SEARCH

SEARCH BY CITATION

Keywords:

  • cytokinin;
  • signal transduction;
  • two-component system;
  • yeast two-hybrid system;
  • protein–protein interaction

Abstract

  1. Top of page
  2. Abstract
  3. Results
  4. Discussion
  5. Conclusions
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information

The signal of the plant hormone cytokinin is perceived by membrane-located sensor histidine kinases and transduced by other members of the plant two-component system. In Arabidopsis thaliana, 28 two-component system proteins (phosphotransmitters and response regulators) act downstream of three receptors, transmitting the signal from the membrane to the nucleus and modulating the cellular response. Although the principal signaling mechanism has been elucidated, redundancy in the system has made it difficult to understand which of the many components interact to control the downstream biological processes. Here, we present a large-scale interaction study comprising most members of the Arabidopsis cytokinin signaling pathway. Using the yeast two-hybrid system, we detected 42 new interactions, of which more than 90% were confirmed by in vitro coaffinity purification. There are distinct patterns of interaction between protein families, but only a few interactions between proteins of the same family. An interaction map of this signaling pathway shows the Arabidopsis histidine phosphotransfer proteins as hubs, which interact with members from all other protein families, mostly in a redundant fashion. Domain-mapping experiments revealed the interaction domains of the proteins of this pathway. Analyses of Arabidopsis histidine phosphotransfer protein 5 mutant proteins showed that the presence of the canonical phospho-accepting histidine residue is not required for the interactions. Interaction of A-type response regulators with Arabidopsis histidine phosphotransfer proteins but not with B-type response regulators suggests that their known activity in feedback regulation may be realized by interfering at the level of Arabidopsis histidine phosphotransfer protein-mediated signaling. This study contributes to our understanding of the protein interactions of the cytokinin-signaling system and provides a framework for further functional studies in planta.

Abbreviations
3AT

3-amino-1,2,4-triazole

AHK

Arabidopsis histidine kinase

AHP

Arabidopsis histidine phosphotransfer protein

ARR

Arabidopsis response regulator

GST

glutathione S-transferase

MAPK

mitogen activated protein kinase

TCS

two-component system

Cytokinins, a class of plant hormones, are recognized by membrane-located sensor histidine kinases. Within the cell, the signal is transduced through a complex type of the two-component signaling system (TCS) [1–7]. Intriguingly, simple TCSs are widespread in prokaryotes, where the different components, their functions and the underlying mechanisms have been well characterized [8]. With only a few exceptions, the TCS in those organisms consists of two proteins, namely a sensor histidine kinase and its cognate response regulator. In eukaryotes, the TCS is found in lower, unicellular genera and in plants, where a third class of proteins, histidine phosphotransfer proteins, participates in signal transfer.

In Arabidopsis thaliana, the model for cytokinin signaling through the TCS predicts that the binding of a ligand to one of the three cytokinin receptors [Arabidopsis histidine kinase 2 (AHK2), AHK3, CRE1/AHK4][6,9–12] leads to autophosphorylation of the receptor at a conserved histidine residue. The phosphoryl group is subsequently transferred to an aspartate in the C-terminal receiver domain of the receptor and from there to a histidine phosphotransfer protein (AHP). After phosphorylation, the AHPs localize to the nucleus, where they phosphorylate B-type response regulator proteins, which in turn activate the transcription of their target genes. One class of target genes codes for A-type response regulators [13–16]. A-type response regulator proteins have been shown to act as regulators of cytokinin signaling [16,17]. The Arabidopsis genome encodes five AHPs and 23 response regulators (ARRs) [5,18]. Among the AHPs, a function in the phosphorelay has been shown in vitro for AHP1, AHP2, AHP3 and AHP5 [1,19]. The ARRs are divided into two major classes: (a) 11 B-type ARRs, which in addition to the response regulator domain also contain a DNA-binding and an activation domain, and several of which function as transcription factors [20–25]; (b) 10 A-type ARRs, which, in addition to the response regulator domain, have only small N-terminal and C-terminal extensions [17]. Besides these two major classes, Kiba and colleagues [26] proposed a third subclass of ARRs. This group consists of at least two members (ARR22 and ARR24), which are structurally related to A-type ARRs; however, their transcription is not induced by cytokinin [26]. For the majority of the members of the Arabidopsis TCS, a role in cytokinin signaling has been shown. There is less evidence for the signaling of other Arabidopsis histidine kinases through the TCS [25], although participation of the TCS in ethylene signaling has been proposed [21].

Phylogenetic analysis of different protein classes of the TCS has revealed that the system has expanded during evolution [5,27]. This raises the question of whether specific pathways are established within the signaling system, e.g. through specific interaction patterns between proteins, and, if so, whether these specific interactions contribute to specifying the many different cytokinin activities in the plant during these early signaling events. Previous studies of Arabidopsis TCS proteins have indicated their promiscuous interactions (supplementary Table S1). In contrast, in the simpler bacterial TCS, there are, with few exceptions, specific links between histidine kinase receptors and their cognate response regulators [28]. Thus, promiscuity of interaction may not be a general feature of two-component signaling, but could be a functionally relevant characteristics of the plant TCS.

To obtain more insight into the complexity of the plant TCS, we attempted to include as many members of the different protein classes as possible in an interaction analysis. This is also relevant because several genetic studies, which analyzed multiple mutants of the different TCS gene families, were hampered by the great level of redundancy in planta[17,25,29–31]. In this situation, knowledge about the ability of proteins to interact can be a valuable contribution to establish signaling links. Using today's methodology, it is difficult to investigate protein–protein interactions in planta on a larger scale. Taking this into account, we decided to employ the heterologous yeast two-hybrid system to conduct a large-scale analysis of the interactions between most members of the Arabidopsis cytokinin signaling pathway and to verify detected interactions by in vitro pull-down assays. The yeast two-hybrid system is more prone to detect false positives than the pull-down assay, which is regarded as the highest standard in protein–protein interaction studies. Therefore, we used the yeast two-hybrid method for a first round of interaction analysis, and then retested the novel interactions in a second step by the in vitro interaction test. This approach is commonly used in similar experiments in other organisms [32–35].

This study was designed to yield information about the protein–protein interaction patterns that could be involved in mediating the flow of cytokinin signal among the different members of this signaling pathway. Using deletion constructs of representative interaction partners, we mapped the interaction domains of this signal transduction pathway. Furthermore, site-directed mutagenesis experiments were performed to begin to study the underlying mechanisms of those interactions.

We detected many novel interactions between TCS proteins, which occur in distinguishable patterns. These interactions support the concept of redundancy in the cytokinin signaling pathway. Interactions between proteins of the same family were found as an exception but may indicate the possibility of the formation of protein homodimers and heterodimers. Importantly, AHPs interact with the majority of the other TCS proteins, supporting their predicted role as a hub. Additionally, amino acid substitution experiments with AHP5 indicate that the canonical phospho-accepting histidine residue is not required for these interactions. The implications of these findings for cytokinin signaling are discussed.

Results

  1. Top of page
  2. Abstract
  3. Results
  4. Discussion
  5. Conclusions
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information

Interaction screen of Arabidopsis TCS proteins

In order to study protein–protein interactions with the yeast two-hybrid system, we were able to clone 24 cDNAs into either the pENTR or the pDONR vectors of the Gateway system. For all three cytokinin receptor histidine kinases, partial cDNAs encoding the C-terminal cytoplasmic parts, according to predictions of the smart program, were used [36]. Of the five AHP genes, we obtained the cDNA clones of four (AHP1, AHP2, AHP3, AHP5). Of the 23 response regulators, we were able to clone 15 (Table 1).

Table 1.   Protein interaction matrix of part of the Arabidopsis two-component signaling system. N, new interactions identified in this study; C, known and confirmed interactions; NC, known interactions that could not be confirmed in this system.
Prey vectorBait vector
AHKsAHPsA-type ARRsB-type ARRs
AHK2AHK3AHK4AHP5aARR3ARR4ARR5ARR6ARR7ARR8ARR9ARR15ARR16ARR1aARR2ARR10ARR14a
  1. a Autoactivating baits suppressed by 5 mm 3AT.

AHKs
 AHK2N  N            N
 AHK3N NN             
 AHK4   C             
AHPs
 AHP1NNC CCN NCC NCCNCN
 AHP2NNC CCN NCC NCCNCN
 AHP3NNC NNN NCC NCCCN
 AHP5NNC NNN NNN NCC N
A-type ARRs
 ARR3   N             
 ARR4   N             
 ARR5   N             
 ARR6   N             
 ARR7   N             
 ARR8   N             
 ARR9   N             
 ARR15   N             
 ARR16   N             
B-type ARRs
 ARR1   C             
 ARR2   C             
 ARR10
 ARR11   N             
 ARR14N              NN
 ARR20

For the in vivo interaction studies in the yeast two-hybrid system, all cloned cDNAs were shuttled via recombination into prey and bait vectors and tested for autoactivation (Table 1). Five clones, namely AHP1, AHP2, AHP3, ARR11 and ARR20, showed strong autoactivating activity and therefore could not be used as baits. However, all clones could be used in the prey configuration. Each protein pair was tested, where possible, in both configurations, leading to a total of 456 possible interactions investigated. Interactions were tested using two different reporter genes (HIS and URA), and growth was evaluated using several dilution steps to reduce the number of false positives. Each possible bait–prey combination was tested at least twice.

Using this approach, 68 interactions were detected. Of these, 42 are novel interactions, which had not been described before. The other 26 interactions have been previously described and were confirmed by our study (Table 1; supplementary Table S1). Of the 24 interactions that we could test in both configurations, 17 took place in both directions, whereas seven took place only in one direction. Most novel interactions were found with AHP5 (Fig. 1). We considered independent confirmation of interactions that were found by other yeast two-hybrid vector–strain combinations or by phosphotransfer studies as complementary evidence and focused additional experiments on the novel interactions.

image

Figure 1.  Protein–protein interactions of Arabidopsis histidine phosphotransfer protein 5 (AHP5) within the two-component signaling system of Arabidopsis thaliana. The figure shows the yeast two-hybrid analysis for AHP5 as bait with 24 different proteins of the two-component system as prey. A cell suspension with an A600 of 0.2 was prepared for each clone. Ten microliters of each cell suspension was transferred to SDII control (+His, +Ura) and SDIV interaction (–His, –Ura) media supplemented with 5 mm 3AT. In addition, two sets of dilutions (1 : 10 and 1 : 100) of each cell suspension were prepared and transferred to SDIV interaction medium.

Download figure to PowerPoint

Confirmation of novel interactions by in vitro coaffinity purification

Novel interactions, detected using the yeast two-hybrid system, were further investigated by coaffinity purifications. One partner was expressed as a glutathione S-transferase (GST) fusion protein, and the other partner was in vitro transcribed/translated in the presence of radioactively labeled methionine. Subsequently, both proteins were copurified via glutathione coupled to beads.

Of the 40 novel interactions tested with this method, 36 were verified in this biochemical assay (Fig. 2). Only four interactions that were detected in vivo could not be verified in vitro by this technique.

image

Figure 2.  Novel interactions confirmed by in vitro interaction assays. (A–C) GST fusion proteins were immobilized on glutathione agarose beads and incubated with [35S]methionine-labeled protein produced by in vitro transcription/translation. After separation by SDS/PAGE, autoradiography was performed. Lane 1: 35S-protein*. Lane 2: glutathione agarose beads + 35S-protein*. Lane 3: GST immobilized on glutathione agarose beads + 35S-protein*. Lane 4: GST fusion protein immobilized on glutathione agarose beads + 35S-protein*. (A) Novel receptor interactions. (B) Novel B-type Arabidopsis response regulator (ARR) interactions. (C) Novel A-type ARR interactions. *Radioactively labeled.

Download figure to PowerPoint

Towards a map of the TCS interaction network

Several interactions were detected between the cytokinin receptors. The cytoplasmic domains of AHK2 interacted with each other (Table 1). In contrast, the cytoplasmic domains of AHK3 and CRE1/AHK4, respectively, did not interact with themselves in our assay. However, AHK3 interacted with the cytoplasmic domains of both AHK2 and CRE1/AHK4 (Table 1). Receptor interaction indicates dimerization, which is a common feature of bacterial sensor histidine kinases [37,38]. The principal interaction motif is located in the N-terminal part of the histidine kinase domain [39,40], which was included in our constructs.

Our data indicate that cytokinin receptors may dimerize in the absence of the ligand, as the cytoplasmic parts of the receptors were sufficient for interaction. In order to test for specificity of the detected interactions, we cloned the cDNAs encoding the cytoplasmic domains of two other receptor histidine kinases, AHK1 and ETR1, and included them in our interaction screen. No interaction of these domains with the cytokinin receptor kinases could be detected, indicating that interactions between these receptors may be specific (data not shown). However, more in planta analyses are required to support the functionality of receptor interactions.

Most interactions were found between the AHPs and other components of the TCS. All tested AHPs interacted with all three receptors and with almost all B-type and A-type ARRs, further supporting their proposed role as signaling hubs (Table 1).

In general, there were no interactions between ARRs of the different types, or within an ARR class. Some exceptions to this general rule were found among the B-type ARRs. ARR14 interacted with itself, indicating the possibility of homodimer formation. Interactions were also detected with ARR2 and, unexpectedly, with the receptor AHK2 (Table 1).

Previously described interactions that we could not detect were between AHP1/AHP2 and ARR10 [41]. Differences in sensitivity between different yeast two-hybrid systems are not unusual and could be the reason for our inability to confirm these interactions. Another four published interactions (ARR1–AHP4; ARR2–AHP4; ARR11–AHP2; ARR22–AHP5; supplementary Table S1) could not be tested, as our set did not contain the corresponding clones (AHP4 and ARR22), or the fusion proteins were strong autoactivators in the bait configuration (AHP2 and ARR11). However, the fact that almost all of the known interactions that we could have detected with our matrix approach were found validates the system used in this study.

Interaction domain mapping of the TCS proteins

The protein–protein interaction domains of the TCS proteins were mapped by testing different truncated versions of representative proteins for each type of interaction (Fig. 3). The cytoplasmic part of the cytokinin receptor AHK2 was separated into four domains; the phospho-accepting part of the histidine kinase (HisKA), the HATPase, a connecting stretch and the receiver domain. The HATPase domain alone or in combination with the HisKA domain was autoactivating in the bait configuration. Thus, these constructs could only be tested in the prey vector. The other truncation constructs were not autoactivating as baits, or the autoactivation could be suppressed by the addition of 5 mm 3-amino-1,2,4-triazole (3AT) to the medium.

image

Figure 3.  Scheme of two-component system (TCS) protein domains used for interaction mapping. The cytoplasmic part of the cytokinin receptor Arabidopsis histidine kinase 2 (AHK2), the histidine phosphotransmitter Arabidopsis histidine phosphotransfer protein 5 (AHP5), the B-type response regulator Arabidopsis response regulator 14 (ARR14) and the A-type response regulator ARR4 were used as representatives of each cytokinin signaling protein family. Numbers above the domains indicate amino acid positions in the full-length protein. HisKA, the phospho-accepting part of the histidine kinase; ATPase, ATP-binding domain; RD, receiver domain; H, phosphotransmitter domain; Myb, DNA-binding domain.

Download figure to PowerPoint

In the case of the self-interaction of the cytoplasmic domain of AHK2, the interaction is mediated via the HisKA and the HATPase domain together. Either domain alone was not sufficient for the interaction (Fig. 4A). No interaction could be detected via the C-terminal half of the receptor. In the case of the interaction between AHP5 and AHK2, the complete cytoplasmic domain of the receptor was necessary for the interaction (Fig. 4A).

image

Figure 4.  Interaction domain mapping of selected two-component system proteins. (A- D) Full-length as well as deletion constructs of selected TCS genes were expressed in yeast together with their interaction partners. Cell suspensions with an A600 of 0.2 and two sets of dilutions (1 : 10 and 1 : 100) of each cell suspension were prepared. Ten microliters of each suspension was transferred to SDIV interaction (–His, –Ura, –Leu, –Trp) media or to SDIV interaction medium supplemented with 5 mm 3-AT. Numbers indicate amino acid positions in the full-length protein. (A) Arabidopsis histidine kinase 2 (AHK2), (B) Arabidopsis histidine phosphotransfer protein 5 (AHP5), (C) Arabidopsis response regulator 14 (ARR14) and (D) ARR4 interaction domain mapping.

Download figure to PowerPoint

The AHPs are rather small and contain only the Hpt domain. To test whether this domain can be divided further into subdomains, the protein was divided, so that both halves overlapped and contained the conserved Hpt core motif (XHQXKGSSXS) (Fig. 3). As an interaction hub of the TCS, AHPs interact with members of all other protein families in this signaling pathway. In all cases tested, the full-length AHP5 was necessary for the interaction (Fig. 4B). In the case of the B-type response regulator ARR14, the protein was divided into an N-terminal part containing the response regulator domain and a C-terminal fragment containing the Myb domain (Fig. 3). The Myb domain mediates the self-interaction of ARR14 (Fig. 4C). This domain was also sufficient for the interaction of ARR14 and ARR2 (Fig. 4C). In contrast, the full-length ARR14 was necessary for the interaction with AHP5 (Fig. 4C). As a representative of the A-type ARRs, ARR4 was chosen, as it has a rather long C-terminal extension after the response regulator domain. Also, in this case the whole protein was required for the interaction with an AHP (in our study AHP5) (Fig. 4D).

The conserved phospho-accepting histidine residue of AHP5 is not necessary for protein–protein interaction

The cytokinin signal is transmitted by the TCS via a multistep phosphorelay. Therefore, it is conceivable that the phosphorylation status − particularly the phosphorylation status of the canonical aspartate and histidine residues participating in this phosphorelay − might influence the interaction between the different proteins of the TCS. As the phosphorylation status of the members of the Arabidopsis TCS in the yeast two-hybrid system is not known, it is unclear whether the phosphorylation of those conserved residues is indeed a prerequisite for interaction or not. To test this further, we chose to investigate the interactions of AHP5 as a test case in more detail, as this protein showed the most interactions in our analysis (Table 1). Mutations were introduced into AHP5, converting the conserved, phospho-accepting histidine into a lysine (H83K) or an alanine (H83A). These mutant proteins were tested for interactions in the yeast two-hybrid system (Fig. 5).

image

Figure 5.  Mutation analysis of Arabidopsis histidine phosphotransfer protein 5 (AHP5) interactions in the two-component signaling system of Arabidopsis thaliana. Yeast two-hybrid analysis was performed by using AHP5 as bait in its native and mutant forms, in which the phospho-accepting histidine residue was substituted by lysine (H83K) or alanine (H83A), respectively. All three baits were screened for interaction with all 24 two-component system preys. A cell suspension with an A600 of 0.2 was prepared for each clone, and 10 µL of each cell suspension was transferred to SDII (+His, + Ura), control and SDIV (–His, Ura) interaction media supplemented with 5 mm 3AT.

Download figure to PowerPoint

The interaction pattern of AHP5 was unchanged with most partner proteins in the yeast two-hybrid assay (Fig. 5). Only two of the 15 tested interactions, namely with CRE1/AHK4 and ARR6, did not yield a positive result. For ARR6, this loss of interaction was dependent on whether the histidine was substituted by a lysine or an alanine (Fig. 5). To test these changed interaction properties independently, coaffinity purification experiments were done with the wild-type as well as with the mutated versions of AHP5 (Fig. 6). The more stringent in vitro binding studies showed for both cases interaction between the original as well as the mutated forms of AHP5. In summary, the results of the yeast two-hybrid and the in vitro coaffinity purification indicate that AHP5 can interact with other proteins of the TCS, regardless of the presence of the canonical histidine residue.

image

Figure 6. In vitro verification of changed interactions of mutated Arabidopsis histidine phosphotransfer protein 5 (AHP5). AHP5 and its mutant proteins H83K and H83A were immobilized on glutathione agarose beads as GST fusion proteins and incubated with [35S]methionine-labeled protein produced by in vitro transcription/translation. After separation by SDS/PAGE, autoradiography was performed. Lane 12: 35S-protein*. Lane 2: glutathione agarose beads + 35S-protein*. Lane 3: GST immobilized on glutathione agarose beads + 35S-protein. Lane 4: GST fusion protein immobilized on glutathione agarose beads + 35S-protein*. *Radioactively labeled.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Abstract
  3. Results
  4. Discussion
  5. Conclusions
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information

Numerous new interactions were identified in the Arabidopsis TCS

The analysis of the well-defined, evolutionarily conserved TCS system of Arabidopsis should reveal whether the large number of proteins involved leads to a high degree of redundancy or to signal specification within the system. A characteristic of the bacterial TCS is specific signaling through a defined set of partner proteins [28], but genetic data as well as a limited set of protein interaction studies had indicated a higher degree of redundancy in the plant TCS [17,25,29–31] (supplementary Table S1).

Performing a matrix screen, we found interactions in 68 of the 456 tested protein combinations. Of those, 42 represent previously unknown interactions and 26 represent interactions that were known from earlier work (supplementary Table S1). In order to reduce the number of false negatives, which could be due to steric hindrance of either bait or prey fusion proteins, TCS components were expressed in both orientations. Of the 24 interactions that we could test in both orientations, 17 took place in both directions and seven others were identified in only one direction, which increased the number of interactions by 29%. This is a common percentage of unidirectional interactions for this type of assay.

As the overlap of results obtained in different yeast two-hybrid experiments is often quite low [42], the fact that we could detect almost all previously known interactions indicates that the particular vector–yeast strain combination that we used is very sensitive. Whereas many large-scale and medium-scale studies of protein–protein interaction rely only on yeast two-hybrid data [34,43,44], we tested the quality of the results obtained from the in vivo interaction test by in vitro pull-down assays. About 90% (i.e. 36 of 40) of the new interactions were confirmed by this assay, compared to 35–82% in other large-scale yeast two-hybrid experiments [33,35]. These numbers illustrate the high quality and stringency of the yeast two-hybrid system used in this study. Many of the interactions reported here were recapitulated by a genome-wide interaction screen with various TCS proteins as bait (our unpublished data).

To be potentially of biological relevance, the interacting proteins need to be expressed in the same spatial and temporal fashion in the plant and occupy the same subcellular compartment. In fact, most of the TCS protein-encoding genes share at least some expression domains [4,17,20,29,30,45]. Furthermore, almost all interactions involve AHPs, which can shuttle between the cytoplasm and the nucleus and thus are expected to have access to most of the other TCS proteins [16]. According to these data, all interactions shown could take place in planta. One unexpected interaction is between the receptor AHK2 and the response regulator ARR14. The interaction between a hybrid histidine kinase and a response regulator is novel and intriguing. As for all interactions, for it to be biologically meaningful, the proteins must be in the same subcellular compartment. However, neither for AHK2 nor for ARR14 are the subcellular localizations currently known. To investigate this interaction in more detail in planta, further investigations are necessary.

Redundancy and specificity within the cytokinin signaling pathway

One of our essential findings, the similar interaction patterns of proteins belonging to the same family, provides a molecular explanation for functional redundancy of the TCS, which has been indicated by genetic studies [17,25,29–31]. Functional redundancy seems to be an important and inherent characteristic of the plant TCS, a signaling system involved in hormonal growth control by integrating intrinsic and extrinsic signals. This appears to be different to the simpler TCS of bacteria; for example, in Escherichia coli there is an almost equal number of receptor kinases and response regulators, and cross-signaling is believed to be a rare event [46]. In fact, using large-scale phosphorelay experiments, the high level of specificity within the bacterial TCS was recently demonstrated by Skerker et al. [28].

In plants, redundant paralogs may act as a buffer in the case of mutation and thus contribute to the genetic robustness of the organism. In yeast, it was shown that mutations in one protein can lead to changes in the transcriptional patterns of the respective paralog, thus compensating for the loss [47]. As many processes in the cell are mediated by protein–protein interactions, the paralog must also be able to interact with the interaction partner of the missing proteins to functionally complement it. In this respect, a certain level of redundancy in the interaction network would be expected in essential pathways of an organism − as we have found to be true for the Arabidopsis TCS (Fig. 7).

image

Figure 7.  Protein–protein interaction network of the cytokinin signaling pathway in Arabidopsis thaliana. In this figure, all interactions among the cytokinin signaling components are summarized. Of 68 interactions shown, 42 are novel and most were confirmed by in vitro interaction assays. The remaining 26 interactions have been previously published (supplementary Table S1). Each dot corresponds to a protein, and verified interactions are indicated by a line. Dashed lines correspond to those interactions found in the yeast two-hybrid analysis only, and white dots to proteins forming homomers. Shaded dots represent proteins that we could not use in this study.

Download figure to PowerPoint

Functional redundancy would also provide a cellular framework to integrate different signals into the TCS pathway, which would have similar consequences for the output. For example, it is known that diverse extrinsic factors such as phosphate availability [48], cold stress [49], circadian rhythm [50] and intrinsic developmental processes [51,52] modulate expression of single or multiple TCS genes [4]. Ultimately, these different input traits might converge at least in part on a common output path, which is cytokinin-controlled growth. The AHPs, which were shown to function as central hub proteins, as they are able to interact with members of all other protein groups within the TCS (Fig. 7), may act as signal integrators in this system. In conclusion, these data support the idea that expansion of an originally prokaryotic signaling system in a multicellular organism might have been functionally adapted to act as a signal collector, and provides only limited specification of downstream responses. This functioning would be consistent with the earlier proposed rheostat-like mechanism of cytokinin signaling [4,53].

Towards the uncovering of the mechanism of the interactions within the TCS of Arabidopsis thaliana

The mapping of the interaction domains of the TCS proteins of Arabidopsis revealed a high degree of similarity to those of the bacterial homologs [54–56]. Exceptions are the interactions of the B-type ARRs, which are mediated via the Myb domain. As this domain is not present in bacteria (Pfam database), homomerizations, if present in these organisms, must be mediated via a different protein domain. However, in higher organisms, the Myb domain is a well-characterized interaction domain, so it is not surprising that this domain also mediates homomerizations in the TCS of Arabidopsis[57,58]. As it was shown that B-type ARRs, at least ARR10, bind DNA as a monomer, it might be that dimerizations comprise a mechanism to regulate protein activity [59].

The underlying mechanism of the detected interactions was also investigated. When the canonical histidine of AHP5, which is relevant for its phosphotransfer function, was mutated, its ability to interact with other TCS proteins was not affected in almost all cases (Figs 5 and 6). Following the histidine mutation, only the interactions of AHP5 with AHK4 and ARR6 were lost in the yeast two-hybrid assay (Fig. 5) but reproduced by in vitro binding analysis (Fig. 6). It could be that the amino acid substitutions weaken these interactions and the yeast two-hybrid system is not sensitive enough to detect them. The in vitro binding assay seems to be more robust, thus enabling the detection of weak interactions even if the involved proteins have a slightly different conformation and/or charge distribution. In summary, these results indicate that the phosphorelay signaling function and the capability to interact with another protein of the TCS can be distinguished, raising the possibility that TCS proteins interact with other proteins irrespective of the phosphorylation status.

This result is especially interesting in the light of the negative feedback regulation of A-type ARRs on cytokinin signal transduction [16,17]. Our data indicate that feedback regulation via AHPs is a possibility, as they are able to interact with the A-type ARRs (Fig. 7). One possible mechanism for this regulation is that the interaction of A-type ARRs with AHPs prevents them from interacting with other proteins and thus interrupts the signaling chain. Another mechanism for such a regulation could be dephosphorylation of phosphorylated – and therefore activated − AHPs by A-type ARRs. The A-type ARRs could thus function as AHP-phosphatases, where one ARR molecule may dephosphorylate many AHPs [19,60]. A third route could be competition of the A-type ARRs with the B-type ARRs for the phosphoryl group of activated AHPs. In this model, the level of feedback inhibition would be proportional to the amount of A-type ARRs present in the cell. For some A-type ARRs, such a mechanism was demonstrated in a bacterial test system [60], whereas in planta experiments demonstrated that aberrant expression of individual A-type ARR genes resulted in no obvious phenotypic changes [51] or only in weakly reduced sensitivity to cytokinin [64]. This difference could be due to the high buffering capacity of the signaling system in planta, which makes it difficult to detect gradual differences.

In either case, interaction of A-type ARRs with AHPs would interfere with AHK-dependent activation of B-type ARRs and thus reduce cytokinin signaling. A role for direct interaction between the A-type and the B-type ARRs in a feedback mechanism appears to be less likely, considering the results of our interaction analysis.

Conclusions

  1. Top of page
  2. Abstract
  3. Results
  4. Discussion
  5. Conclusions
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information

Taken together, our findings have helped to complete the picture of an internally redundant cytokinin signaling network. The known expression domains of the corresponding genes in planta and the known subcellular localizations of the proteins in planta are compatible with the proposed interaction pattern. However, the implications of the protein–protein interaction analysis should be further investigated with more functional high-throughput studies.

Redundancy appears to be an important inherent property of the TCS. This could be a means to fine-tune the signaling system. The redundancy of the system raises the question how signal specificity can be achieved in the TCS to generate the many specific cytokinin responses. An important contribution to signal specification could come from specific interactions of TCS proteins with the rest of the Arabidopsis proteome. It was shown in yeast that differences in signaling output of the mitogen activated protein kinase (MAPK) pathway are achieved by different MAPK interaction partners, e.g. scaffolding proteins [61]. Our results from genome-wide interaction screens with numerous TCS proteins indicate that each of these proteins has a specific set of interaction partners, thus supporting this hypothesis (Dortay, Mehnert, Pfeifer, Schwerdtner, Liu, Schmülling & Heyl, unpublished results).

Experimental procedures

  1. Top of page
  2. Abstract
  3. Results
  4. Discussion
  5. Conclusions
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information

Cloning of cDNAs encoding TCS signaling components

The cDNAs encoding the cytokinin signaling components were amplified either by PCR using a cDNA library from Arabidopsis thaliana C24 [62] or by RT-PCR using total RNA extract from A. thaliana Col-0 primary leaves. Primers containing the Gateway™attB1 and attB2 sites or TOPO sites were used for recombination of isolated cDNAs into the Gateway™ Entry vectors pDONR201 and pDONR222 or pENTR/D-TOPO (Invitrogen, Carlsbad, CA, USA). For the primer sequences, see supplementary Table S2. Deletion constructs of TCS signaling components were amplified by PCR using the cloned full-length cDNAs in the Entry vectors. Primers containing attB1 and attB2 sites were used for recombination into pDONR222. For the primer sequences, see supplementary Table S3. The identities of all cloned cDNAs were verified by sequencing.

Plasmid constructions and yeast two-hybrid analysis

The yeast two-hybrid analysis was done by using a LexA DNA-binding domain encoding bait vector (pBTM116c-D9) and a Gal4 activation domain encoding prey vector (pACT2; Clontech, Mountain View, CA, USA) adapted to the GatewayTM system (Invitrogen) in the yeast strain L40ccαU (MATa his3Δ200 trp1-901 leu2-3 112 LYS::(lexAop)4-HIS3 URA3::(lexAop)8-lacZ, ADE2::(lexAop)8-URA3 GAL4 gal80 can1 cyh2) [35]. The cDNAs were in vitro recombined from the Gateway™ Entry vectors into the bait and prey vectors according to the instructions of the manufacturer. Based on the lithium acetate method, yeast transformations were performed as previously described [63]. Preselection of bait and prey hybrid proteins was done by testing for self-activation on interaction medium by coexpression of the bait hybrid proteins with the Gal4 activation domain encoded by the empty prey vector and by coexpression of the prey hybrid proteins with the LexA DNA-binding domain encoded by the empty bait vector. Weakly autoactivating hybrid proteins were totally suppressed by supplementing the interaction medium with 5 mm 3AT, whereas strong autoactivating hybrid proteins that could not be suppressed even in the presence of 20 mm 3AT were omitted from the matrix screen. All yeast clones were incubated for 7 days at 30 °C.

Expression of fusion proteins and in vitro interaction assay

For protein expression and purification, cDNAs were in vitro recombined into the Gateway vector pDEST15 (Invitrogen) encoding an N-terminal GST-tag, and transformed into the E. coli strain BL21(DE)pLysS (Stratagene, La Jolla, CA, USA). Expression of GST and GST fusion proteins was induced at 30 °C by 1 mm isopropyl thio-β-d-galactoside for 3 h. Harvested cells were lysed with GST-lysis buffer [20 mm sodium phosphate buffer, pH 7.3, 150 mm NaCl, 1 mm EDTA, 0.2% Triton X-100, 1 mm dithiothreitol, 1 mm phenylmethanesulfonyl fluoride, 10 µg·mL−1 aprotonin, 10 µg·mL−1 leupeptin, 2 mm benzamidin] and ultrasound treatment. The supernatant of centrifuged samples was used for determining concentrations of GST and GST fusion proteins on SDS/PAGE by using BSA standards. In vitro interaction assays were performed by using the MagneGST™ pull-down system (Promega, Madison, WI, USA) according to the instructions of the manufacturer. For this, 20 µL of magnetic glutathione agarose beads were pretreated according to the instructions of the manufacturer and incubated at room temperature for 30 min with 10% BSA. Equal amounts of GST and GST fusion proteins (200 ng) were purified on pretreated magnetic glutathione agarose beads in the presence of 0.5% Nonidet-P40 and 10% BSA by incubation at room temperature for 30 min. The respective cDNAs were shuttled into the His-tag containing Gateway™ vector pDEST17 (Invitrogen) and in vitro transcribed/translated in the presence of [35S]methionine. Five microliters out of 20 µL of immobilized GST and GST fusion proteins and in vitro synthesized radioactive protein were incubated at room temperature for 60 min and washed 10 times with 400 µL of washing buffer supplemented with 0.5% Nonidet-P40. Eluted proteins were separated on SDS/PAGE, and the dried gel was used for autoradiography.

Site-directed mutagenesis

For site-directed mutagenesis, the QuikChange Site Directed Mutagenesis Kit (Stratagene) was used according to the instructions of the manufacturer. Sequences of primers used for the mutagenesis of AHP5 in the entry vector were: H83K sense, 5′-CAGGTGGATTCAGGTGTTAAGCAACTCAAGGGTAGTAGC-3′; H83K antisense, 5′-GCTACTACCCTTGAGTTGCTTAACACCTGAATCCACCTG-3′; H83A sense, 5′-CAGGTGGATTCAGGTGTTGCCCAACTCAAGGGTAGTAGC-3′; H83A antisense, 5′-GCTACTACCCTTGAGTTGGGCAACACCTGAATCCACCTG-3′. Mutagenesis was verified by sequencing, and mutated AHP5 cDNA was in vitro recombined from the entry vector into the bait vector.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Results
  4. Discussion
  5. Conclusions
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information

We thank Professor Erich Wanker (MDC, Berlin) for the bait plasmid and the yeast strain. We also wish to thank Wolfgang Schuster, Hanjo Hellmann for technical help and Marion Amende for help with the cloning of many genes. We also thank Anahid Powell for critically reading the manuscript. This work was supported by the AFGN Program of the DFG and by a NaFöG stipend to HD.

References

  1. Top of page
  2. Abstract
  3. Results
  4. Discussion
  5. Conclusions
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. Results
  4. Discussion
  5. Conclusions
  6. Experimental procedures
  7. Acknowledgements
  8. References
  9. Supporting Information
FilenameFormatSizeDescription
FEBS_5467_sm_TableS1-S3.pdf53KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.