepIII-CDKA;1) were generated. CDKA;1 was amplified with primers 266 and 95, primers 324 and 95, and primers 323 and 95 to conjugate a TEV protease recognition sequence, a Strep-tag III sequence and attB recombination sites, respectively. The PCR product was cloned into the GATEWAY entry vector pDONR223 (Invitrogen) using BP Clonase II (Invitrogen), and then recombined into the destination vector pAMPAT(pCDKA;1) (Nowack et al., 2006) by using LR Clonase II (Invitrogen). Heterozygous cdka;1-1 plants (SALK_109806) were transformed by a modified version of the floral-dip method (Clough and Bent, 1998). Standard lines were established by isolating T3 plants homozygous for the transgene in cdka;1-1/−.

Protein expression and purification

To express HisGST-fused proteins, the cDNA of the respective entry clone was recombined into the destination vector pHGGWA (Busso et al., 2005) by using LR Clonase II (Invitrogen), and E. coli BL21-AI cells (Invitrogen) were transformed with the resulting vector. E. coli cells were grown to OD600 = 1.0, and the production of the fusion protein was induced by adding 0.3 mm isopropyl-β-d-thio-galactoside (IPTG) and 0.2% arabinose for 6 h at 37°C (MT2A, GPX2 and GSTF2), or overnight at 18°C (HSC70.1). Cells were harvested by centrifugation and re-suspended in Ni-NTA binding buffer (50 mm NaH2PO4, 100 mm NaCl, 10% (v/v) glycerol, 25 mm imidazole, pH 8.0), and lysed by sonication. After the addition of Triton X-100 to 0.2%, the cell slurry was incubated at 4°C and clarified by centrifugation. The supernatant was passed through a column packed with Ni-NTA resin (Qiagen,, which was washed sequentially with Ni-NTA binding buffer and eluted with HIS elution buffer (Ni-NTA binding buffer containing 225 mm imidazole). The eluate was sequentially purified with a column packed with Glutathione-agarose (Sigma-Aldrich,, which was equilibrated with phosphate-buffered saline (PBS) buffer (140 mm NaCl, 2.7 mm KCl, 10.1 mm Na2HPO4, 1.8 mm KH2PO4, pH 7.3). HisGST-fused proteins were eluted with GST elution buffer (50 mm Tris-HCl, pH 8.0, 10 mm glutathione), and the buffer was exchanged for kinase buffer (50 mm Tris-HCl, pH 7.5, 10 mm MgCl2, 1 mm EGTA), with a PD-10 column (GE Healthcare, The concentration of proteins was adjusted to 1 mg ml−1. p13suc1-sepharose beads were prepared as described previously (Harashima and Sekine, 2011).

Kinase reaction

To purify StrepIII-CDKA;1 protein complexes, seedlings of StrepIII-CDKA;1/+cdka;1-1/− plants were collected and frozen in liquid N2, then ground with a mortar and a pestle or a TissuLyser II (Qiagen). The resulting fine powder was thawed and suspended in IP buffer [25 mm Tris-HCl, 75 mm NaCl, 15 mm MgCl2, 15 mm EGTA, 0.1%(w/v) NP-40, pH 7.5] containing protease inhibitors (Complete, EDTA-free; Roche Applied Science, and phosphatase inhibitors (PhosSTOP; Roche Applied Science). Cell debris was pelleted by centrifugation at 20 000 g, held at 4°C, for 10 min, and then the supernatant was again clarified by centrifugation at 20 000 g, 4°C, for 20 min. The protein concentration of the supernatant was measured with a Bradford kit (Bio-Rad, Total protein (300 μg/200 μl IP buffer) was incubated with 20 μg of avidin (IBA, for 30 min at 4°C, and then mixed with 20 μl of 50% slurry Strep-Tactin Sepharose beads (IBA) and incubated overnight at 4°C with gentle agitation. After the beads had been washed three times with IP buffer followed by three washes with kinase buffer (50 mm Tris-HCl, pH 7.5, 10 mm MgCl2, 1 mm EGTA) (30 sec of centrifugation at 500 g is sufficient to sediment the beads), StrepIII-CDKA;1 protein complexes were analysed by kinase assays. To purify p13suc1-associated proteins, 300 μg of crude extracts of Col-0 inflorescences in IP buffer were mixed with 20 μl 50% slurry p13suc1-sepharose beads and incubated overnight at 4°C with gentle agitation. Then the beads were washed three times with IP buffer, followed by three times with kinase buffer. Kinase reactions were performed in kinase buffer containing 2.5 μg of HisGST-fused proteins, and 92.5 kBq of [γ-32P]ATP. After incubation for 30 min at 30°C, kinase reactions were stopped by the addition of sample buffer and boiling. Samples were analyzed on a 12% TGX gel (Bio-Rad), and after gels were stained with Coomassie Brilliant Blue (CBB) they were autoradiographed.


The authors thank Maren Heese (Institut de Biologie Moléculaire des Plantes, Strasbourg) for her help in analyzing the data. The authors are grateful to John Larkin (Louisiana State University, Baton Rouge) for helpful comments and a critical reading of the manuscript. We thank Laurent Deslandes, Iris Ottenschläger, Klaus Palme, Filipa Santos, Imre E. Somssich (Max-Planck-Institute for Plant Breeding Research, Cologne) for the kind gift of materials used in this study. This work was supported by an ATIP grant from the Centre National de la Recherche Scientifique (CNRS), ERC starting grant from the European Union and a grant from the Volkswagen Foundation to AS.

Accession numbers: The sequence data from this article are as follows: CDKA;1 has gene code At3g48750 and GenBank accession number NM_114734; CDC6 has gene code At2g29680 and GenBank accession number NM_179806; CDT1 has gene code At2g31270 and GenBank accession number NM_128683.3; CKS1 has gene code At2G27960 and GenBank accession number NM_128355.2; CYCB1;2 has gene code At5g06150 and GenBank accession number NM_120697.2; CYCD3;1 has gene code At4g34160 and GenBank accession number NM_119579.2; F20B24.1 has gene code At1G10590 and GenBank accession number NM_179301.3; GPX2 has gene code At2g31570 and GenBank accession number NM_128714.3; GSTF2 has gene code At4g02520 and GenBank accession number NM_116486.2; HSC70.1 has gene code At5G02500 and GenBank accession number NM_001125684.1; histone H1 has gene code At1g06760 and GenBank accession number NM_100553.2; MCF19.18 has gene code At5G45510 and GenBank accession number NM_001085250.1; MT2A has gene code At3g09390 and GenBank accession number NM_111773.3; MQK4.26 has gene code At5G16510 and GenBank accession number NM_180500.2; UBC35 UBC13A has gene code At1G78870 and GenBank accession number NM_001036221.2.