The Arabidopsis B-type response regulator 18 homomerizes and positively regulates cytokinin responses




In higher plants, the two-component system (TCS) is a signaling mechanism based on a His-to-Asp phosphorelay. The Arabidopsis TCS involves three different types of proteins, namely the histidine kinases (AHKs), the histidine phosphotransfer proteins (AHPs) and the response regulators (ARRs). The ARRs comprise three different families, namely A, B and C types, according to their protein structure. While some members of the B-type family of ARRs have been studied extensively and reported to act as DNA-binding transcriptional regulators, very limited information is available for other B-type ARRs such as ARR18. In this study, we characterize in detail the molecular and functional properties of ARR18. ARR18 acts as a transcriptional regulator in plant cells and forms homodimers in planta as shown by FRET–FLIM studies. As demonstrated by mutational analysis, the aspartate at position 70 (D70) in the receiver domain of ARR18 acts as crucial phosphorylation site. The modification of D70 affects the response regulator’s ability to homodimerize and to activate its target genes. Furthermore, physiological investigations of Arabidopsis lines ectopically expressing ARR18 introduce ARR18 as a new member within the cytokinin-regulated response pathway regulating root elongation.


Bacteria and higher eukaryotes like yeasts and plants rely on signaling mechanisms such as the two-component system (TCS) to sense and react to a broad spectrum of environmental and endogenous stimuli. In general, the transduction and amplification of a signal within TCS is achieved through a histidine (His) to aspartate (Asp) phosphorelay mechanism. In bacteria, the conserved basic phosphotransfer happens predominantly between a histidine protein kinase (HK) and a response regulator (RR) protein (Stock et al., 1989). Upon stimulus, the bacterial HK autophosphorylates and regulates the phosphorylation status of the RR. The bacterial RRs are fundamental control elements functioning as phosphorylation-triggered switches that mediate many cellular responses. Most of the bacterial RRs are transcription factors that consist of a conserved N-terminal receiver domain and a C-terminal DNA-binding domain. The phosphorylated RRs modulate the transcriptional activity of their target genes (West and Stock, 2001). Activation of RR transcription factors in bacteria typically involves dimerization or higher-order oligomerization (Fiedler and Weiss, 1995; Stock et al., 2000; Jeon et al., 2001; Galperin, 2006). The recently solved structure of the Escherichia coli RR PhoB indicates that the protein monomers associate with two different structural dimers depending on the phosphorylation state and exhibit a differential DNA-binding capacity and ability for transcriptional regulation (Mack et al., 2009).

Higher plants, such as Arabidopsis thaliana, have developed a TCS phosphorelay that includes hybrid histidine kinases (AHK), response regulators (ARR) and an intermediate element, the histidine phosphotransfer proteins (AHP) (Stock and Da Re, 2000; Lohrmann and Harter, 2002; Schaller et al., 2008). Arabidopsis has been reported to have 11 AHKs or AHK-like proteins. Among them, three AHKs are shown to be cytokinin receptors, namely AHK2, AHK3 and AHK4/CRE1/WOL (Ueguchi et al., 2001; Yamada et al., 2001; Nishimura et al., 2004). Other AHKs are involved in ethylene signaling (Hua et al., 1998; Cho and Yoo, 2007; Grefen et al., 2008), megagametophyte development (Pischke et al., 2002), stress signaling (Tran et al., 2007) and H2O2-regulated stomatal closure, salt sensitivity and resistance to bacterial and fungal pathogens (Desikan et al., 2008; Pham et al., 2012). The five canonical AHP family members AHP1 through AHP5 encode proteins that contain the conserved His residue that is required for the phosphorelay (Hutchison et al., 2006), while the AHP6 gene encodes a non-phosphorylatable AHP and is, therefore, considered a pseudo AHP (Suzuki et al., 1998). Analysis of the complete genome of Arabidopsis reveals the existence of 23 ARRs. According to their domain composition and molecular function, the canonical ARRs can be classified as either A, B or C types (Grefen and Harter, 2004; Kiba et al., 2004). ARR18 is one of the 11 members of the B-type family. The output domain in all members of the B-type ARRs family contains at least one nuclear localization signal (NLS), a transactivation domain and a DNA-binding GARP domain (Grefen and Harter, 2004). The DNA-binding domain of some B-type ARRs, namely ARR1, ARR2, ARR10 and ARR11, has been shown to bind to the nucleotide motif 5′-(A/G)GAT(T/C)-3′in vitro in a sequence-specific manner (Lohrmann et al., 1999; Sakai et al., 2000; Imamura et al., 2003). These sequence motifs are enriched in the promoters of primary cytokinin response genes like those of the A-type response regulators ARR4, ARR5, ARR6 and ARR7 (Rashotte et al., 2003). Furthermore, these B-type response regulators are able to activate the transcription of an ARR6-promoter-uidA reporter gene (PARR6-uidA) in Arabidopsis protoplasts (Hwang and Sheen, 2001). In conclusion, although it is of general consensus that, due to their highly conserved structure, all B-type response regulators act as transcriptional regulators, among the 11 B-type ARRs, this situation has only been directly demonstrated for ARR1, ARR2, ARR10 and ARR11 (Sakai et al., 2000; Hwang and Sheen, 2001; Imamura et al., 2003; Mason et al., 2005).

Within the B-type ARRs, ARR18 belongs to the cytokinin-responsive subfamily-1 that comprises seven proteins, namely ARR1, ARR2, ARR10, ARR11, ARR12, ARR14 and ARR18. Among them, ARR1, ARR2, ARR10 and ARR12 have been implicated previously in cytokinin responses based on the phenotype of their corresponding loss-of-function mutants and overexpression studies in Arabidopsis plants or protoplasts (Hwang and Sheen, 2001; Sakai et al., 2001; Imamura et al., 2003; Mason et al., 2005). The majority of subfamily-1 ARR genes, including ARR18, is expressed particularly in tissues and organs where cytokinin plays a crucial regulatory role such as in the apical shoot meristem and developing leaves (Mason et al., 2004). Single mutants of the subfamily-1 ARRs exhibit a decreased sensitivity to cytokinin with respect to root elongation and shoot regeneration from calli (Mason et al., 2005; Ishida et al., 2008). This cytokinin insensitive phenotype increases in higher-order mutants (Mason et al., 2005; Ishida et al., 2008).

In the plant TCS, ARRs are involved in a multi-step phosphorelay and receive the phosphate group at a conserved aspartate residue (Grefen and Harter, 2004). By mutation of this aspartate (D) to either a glutamate (E) or asparagine (N), two ARR protein versions can be engineered that either mimic a constitutive active state (D to E mutation) or a constitutive inactive state (D to N mutation) of the protein (Hwang and Sheen, 2001; Hass et al., 2004; Mira-Rodado et al., 2007). Such mutants have been essential to understand, for example, how phosphorylation of ARR2 is required for the control of leaf longevity (Kim et al., 2006). In addition, a constitutive active state version ofARR2 was shown to regulate the transactivation ability of ARR2 on the ERF1 and ARR6 promoters (Hass et al., 2004; Kim et al., 2006).

Several Arabidopsis TCS elements are reported to form dimers similar to the members of bacterial TCS. The cytoplasmic transmitter domains of AHK2, 3 and 4 (Dortay et al., 2006, 2008), the full-length AHK3 and 4 proteins (Caesar et al., 2011b), and the full-length ethylene receptors form homo- and heterodimers (Gao et al., 2008; Grefen et al., 2008). Three AHPs, namely AHP1, AHP2 and AHP3, have equally been reported to form homo- and heterodimers (Dortay et al., 2006, 2008; Punwani et al., 2010). However, although RRs are hypothesized to undergo conformational changes after receiving the phosphate group from the AHPs that would allow the formation of functional dimers (Kim, 2008), no experimental proof has yet been provided for any plant ARR including ARR18.

In this paper, we show that ARR18 acts as a transcriptional regulator and forms homodimers in planta. Moreover, we show that the aspartate at position 70 (D70) in the receiver domain of ARR18 functions as a crucial phospho-accepting site. The modification of D70 in ARR18 interferes with its homomerization capacity and transcriptional activity. The results of our physiological investigations introduce ARR18 as a new member of the cytokinin-regulated root elongation response.


ARR18 is a transcription factor

Sequence analyses indicate that ARR18 is a member of the B-type ARR subfamily. Its N-terminal receiver domain (RD) is followed by a long C-terminal region [output domain (OD)], where a potential NLS, a DNA-binding domain and a transactivation domain are located (Figures 1a and S1). As this architecture is found in other B-type ARRs that act as transcription factors (Sakai et al., 2000; Hass et al., 2004; Mason et al., 2004), the question of whether ARR18 might also function as a nuclear transcriptional regulator arose. Firstly, we tested whether the predicted NLS in the OD of ARR18 protein is functional. Therefore, the subcellular localization of the GFP fusions of full-length ARR18 as well as its receiver and output domains (ARR18RD and ARR18OD) were examined in transiently transformed tobacco (Nicotiana benthamiana) epidermal leaf cells by confocal laser scanning microscopy (CLSM). An ARR2-RFP fusion served as a nuclear marker (Hass et al., 2004) as it represents one of the few B-type ARRs in which both the nuclear localization and function as transcription factor has been proven (Sakai et al., 2000; Lohrmann et al., 2001). The analysis revealed localization of full-length ARR18–GFP and ARR18OD–GFP in the nucleus, whereas ARR18RD–GFP lacking the putative NLS sequence was excluded from the nuclear compartment (Figure 1b). These data show that the output domain of ARR18 is essential for its nuclear localization and suggest that the single predicted NLS is responsible for ARR18 nuclear localization.

Figure 1.

 ARR18 localizes to the nucleus and activates the cytokinin-responsive pARR5:uidA reporter gene.
(a) Scheme of the structural characteristics of the B-type response regulator ARR18. The conserved aspartate at position 70 (D70) is indicated. Oval, Receiver domain; bar, NLS; rhomboid, GARP DNA-binding domain; parallelogram, transactivation domain.
(b) Confocal images of tobacco (Nicotiana benthamiana) epidermal leaf cells expressing the indicated GFP and RFP fusion proteins. Images from the emission channels for GFP, RFP and overlays with the bright field images are shown. ARR2-RFP was used as nuclear marker. ARR18RD and ARR18OD stand for ARR18 receiver (RD, including amino acids from position 1 to 144) and output domains (OD, including amino acids from positions 145 to 636) respectively. The bars represent 10 μm.
(c) Reporter gene activation assay. Arabidopsis protoplasts were co-transfected with either the PARR5:uidA reporter construct and the P35S:NAN transformation efficiency standard alone (control) or together with an effector plasmid expressing either wild type ARR18 (18), bZIP63 or ARR2. The relative GUS/NAN activity is shown on Y-axis. Mean values and standard deviations were calculated from three independent transfection experiments. *Indicates statistically significant differences to the control ( 0.005 with t-test).

We next addressed the question of whether ARR18 is able to bind DNA and to activate known B-type ARRs target genes. The capacity of ARR18 to bind to the promoter region of A-type ARRs has recently been reported by chromatin immunoprecipitation assays where, a mutated form of ARR18 lacking the initial 45 residues of the protein was able to bind to the promoter region of the ARR6 gene (Liang et al., 2012). We tested the transcription activation capacity of full-length ARR18 on the ARR5 promoter that, like ARR6, contains the sequence motif 5′-(A/G)GAT(T/C)-3′ (Rashotte et al., 2003), using an activator/reporter gene assay in Arabidopsis protoplasts (Walter et al., 2004). Therefore, the ARR5 promoter sequence was fused to the uidA reporter gene (PARR5:uidA) (Brandstatter and Kieber, 1998) and its activity determined relative to that of the NAN transformation efficiency control (Kirby and Kavanagh, 2002) in the presence of either ARR18 or ARR2 (positive control; Hwang and Sheen, 2001). The Arabidopsis bZIP transcription factor 63 (AtbZIP63) served as a negative control as it does not recognize the ARR5 promoter but binds to G-/C-boxes present in other type of genes (Kirchler et al., 2010). As shown in Figure 1(c), we observed an increased activity of the PARR5:uidA reporter gene when full-length ARR18 and ARR2 were used as effector proteins. In contrast, bZIP63 failed to activate the reporter gene (Figure 1c). These results indicate that ARR18 recognizes the ARR5 promoter in vivo and functions as a transcriptional regulator.

ARR18 forms homomers in planta

As described above, the dimerization of bacterial response regulators has been known for a long time (Fiedler and Weiss, 1995). In contrast, in Arabidopsis and other plants, there are no reports on the capacity of B-type response regulators to form dimers. This failure might be mainly due to the strong auto-transactivation capacity of the B-type ARRs in the yeast two-hybrid system.

To analyze, whether ARR18 is able to form homodimers, we conducted in planta Förster resonance energy transfer-fluorescence lifetime imaging microscopy (FRET–FLIM) analysis in nuclei of transiently transformed tobacco epidermal leaf cells, as described previously by Caesar et al. (2011a). Hence, the ARR18 protein was fused to GFP (ARR18–GFP) providing the donor fusion and to mCherry (ARR18–mCherry) providing the acceptor fusion for FRET–FLIM (Figure 2a). It is important to note that C-terminal GFP fusions of ARR18 are functional in planta (for more details see Figure 5). ARR18–GFP alone and the TCS-unrelated NLS-fused Lesion Simulating Disease 1-mCherry fusion (LSD1–mCherry) served as negative controls in the FRET–FLIM study. LSD1 is a negative regulator of cell death by acting as a rheostat for reactive oxygen species (Dietrich et al., 1997; Kaminaka et al., 2006). To determine the maximal FRET–FLIM efficiency, a GFP–mCherry fusion was attached to ARR18 resulting in ARR18–GFP–mCherry (Bisson et al., 2009). The expression of all protein fusions was under the control of the estradiol-inducible promoter to ensure similar levels of protein accumulation (Bisson et al., 2009). Determined by CLSM, the fusion proteins accumulated to very similar levels in the nuclei 4 h after application of estradiol (Figure 2b). As expected, a strongly reduced donor GFP fluorescence lifetime (FLT) was observed for the ARR18–GFP–mCherry fusion (Figure 2a). When the FLT of ARR18–GFP was recorded in the presence ARR18–mCherry a significantly shorter lifetime was observed compared with that of ARR18–GFP alone and the ARR18–GFP/LSD1–mCherry control protein pair (Figure 2a). These FRET–FLIM data leads to the conclusion that ARR18–GFP and ARR18–mCherry are located very closely to each other, indicating their homodimerization or oligomerization inside the nucleus.

Figure 2.

 ARR18 forms homomers in planta.
(a) FRET–FLIM interaction assay of ARR18 homomerization. The indicated GFP and mCherry fusion proteins were expressed in tobacco cells and the fluorescence lifetime (FLT) of GFP was recorded. Single transformation of ARR18 fused to both GFP and mCherry (ARR18–GFP–mCherry) served as positive control. Single transformation of ARR18 fused to GFP (ARR18–GFP) was used to determine the FLT background value. Values are means and standard deviations of at least 15 independent measurements per experiment. *Indicates significant FLT differences to the ARR18–GFP background control ( 0.005 with t-test).
(b) Confocal images of tobacco epidermal leaf cells expressing the indicated GFP and mCherry fusion proteins. The emission channels for the GFP and RFP fluorescence and the overlays with the bright field images are shown. The bars represent 10 μm.
(c) Yeast membrane-based Split-Ubiquitin protein–protein interaction analysis. ARR18-Cub was co-transformed with either ARR18-Nub, pAI-Alg5 (positive control) or pDL2-Alg5 (negative control) in yeast. After transformation, the yeast was grown on vector selective media (CSM-L, W) and on interaction selective medium (CSM-L, W, H, Ade) in a dilution series of OD600 from 1.0 to 0.01.

In order to corroborate these results, we also performed protein–protein interaction studies in a membrane-based Split-Ubiquitin system (Figure 2c). This method provides a tool that overcomes the auto-activation problems that would arise in a classical yeast two-hybrid system due to the transcriptional activation properties of transcription factors such as ARR18 (Mockli et al., 2007). As shown in Figure 2(c), ARR18 formed dimers in yeast as well, supporting the in planta FRET–FLIM data. Thus, the homomerization of response regulators is not only restricted to bacteria but also occurs in plants.

Homomerization of ARR18 is Asp70-dependent

In bacteria, the response regulators’ RD mediates the formation of dimers in an aspartate phosphorylation-dependent manner (Chen et al., 2003; Toro-Roman et al., 2005; Mack et al., 2009). We analyzed next whether an analogous aspartate phosphorylation in the RD of ARR18 is required for its homomerization in plant cells. The amino acid sequence alignment of the 11 B-type ARR RDs revealed that only one aspartate is conserved throughout the complete family of response regulators (Figure S1). In ARR2, it corresponds to the aspartate at position 80 (D80) and has been proven to be the phosphorylation site of the response regulator (Hass et al., 2004). By in silico analysis, in ARR18, this aspartate is placed at position 70 (D70) and, thus, represents the putative phosphorylation site of ARR18 (Figure S1). Due to the thermodynamic instability of phospho-Asp, a direct proof of ARR phosphorylation by biochemical or mass spectrometric approaches is very difficult. This problem is bypassed by a widely accepted molecular genetics approach, in which the putative phosphorylated amino acid is mutated. Hence, we substituted the D70 of ARR18 to either a glutamate (D70E) or asparagine (D70N) by site-directed mutagenesis, engineering ARR18 proteins that either mimic a constitutive phosphorylated state (ARR18D70E, gain-of-function version) or a constitutive non-phosphorylated state (ARR18D70N, loss-of-function version) as has already been reported for both A- and B-type ARRs (Hwang and Sheen, 2001; Hass et al., 2004; Kim et al., 2006; Mira-Rodado et al., 2007). Then, donor (GFP) and acceptor (mCherry) fusion constructs of these ARR18 mutant versions were generated, and the encoded fusion proteins were tested by FRET–FLIM for their association in the nuclei of transiently transformed tobacco cells (Figure 3). Again, on the basis of CLSM images, the fusion constructs were expressed to very similar levels (Figure S2). As shown in Figure 3(a), we observed a significantly reduced FLT for the ARR18–GFP/ARR18D70E–mCherry protein pair. When the loss-of-function version of ARR18 (ARR18D70N) was used as acceptor, the FLT of ARR18–GFP was not changed (Figure 3a). This finding implies that wild type ARR18 can form homomers with gain-of-function ARR18D70E but not with the loss-of-function version of ARR18 (ARR18D70N). In order to further confirm the function of the putative phosphorylation site and, thus, the importance of the D70 for ARR18 homomerization, we also generated mutations in a nearby aspartate at position 75 (D75) (Figure S1) in ARR18 to either glutamate (ARR18D75E) or asparagine (ARR18D75N) and used these constructs for FRET–FLIM experiments (Figure 3b). We observed a significant reduction of the FLT of GFP, when both the ARR18D75N and ARR18D75E mutant versions were tested against wild type ARR18–GFP (Figure 3b). This finding provides a clear hint that the mutation of another aspartate different to D70 in the RD of ARR18 does not affect the homomerization capability of ARR18. This situation strongly suggests that D70 is the in planta phosphorylatable amino acid. Similar to wild type ARR18, gain-of-function ARR18D70E homomerized with wild type ARR18 and ARR18D70E but not with ARR18D70N (Figure 3c). However, when the loss-of-function ARR18D70N version was tested for its homomerization ability, it associated only with ARR18D70N and not with wild type ARR18 or ARR18D70E (Figure 3d). This finding implies that ARR18 can form homodimers only when the ARR18 monomers are in the same modification state. Our data also imply that wild type phosphorylated ARR18 is able to interact with ARR18D70E but not with ARR18D70N.

Figure 3.

 ARR18 homomerizes in a phosphorylation-dependent manner.
FRET–FLIM studies of ARR18–GFP interaction with different ARR18 D70 mutants: ARR18D70E–mCherry, ARR18D70N–mCherry (a), ARR18D75E and ARR18D75N (b). Also depicted are interaction studies of ARR18D70E–GFP (c) and ARR18D70N–GFP (d) with ARR18D70E–mCherry, ARR18D70N–mCherry and wild type ARR18–mCherry. The fusion proteins were expressed in tobacco cells and the fluorescence lifetime (FLT) of GFP was recorded. Single transformation of ARR18 fused to GFP–mCherry (ARR18–GFP–mCherry) served as positive control. The FLT background values were measured in cells transformed with either ARR18–GFP (a, b), ARR18D70E–GFP (c) or ARR18D70N–GFP (d). Values are presented as mean and standard deviation of at least 15 independent measurements per experiment. *Indicates significant FLT differences to the ARR18–GFP, ARR18D70E–GFP or ARR18D70N–GFP background controls ( 0.05 with t-test).

The transcriptional activity of ARR18 is Asp70-dependent

As shown above, the ability of ARR18 to form homomers depends on D70 and its phosphorylation state. Therefore, the question arises whether D70 phosphorylation also affects the transcriptional activity of ARR18. This scenario was tested in Arabidopsis cell culture protoplasts using again the PARR5:uidA as the reporter gene (Figure 4a). Wild type ARR18 and the different mutant versions of ARR18, namely ARR18D70E and ARR18D70N were co-transformed with PARR5:uidA, P35S:NAN and the relative GUS/NAN activity was determined. As already shown in Figure 1(c), wild type ARR18 activated PARR5:uidA (Figure 4a). The gain-of-function ARR18D70E activated the reporter gene to an even higher extent than wild type ARR18 (Figure 4a). In contrast, loss-of-function ARR18D70N activated the promoter to a significantly less extent (Figure 4a). The residual activation capacity observed in the ARR18D70Nis a phenomenon already described for loss-of-function ARR2D80N mutants (Hwang and Sheen, 2001; Kim et al., 2006). The outcome of competition binding assays in Arabidopsis cell culture protoplasts using again, PARR5:uidA as reporter gene, suggests that the phosphorylation state and, thus, the homomerization state of ARR18 have no influence on the DNA-binding capacity of the response regulator (Figure S3).

Figure 4.

 Asp70-to-Glu (D70E) mutation enhances the transactivation capacity of ARR18 but not its intracellular distribution.
(a) Reporter gene activation assay. Arabidopsis protoplasts were co-transfected with either the PARR5:uidA reporter gene construct and the P35S:NAN transformation efficiency standard alone (control) or in combination with different effector plasmids expressing either wild type ARR18 (18), ARR18D70N (18D70N, ARR18 loss-of-function), ARR18D70E (18D70E, ARR18 gain-of-function) or ARR18D75N (18D75N, ARR18 carrying an Asp to Asn mutation in Asp75 residue that is not targeted for phosphorylation in the TCS phosphorelay) proteins. Mean values and standard deviations of GUS/NAN activity were calculated from four independent transfection experiments. *Indicates statistically significant differences to the control ( 0.05 with t-test).
(b) Reporter gene activation assay as described in (a). The effector proteins in this case were: ARR18 (18), ARR18D70N (18D70N, ARR18 loss-of-function) or ARR18D70E (18D70E, ARR18 gain-of-function). The transfected protoplasts were incubated with 100 nmt-zeatin (white) or without the hormone (grey). Immunoblot with an anti-GFP antibody shows protein level of ARR18, ARR18D70E, and ARR18D70N. *Indicates statistically significant differences to each non-hormone treated probe ( 0.05 with t-test).
(c) Confocal images of tobacco epidermal leaf cells expressing the indicated RFP and GFP fusion proteins. Images from the emission channels for RFP, GFP and the overlays with the bright field images are shown. ARR18–GFP was used as nuclear marker. The bars represent 10 μm.

In order to corroborate again that D70 represents the aspartate targeted for phosphorylation in the ARR18 protein, we tested whether mutation of the next highly conserved aspartate (D75) into an asparagine (ARR18D75N) affected the activation capacity of the protein. In contrast to the ARR18D70N loss-of-function mutant, ARR18D75N activated the ARR5 promoter to the same extend as wild type ARR18 (Figure 4a). These results indicate again that the transcriptional activity of ARR18 target genes is dependent on D70 and its phosphorylation state and that aspartate 70 represents the crucial site for ARR18 phosphorylation in the receiver domain.

It has been reported that cytokinin further enhances the transcriptional activity of the B-type response regulators ARR2 and ARR10 in Arabidopsis (Hwang and Sheen, 2001; Hass et al., 2004; Kim et al., 2006). We, therefore, tested whether the cytokinin t-zeatin had a similar effect on ARR18. In the presence of the cytokinin, wild type ARR18 up-regulated the reporter gene activity to a level equivalent to that of the non-treated gain-of-function ARR18D70E mutant (Figure 4b). t-zeatin did not alter the activity of the PARR5:uidA reporter gene in either ARR18D70E or ARR18D70N versions (Figure 4b).

Regarding the ability described above of the various ARR18 versions to differentially regulate the ARR5 promoter, we also concentrated on whether these changes could be caused by their differential subcellular localization. To address this factor, we constructed a set of different ARR18 mutant proteins fused to RFP (ARR18D70E–RFP, ARR18D70N–RFP) and independently co-expressed these fusions with wild type ARR18–GFP in tobacco cell culture protoplasts. The co-localization study revealed that both mutant ARR18 fusions localized to the nucleus like wild-type ARR18 (Figure 4c). This finding indicates that the phosphorylation state of ARR18 does not affect its subcellular distribution.

ARR18 is involved in root elongation in a cytokinin-dependent manner

ARR18 belongs to the subfamily-1 of B-type ARRs. Because some of its members have already been implicated in cytokinin signaling, we studied whether ARR18 is also integrated in this response. Thus, we tested primary root growth, a cytokinin-mediated response, in the presence of exogenously applied cytokinin in two independent transgenic Arabidopsis lines that ectopically expressed ARR18–GFP (ARR18–GFPOX1 and ARR18–GFPOX2) (Figure S4a,b) and an ARR18 loss-of-function mutant (arr18) (Mason et al., 2005) (Figures 5 and S4c). The root length of wild-type seedlings was reduced strongly when treated with different concentrations of kinetin (Figure 5a,b). In contrast, arr18-1 seedlings showed a slightly decreased sensitivity to cytokinin resulting in longer roots than those of wild type plants (Figure 5a,b). Both ARR18 overexpressors (ARR18–GFPOX1 and ARR18–GFPOX2) were more sensitive to cytokinin than wild-type plants, resulting in a strongly reduced root length (Figure 5a,b). Although there is functional redundancy in the subfamily-1 of B-type ARRs, our data proves that ARR18 plays an additional role in the cytokinin-dependent root elongation response in Arabidopsis. Moreover we found that ARR18 interacts with AHP1, 2, 3 and 5 (Figure S5), which have been shown to function as positive regulators in cytokinin TCS signaling (Hutchison et al., 2006) and interact with the transmitter domain of the cytokinin receptors AHK2, AHK3 and AHK4 (Suzuki et al., 2001; Geisler-Lee et al., 2007; Dortay et al., 2008).

Figure 5.

 ARR18-overexpression lines and an arr18 loss-of-function mutant shows altered root growth response in the presence of cytokinin.
(a) Arabidopsis seedlings from an ARR18 T-DNA insertion mutant (arr18), two ectopic full-length ARR18–GFP overexpressors (ARR18–GFPOX1 and ARR18–GFPOX2) and their corresponding wild type (Ws-2) were grown on ½MS plates or ½MS medium supplemented with 1 μg L−1 kinetin under constant light conditions. Root length in the presence of kinetin was calculated as percentage relative to the non-kinetin treated control. *Indicates statistically significant differences compared with the wild type ( 0.005 with t-test).
(b) Cytokinin dose response analysis of root growth of the indicated Arabidopsis lines (Ws-2, arr18, ARR18–GFPOX1, ARR18–GFPOX2). Seedlings were grown vertically on control plates and plates supplemented with the indicated concentrations of kinetin (0.1 and 1 μg L−1). Root elongation was measured 8 days after sowing. Means and standard deviations were calculated from of at least three independent replicates (> 25 per replicate).


Dimerization of TCS family members is a process well known to occur in bacteria (Fiedler and Weiss, 1995; Stock et al., 2000; Jeon et al., 2001; Galperin, 2006; Mack et al., 2009). Only recently, it has been reported that, with the exception of RRs, dimerization of TCS members such as HKs and HPs also occurs in plants (Dortay et al., 2006, 2008; Gao et al., 2008; Grefen et al., 2008; Punwani et al., 2010; Wulfetange et al., 2011). The data on ARR18 homomerization (Figure 2) represent, therefore, a novel phenomenon in plants, as response regulators dimerization was, up to now, a process only found in bacteria (Benda et al., 2004; Bachhawat et al., 2005). Interestingly enough, the ARR18 homomerization process seems to resemble that of bacteria in which the receiver domain mediates the formation of dimers in a phosphorylation-dependent manner (Chen et al., 2003; Toro-Roman et al., 2005; Mack et al., 2009). Similarly, ARR18 homomerization can only occur between proteins that are in the same phosphorylation state and, thus, form either gain-of-function (ARR18D70E) or loss-of-function (ARR18D70N) homomers (Figure 3).

Aspartate 70 in ARR18 coincides in position with the aspartate responsible for ARR2 phosphorylation (Hwang and Sheen, 2001; Hass et al., 2004; Kim et al., 2006). This situation, together with the fact that it represents the only aspartate residue conserved throughout the family of B-type response regulators, provides a clear hint that D70 is the target amino acid for ARR18 phosphorylation. Supporting this idea, gain- or loss-of-function mutations of another aspartate in the RD besides D70 (ARR18D75E and ARR18D75N) have no influence on the homomerization ability of ARR18 (Figure 3b). We thus conclude that the modification status of aspartate 70 regulates the homomerization of ARR18.

We consequently postulate that ARR18, like bacterial RRs, forms active phosphorylated or inactive non-phosphorylated dimers (Baikalov et al., 1996; Park et al., 2002a,b; Nixon et al., 2005; Gao and Stock, 2010). In the bacterial Sinorhizobium meliloti DctD response regulator, the RD undergoes conformational changes upon phosphorylation that affect its transcriptional activity (Park et al., 2002a,b; Nixon et al., 2005). DctD crystallization revealed that in the non-phosphorylated state, the DctD dimer conformation prevents the formation of a ring-like structure required for the activity of the response regulator. In contrast, upon phosphorylation, DctD response regulator dimers undergo a conformational change that results in the formation of the functionally active ring-like structure (Park et al., 2002a,b; Nixon et al., 2005). Whether a similar mechanism also occurs in plants will be revealed when plant RRs crystal structures are solved. Until then, the ability of ARR18D70N to homomerize suggests that, similar to bacterial response regulators (Mack et al., 2009), non-phosphorylated inactive Arabidopsis response regulator homomers may also have a repressing function on the transcriptional activity of their target genes. This hypothesis is supported by the fact that the transcriptional activity of the loss-of-function ARR18D70N mutant is strongly impaired and clearly increased in the gain-of-function ARR18D70E mutant (Figure 4a). This is also observed in competition binding assays in Arabidopsis cell protoplasts, in which the ratio of either gain- or loss-of-function versions of the ARR18 protein (ARR18D70E and ARR18D70N) are increased while the amount of wild-type ARR18 protein is kept constant (Figure S3). In this case, only an increased ratio of the ARR18D70N protein resulted in an inhibition of the reporter gene activity when compared with the wild type ARR18 single transformation (Figure S3). Whether ARR18D70E competes with wild-type ARR18 for the DNA-binding or whether the competition to activate the 5′-(A/G)GAT(T/C)-3′ promoter region is achieved differently remains unsolved. It is, in any case, possible that as in bacterial systems both inactive ARR18D70N and active ARR18 homomers bind the promoter regions of their target genes but only phosphorylated ARR18 homomers are active at a transcriptional level.

Again, the importance of aspartate 70 is underlined by the fact that mutations in other aspartate residues such as D75 resulted in ARR18 mutants with an activation capacity equal to that of wild-type ARR18 proteins (Figure 4a). Thus, D70 phosphorylation does not only modify the homomerization ability of ARR18 but also its transactivation capacity.

The failure to form heterodimers of the phosphorylated and non-phosphorylated forms would have the consequence that the ARR18-mediated transcriptional responses become more precise: Only when a significant pool of phosphorylated ARR18 is generated by an incoming phosphorelay, the active homomers can compete out the non-phosphorylated homomers at the cognate cis-acting element. The permanence of ARR18 transcriptional activity on its target genes depends, on the other hand, on the half-life of the phosphate residue at D70. However, phosphorylation half-lives are not known for any plant RR. Thus, as in bacterial RRs, phosphorylation-dependent dimerization or higher-order oligomerization might represent an essential component of transcriptional regulation of plant response regulators.

Here, we show that ARR18 does not only function as a transcription factor in a phosphorylation-dependent manner but also that its activation ability is regulated by cytokinin. The fact that wild type ARR18 activation capacity is increased upon hormone treatment (Figure 4b) suggests that, in the absence of cytokinin, the ectopically expressed wild-type ARR18 pool is not entirely phosphorylated in protoplasts. We postulate that upon cytokinin application, the amount of ‘active’ phosphorylated ARR18 protein increases, which, in turn, is responsible for the observed enhancement in ARR18 activation capacity. This idea is supported by our data on the gain-of-function ARR18D70E version. Here, the greater part of the ARR18 pool mimics a phosphorylated and thus an ‘active’ form of the protein, a finding that explains the higher level of activity observed for the ARR18D70E mutant in the absence of hormone. Consequently, we expect that cytokinin addition would not affect this ‘active-ARR18′-saturated system as observed in our results (Figure 4b). Similarly, in the loss-of-function ARR18D70N mutant, the greater part of the ARR18 pool consists of inactive protein that cannot be phosphorylated and activated in response to t-zeatin application.

Recently, it has been described that the deletion of the RD of ARR18 results in mutant plants that mimic a strong cytokinin response in the absence of the hormone (Liang et al., 2012). This phenomenon has been already described in other B-type ARRs (Sakai et al., 2001; Imamura et al., 2003) and indicates that the non-phosphorylated RD suppresses the transcriptional activity of B-type ARRs. Our data support this hypothesis and suggest that the phosphorylation of subfamily-1 ARRs, such as ARR18, is a cytokinin-mediated response regulating its transcriptional activity. The function of ARR18 in cytokinin signaling is further substantiated by the altered cytokinin-induced root length phenotype in both ARR18-overexpressing and arr18 loss-of-function lines in comparison with wild-type Arabidopsis (Figure 5).

In conclusion, our findings provide a new perspective on how the B-type response regulators might bind to their cognate DNA sequence and exert their function on target genes in plants.

Experimental Procedures

Cloning and site-directed mutagenesis

All the clones used in our experiments were constructed using Gateway™ technology (Invitrogen, The entry clones were either obtained by BP-reaction in pDONR207 or through TOPO-reaction using the pENTR/D-TOPO vector. cDNA preparations, derived from Arabidopsis flowers, were used as template to clone ARR18 (AT5G58080). For the generation of truncated ARR18 versions, the ARR18 RD (ARR18RD, including amino acids from positions 1 to 144) and the output domain (ARR18OD, including amino acids from positions 145 to 636) were amplified by polymerase chain reaction (PCR) using the ARR18 entry clone. Site-directed mutagenesis of ARR18 was carried out on the ARR18 entry clones using QuikChange® Site-Directed Mutagenesis Kit (Stratagene, The primer pairs used for cloning the D70E, D70N, D75E and D75N mutants are described in Table S1.The binary constructs for GFP and RFP fusion protein expression under the control of the 35S promoter were obtained via LR reaction using the corresponding ARR18, ARR18D70E and ARR18D70N entry clones and destination vectors pH7WGF2.0 and pB7WGR2.0 (Karimi et al., 2002).

FRET–FLIM interaction studies

ARR18, LSD1 and different ARR18 mutant entry clones (ARR18D70E, ARR18D70N, ARR18D75E and ARR18D75N) were recombined via LR reaction into estradiol-inducible pABindGFP–mCherry, pABindGFP and pABindmCherry expression vectors (Bisson et al., 2009) for the FRET analysis. All binary vectors were transformed into Agrobacterium tumefaciens strain GV3101 pMP90 and infiltrated into tobacco leaves (Schutze et al., 2009). The p19 protein from tomato bushy stunt virus cloned in pBIN61 (Voinnet et al., 2000) was used to suppress gene silencing. Transgene expression driven by the estradiol-inducible promoter was induced 2–3 days after infiltration by brushing the leaves with 20 μmβ-estradiol supplemented with 0.1% Tween 20. The fluorescence lifetime was determined after 4 h of induction using a custom-built confocal stage scanning microscope [CSSM; (Elgass et al., 2009)].

Arabidopsis protoplast transient expression assay

The constructs for the protoplast transformation assay were obtained via LR reaction with the entry clones ARR18, ARR18D70E, ARR18D70N, ARR2, bZIP63 and ARR18D75N and destination vectors pUBC–GFP (Grefen et al., 2010). The reporter construct PARR5:uidA and the internal control construct P35S:NAN described previously in the literature were also used in these experiments (Brandstatter and Kieber, 1998; Kirby and Kavanagh, 2002). Arabidopsis cell culture protoplasts were isolated and transfected as described previously (Schutze et al., 2009). The amount of each transformed plasmid was constant throughout the entire experimental set. To keep the total DNA quantity at the same level in each sample, unrelated DNA (pB7WGR2.0 empty vector used in the previous section for RFP localization studies; Karimi et al., 2002) was added when required. Transfected protoplasts were incubated in the dark overnight at 23°C. The samples were incubated for a further 4 h with or without 100 nmt-Zeatin. The P35S:NAN construct was used as an internal control to normalize the variations of each transfection due to transformation efficiency and cell viability. GUS and NAN enzyme assays were performed as described previously (Ehlert et al., 2006). The results are shown as the means of relative GUS/NAN activities. All transient experiments were repeated at least four times.

Yeast membrane-based Split-Ubiquitin protein–protein interaction assay

The membrane-based Split-Ubiquitin system for protein–protein interaction analysis was performed according to the DUALhunter kit manual (Dualsystems Biotech AG, (Mockli et al., 2007). Both ARR18-Cub and ARR18-Nubconstructswere cloned via LR reaction using the ARR18 entry clone described above and the pDHB1 and pPR3-N destination vectors respectively (supplied in the kit). ARR18-Cub was co-transformed with either ARR18-Nub, pAI-Alg5 or pDL2-Alg5 in the yeast strain NMY51 using the lithium acetate/SS-DNA/PEG method (Grefen et al., 2009). After transformation, the yeast was grown on vector selective media (CSM-L, W) and incubated for 3 days at 28°C. Subsequently, five independent yeast clones were pooled and grown in liquid vector selective medium to an OD600 of 1.0. To test protein interaction, 7 μl of each culture in a dilution series of OD600 from 1.0 to 0.01 were dropped on interaction selective medium (CSM-L, W, H, Ade) and incubated for 3 days at 28°C. Co-transformations of ARR18-Cub with either pAI-Alg5 or pDL2-Alg5, supplied in the kit, served as positive and negative controls respectively.

Root growth inhibition assay

Seedlings were surface sterilized using 70% ethanol with 0.01% Triton X-100 for 10 min followed by 10 min incubation in 95% ethanol. Stratification for 24 h in the dark at 4°C was performed to break the seed dormancy and initiate synchronized germination. Seedlings were grown vertically under constant light on ½ strength Murashige and Skoog (MS) medium without sucrose and supplemented with appropriate concentrations of kinetin for the root elongation assay. Root lengths were measured after 8 days and seedlings that had not germinated within 2 days after stratification were excluded from the analysis. Means and standard deviations were calculated from at least 25 seedlings per line and at least three independent replicates were measured. To calculate root relative values, first, both non-treated (X) and kinetin treated (XK) root length mean values were calculated for each line of seedlings following the formula: inline image, where n represents the number of seedlings measured per line and Xi and inline image, the hole data set of root length values per treatment. Finally, each line could be addressed a relative mean value (XR) calculated as follows: XR = XK/X.

In order to compare the t-tests of relative data, we did an estimation of the standard deviation (SD) of the relative root lengths according to the Gaussian rules of error propagation. The relative SD (SXR) was calculated as follows with mean values X, XK and XR of data samples Xi, inline image and inline image and SD values SX, SXK and SXR respectively:


The different values for the data sample inline image, required to compare the t-tests of relative data, were calculated from each kinetin treated sample value (inline image) using the following formula: inline image.


We thank the transformation and green house facility of the ZMBP for their support, and Felicity de Courcy for proofreading the manuscript. We appreciate A. Bleckmann, R. Simon, K. Schumacher, C. Grefen and J. Horak for provision of vectors and E.G. Schaller and J.J. Kieber for seeds and the pARR5:uidA construct. This work was supported by a DFG grants to K.H. (HA 2146/8-2; HA 2146/10-1), a DAAD PhD fellowship to M.V. and a PhD fellowship of the University of Tübingen to K.E.