Abscisic acid-activated SNRK2 protein kinases function in the gene-regulation pathway of ABA signal transduction by phosphorylating ABA response element-binding factors


(fax +81 52 789 5215; e-mail hattori@agr.nagoya-u.ac.jp).


The plant hormone abscisic acid (ABA) induces gene expression via the ABA-response element (ABRE) present in the promoters of ABA-regulated genes. A group of bZIP proteins have been identified as ABRE-binding factors (ABFs) that activate transcription through this cis element. A rice ABF, TRAB1, has been shown to be activated via ABA-dependent phosphorylation. While a large number of signalling factors have been identified that are involved in stomatal regulation by ABA, relatively less is known about the ABA-signalling pathway that leads to gene expression. We have shown recently that three members of the rice SnRK2 protein kinase family, SAPK8, SAPK9 and SAPK10, are activated by ABA signal as well as by hyperosmotic stress. Here we show that transient overexpression in cultured cell protoplasts of these ABA-activated SnRK2 protein kinases leads to the activation of an ABRE-regulated promoter, suggesting that these kinases are involved in the gene-regulation pathway of ABA signalling. We further show several lines of evidence that these ABA-activated SnRK2 protein kinases directly phosphorylate TRAB1 in response to ABA. Kinetic analysis of SAPK10 activation and TRAB1 phosphorylation indicated that the latter immediately followed the former. TRAB1 was found to be phosphorylated not only in response to ABA, but also in response to hyperosmotic stress, which was interpreted as the consequence of phosphorylation of TRAB1 by hyperosmotically activated SAPKs. Physical interaction between TRAB1 and SAPK10 in vivo was demonstrated by a co-immunoprecipitation experiment. Finally, TRAB1 was phosphorylated in vitro by the ABA-activated SnRK2 protein kinases at Ser102, which is phosphorylated in vivo in response to ABA and is critical for the activation function.


The plant hormone abscisic acid (ABA1) plays important roles in various physiological and developmental processes. Seed maturation and dormancy, and adaptive responses to water deficit, are the two major processes involving ABA signalling (Bray, 1997; McCarty, 1995). ABA levels are elevated in response to developmental or environmental cues produced in such processes. Stomatal aperture is reduced by the rise in ABA levels on water deficit, to prevent further loss of water. Stomatal closure is mediated by turgor reduction in guard cells, which is caused by an efflux of K+ and anions from guard cells (MacRobbie, 1997). Studies on ABA signalling in guard cells have identified a number of signalling proteins (protein kinases and phosphatases, and G-proteins), and small molecules as second messengers (Ca2+, cADP-ribose, H2O2 and phosphoinositides) and proteins that generate them (DeWald et al., 2001; Munnik et al., 1998, 1999, 2000; Pical et al., 1999; Wang, 1999; Xiong et al., 2002b). It appears that these factors cannot be placed along a linear-signalling cascade, but rather the ABA-signalling system in guard cells constitutes a complex network which includes parallel pathways and both positive- and negative-feedback loops (Leonhardt et al., 2004). Recent studies indicate that the expression of many of these signalling factors themselves is up or downregulated by ABA (Leonhardt et al., 2004). The involvement of proteins related to RNA metabolism has also been implicated in ABA signalling in stomatal regulation, as well as in other responses (Hugouvieux et al., 2001; Koiwa et al., 2002; Kuhn and Schroeder, 2003; Lu and Fedoroff, 2000; McCourt, 2001; Ng et al., 2004; Xiong et al., 2002b). These observations further illustrate the complexity of ABA signal transduction. Increases in ABA level also lead to the expression of specific sets of genes, and modify cellular functions. Compared with ABA signalling in stomatal regulation, less information has been available on the ABA-signalling pathway leading to gene expression. Early ABA-signalling events, however, appear to be shared between stomatal regulation and gene-expression pathways because many mutations, including abi1-1, abh1, rcn1 and ost1/srk2e, have been shown to affect both ABA-induced stomatal behaviour and gene expression (Finkelstein et al., 2002; Hugouvieux et al., 2001; Kwak et al., 2002; Yoshida et al., 2002). It is intriguing to know where in the cascade the stomatal and gene-regulation pathways branch, or what signalling events are specific to the gene-regulation pathway. In contrast to the intermediary-signalling events, information has been accumulating about the terminal steps in the primary gene-regulation pathway. ABA-response elements (ABREs) responsible for primary ABA-induced transcription have been identified (Busk and Pages, 1998). These elements have a core consensus of ACGTG/TC (Hattori et al., 2002), and function in concert with another element, the coupling element (CE) (Hobo et al., 1999a; Shen and Ho, 1995; Shen et al., 1996). A group of bZIP transcription factors have been identified as the trans-acting factors that bind to ABRE and mediate ABA-induced transcriptional activation (Choi et al., 2000; Finkelstein and Lynch, 2000; Hobo et al., 1999b; Kang et al., 2002; Lopez-Molina and Chua, 2000; Uno et al., 2000). These ABRE-binding factors (ABFs) stimulate transcription in two ways. Firstly, ABFs are activated via ABA-dependent phosphorylation (Kagaya et al., 2002). Secondly, the protein levels of ABFs are elevated in response to ABA transcriptionally as well as post-translationally (Choi et al., 2000; Kang et al., 2002; Lopez-Molina et al., 2001). We have shown previously that a rice ABF, TRAB1, undergoes ABA-dependent phosphorylation at Ser102, which is required for the activation function of TRAB1 (Kagaya et al., 2002).

As in studies of other signalling systems, genetic approaches have been exploited to dissect ABA signal transduction. Most genes identified from recessive ABA-insensitive mutants encode transcription factors (Finkelstein et al., 2002). In addition to these transcription factors, more negative regulators than positive ones have been identified and cloned through genetic analyses (Finkelstein et al., 2002; Himmelbach et al., 2003). In other words, more ABA-hypersensitive recessive mutants have been characterized. OST1/SRK2E is one of a few examples of the genetically identified genes that encode positive regulators other than transcription factors (Mustilli et al., 2002). OST1/SRK2E was found to be a homologue of a Vicia faba ABA-activated protein kinase, AAPK, which had been shown to be involved in ABA signalling in guard cells (Li et al., 2002). These kinases are members of the SnRK2 protein kinase family (Hrabak et al., 2003). A wheat ABA-inducible protein kinase, PKABA1, which has been suggested to be involved in ABA suppression of gibberellic acid signalling in the aleurone layers, is also a member of the SnRK2 family (Gomez-Cadenas et al., 1999, 2001). Both Arabidopsis and rice genomes encode 10 members of the SnRK2 protein kinase family (Hrabak et al., 2003; Kobayashi et al., 2004). We have shown recently that all members of the rice SnRK2 family, designated SAPK1–10, are activated in response to hyperosmotic stress via phosphorylation by an unidentified upstream protein kinase, and that SAPK8, SAPK9 and SAPK10 are also activated by ABA (Kobayashi et al., 2004). Similar results have been reported for Arabidopsis SnRK2 (Boudsocq et al., 2004). From phylogenetic analysis, we have classified the SnRK2 members into three subgroups. SAPK8–10 constitute subgroup 3, to which AAPK and OST1/SRK2E also belong (Kobayashi et al., 2004). Yoshida et al. (2002) reported that the srk2e mutant is impaired not only in stomatal response, but also in ABA-induced gene expression.

Here we show that ABA-activated members of the rice SnRK2 family are involved in the gene-regulation pathway of ABA signalling, and provide evidence that these protein kinases phosphorylate TRAB1 directly to activate transcription in response to ABA.


ABA-activated rice SnRK2 protein kinases are involved in ABA signal transduction leading to gene induction

To confirm that the rice ABA-activated SnRK2 protein kinases function in the gene-regulation pathway of ABA signalling like OST1/SRK2E (Yoshida et al., 2002), we tested whether overexpression of these kinases affects the expression of an ABA-regulated promoter. SAPK10 was transiently overexpressed in rice cultured cell protoplasts together with the 55-bp TATA–GUS reporter gene, in which GUS-reporter expression was driven by a chimera promoter containing a CaMV 35S minimal promoter and two copies of ABREs derived from the Osem promoter (Hobo et al., 1999a) (Figure 1a). SAPKs expressed in protoplasts were in the double-HA epitope and His-affinity (dHA-His)-tagged forms (Kobayashi et al., 2004) in all the experiments described here. Co-expression of SAPK10 resulted in a roughly twofold stimulation of ABA-induced expression of the reporter gene (Figure 1b). In the absence of ABA, fivefold activation of the reporter gene by SAPK10 was observed (Figure 1b). The reporter gene activation in the absence of ABA probably depended on the overexpressed SAPK10, with residual activities in an unactivated state or a small proportion of the active enzyme. SAPK8 and SAPK9 were also able to activate the reporter gene in the absence of ABA (Figure 1c). No activation was observed for a control reporter gene, which lacked ABREs, or by an SnRK3 protein kinase (data not shown). These results suggest that all these ABA-activated SnRK2 members can function in ABA signal transduction to activate ABRE-regulated promoters. The level of reporter activation was highest with SAPK9 and lowest with SAPK8. The weaker reporter activation by SAPK8 can be attributed to the low levels of SAPK8 protein expression (Kobayashi et al., 2004). It is less likely that the difference in activation levels between SAPK9 and SAPK10 could be because of the difference in substrate specificities, considering the high levels of sequence conservation among these subclass 3 ABA-activated kinases. It could rather be because of the difference in the basal activities of the unactivated enzymes.

Figure 1.

Activation of ABRE by ABA-activated SnRK2 protein kinases.
(a) Schematic illustration of the 55-bp TATA–GUS reporter gene.
(b) An expression plasmid for dHA-His-tagged SAPK10 or an empty vector (p35S-Sh-dHA/His) were transfected together with the 55-bp TATA–GUS reporter plasmid into protoplasts of rice suspension-cultured cells by electroporation. Each transfection also included 5 μg of a ubiquitin promoter::luciferase plasmid as an internal standard. After culturing electroporated cells for 36 h in the presence of ABA at the concentrations indicated, GUS and luciferase activities were determined. GUS activities were normalized by luciferase activity. Each bar represents the mean of the activities obtained from three independent electroporations with standard deviation. Inset, values at 0 μm ABA.
(c) Activation of the 55-bp TATA–GUS reporter gene by SAPK8, SAPK9 or SAPK10 was tested as in (b) in the absence of ABA.

Kinetics of ABA-induced activation of SAPK10 and Ser102-phosphorylation of TRAB1 support direct phosphorylation of ABFs by SnRK2 protein kinases

Because ABA-activated SnRK2 protein kinases were implicated in the gene-regulation pathway of ABA signal transduction, the possibility was tested that ABA-activated SnRK2 protein kinases directly phosphorylate a rice ABF, TRAB1, which had been shown to be activated by phosphorylation in response to ABA signals. Firstly, the kinetics of SAPK activation and TRAB1 phosphorylation were compared (Figure 2). Unless noted, TRAB1 expressed in protoplasts was also a dHA-His-tagged form in all the experiments.

Figure 2.

Kinetics of SAPK10 activation and TRAB1 phosphorylation.
(a) Rice cultured-cell protoplasts expressing dHA-His-tagged SAPK10 were treated with 50 μm ABA for the time indicated. Tagged SAPK10 was recovered from the cell extracts by Ni-agarose resin, and subjected to the in-gel kinase assay using MBP as a substrate (in-gel assay), or to immunoblot analysis with anti-HA antibody (Immunoblot).
(b) Rice cultured-cell protoplasts expressing dHA-His-tagged TRAB1 were treated with 50 μm ABA for the time indicated. The tagged protein was recovered and subjected to immunoblot analysis as in (a). Arrow indicates position of the S-band corresponding to the Ser102-phosphorylated form.
(c) SAPK10 activities (closed squares) and TRAB1 phosphorylation levels (open triangles) were quantified using the radioactive and luminescence images obtained in the experiments described in (a,b), respectively. SAPK10 activities were normalized by immunoblot signals. Changes in the levels of Ser102-phosphorylation in TRAB1 were expressed as the increase in S-band intensity/total band intensity at each time point relative to the 0-min value.

Activation of SAPK10 derived from the transfected expression plasmid was first observed at 1 min after ABA treatment, although the level was very low (Figure 2a,c). The activity then increased linearly up to 15 min and remained constant thereafter. Previously, we have shown that TRAB1 becomes active via phosphorylation at Ser102 (Kagaya et al., 2002), which can be monitored by the increase in intensity of the slow migration band (S-band) of the triplet band observed in the immunoblot analysis. The increase in the Ser102 phosphorylation levels of TRAB1 was not obvious at 1 min, but was first observed 5 min after ABA treatment (Figure 2b,c). The phosphorylation level continued to increase up to 30 min with some decline in the rate of increase after 15 min. Although it was difficult to determine whether there was a very small increase in the TRAB1 phosphorylation level at 1 min because of the relatively high basal levels of phosphorylation, the kinetic patterns indicate that the increase in TRAB1 phosphorylation immediately follows SAPK10 activation. These results are consistent with the hypothesis that SAPK10 and other ABA-activated SnRK2 protein kinases phosphorylate TRAB1, and possibly other ABFs as well, in response to ABA in vivo.

TRAB1 is phosphorylated in response to hyperosmotic stress

We reported previously that the rice ABA-activated SnRK2 protein kinases are also rapidly activated by hyperosmotic stress, like other SnRK2 family members (Kobayashi et al., 2004). Thus the induction of TRAB1 phosphorylation in response to hyperosmotic treatment would be expected if the ABA-activated SnRK2 protein kinases are responsible for TRAB1 phosphorylation. This possibility was tested with protoplasts transfected with the TRAB1-expression plasmid. As expected, the intensity of the TRAB1 S-band was increased by either 250 or 400 mm NaCl treatment (Figure 3a). Absence of the S-band in NaCl-treated cells expressing the S102A mutant of TRAB1, as in ABA-treated cells, confirmed that the hyperosmotic stress-induced increases in the S-band were because of Ser102 phosphorylation (Figure 3a). The proportion of the Ser102 phosphorylated form of TRAB1 (S-band) was increased slightly at 1 min after hyperosmotic treatment, then at a high level at 5 min, and declined thereafter (Figure 3b). The kinetics of Ser102 phosphorylation being faster in response to hyperosmotic stress than in response to ABA was consistent with the previously observed faster activation of SAPKs in response to hyperosmotic stress (Kobayashi et al., 2004). These results further support the in vivo phosphorylation of TRAB1 by SAPK10. It should be noted that the increase in the Ser102 phosphorylated form in response to hyperosmotic stress was rather transient, peaking at 5 min, while the levels of the phosphorylated form increased continuously up to 30 min when stimulated by ABA.

Figure 3.

Phosphorylation of TRAB1 in response to hyperosmotic stress.
(a) Rice cultured-cell protoplasts expressing the wild-type (WT) or S102A mutant (S102A) forms of dHA-His-tagged TRAB1 were treated with 50 μm ABA, or 250 or 400 mm NaCl for 15 min, or given control treatment (cont). The tagged protein was recovered and subjected to immunoblot analysis as in Figure 2. Arrows indicate S-bands representing Ser102-phosphorylated forms of TRAB1.
(b) Rice cultured-cell protoplasts expressing WT dHA-His-tagged TRAB1 were treated with 400 mm NaCl for the time indicated, and analysed as in (a).

SAPK10 physically interact with TRAB1 in vivo

To further confirm the direct action of SnRK2 protein kinases on ABFs, physical interaction between TRAB1 and SAPK10 was tested by co-immunoprecipitation experiments. Extracts were prepared from rice protoplasts expressing both Myc-tagged TRAB1 (TRAB1-myc) and dHA-His-tagged SAPK10 (SAPK10-dHA-His), or either alone, and subjected to immunoprecipitation with anti-myc antibody–agarose beads. Immunoblot analysis of the immunoprecipitate with anti-myc antibody revealed that SAPK10-dHA-His was co-precipitated from the extracts of the protoplasts expressing both TRAB1-myc and SAPK10-dHA-His, but not from those expressing SAPK10-dHA-His alone (Figure 4). These results indicate that SAPK10 and TRAB1 form a complex in vivo, again supporting the direct phosphorylation of ABFs by ABA-activated SnRK2 protein kinases.

Figure 4.

Co-immunoprecipitation of TRAB1 and SAPK10.
Extracts were prepared from rice cultured-cell protoplasts that had been transfected with an expression plasmid for Myc-tagged TRAB1 (TRAB1-myc) or dHA-His-tagged SAPK10 (SAPK10-dHA-His) alone, or with both. After culturing transfected protoplasts, cell extracts were prepared and subjected to immunoprecipitation with anti-myc antibody–agarose beads. The immunoprecipitates were subjected to immunoblot analysis with anti-HA antibody. Inclusion or exclusion of the plasmid indicated in the transfection is shown by (+) or (−), respectively.

ABA-activated SnRK2 protein kinases phosphorylate TRAB1 in vitro

All the experiments described above supported the involvement of ABA-activated SnRK2 protein kinases in the gene-regulation pathway of ABA signal transduction and direct phosphorylation of ABFs (TRAB1) by these kinases. The ability of the ABA-activated SnRK2 protein kinases to phosphorylate TRAB1 was then tested in vitro, using a recombinant thioredoxin (TRX)-TRAB1 fusion protein expressed in Escherichia coli for a substrate, and SAPK10 expressed in and purified from protoplasts. Because it was difficult to produce a full-length TRAB1 fusion protein in E. coli of sufficient quantity and purity, a fusion protein with an N-terminal fragment (amino acids 57–171) was used for the substrate (Figure 5a). A further truncated form (57–90) and the S102A mutant forms of the fusion protein were produced, as well as TRX itself.

Figure 5.

Phosphorylation of recombinant TRAB1 by ABA-activated SnRK2 protein kinases.
(a) Schematic illustration of the recombinant protein structure consisting of TRX and a TRAB fragment. The fusion proteins with WT and carboxyl-terminally truncated (DC) forms contained a TRAB1 fragment with amino acids 57–171 and 57–90, respectively. Shaded areas indicate conserved sequence blocks identified previously (Hobo et al., 1999b).
(b) ABA-activated SAPK8, SAPK9 and SAPK10 were obtained as in Figure 2(a) and subjected to the in-gel kinase assay using MBP or recombinant TRX-TRAB1 (57–171) as indicated.
(c) The affinity-recovered ABA-activated SAPK8, SAPK9 or SAPK10 was subjected to immunoprecipitation with anti-HA antibody. The immunocomplex on protein-G agarose beads were used for the in vitro kinase assay with the substrate indicated. A control reaction was performed with immunoprecipitates obtained in the same way as in the other reaction, except that cell extracts were prepared from protoplasts transfected with the empty vector. The autoradiogram (upper panel) and the CBB-stained image (lower panel) of the SDS–polyacrylamide gel after separation of the reaction products are shown. Asterisks indicate positions of the substrate/product bands. Arrow, position of the immunoglobulin light chain.

When the activities of hyperosmotic or ABA-activated forms of SAPK8, 9 or 10 were analysed by the in-gel kinase assay, these kinases phosphorylated TRX-TRAB1 as efficiently as the standard substrate myelin basic protein (MBP) (Figure 5b). Phosphorylation of TRAB1 by the ABA-activated SAPK8, 9 or 10 was also detected by the immunocomplex kinase assay (Figure 5c). For this assay, ABA-activated SAPKs recovered via the His-affinity tag were immunoprecipitated further with anti-HA antibody. Phosphorylation by these kinases was undetectable when the further truncated form [TRX-TRAB1 (57–90)] or the unfused TRX form was utilized as a substrate (Figure 5c). These results indicated that SAPK8–10 can phosphorylate a specific region (amino acids 91–171) of TRAB1.

ABA-activated SnRK2 protein kinases phosphorylate Ser102 of TRAB1 in vitro

Because the region of TRAB1 phosphorylated by SAPKs contained Ser102, which is phosphorylated in response to ABA in vivo, the possibility was tested that these kinases phosphorylate Ser102 of TRAB1 by comparing the activity against the wild-type (WT) and the S102A mutant form of TRX-TRAB1 (57–171) by the immunocomplex kinase assay (Figure 6a). Phosphorylation efficiency against the S102A mutant substrate was found to be approximately 50–70% of that against the WT substrate (Figure 6b). This result supported the phosphorylation at Ser102, and also indicated that SAPKs phosphorylate other sites as well. It should be noted that mobility differences were observed between the 32P-labelled product bands derived from the wild type and S102 mutant substrate, whereas the motilities of the unlabelled substrates were not different between wild type and mutant (Figure 6a). The difference was most remarkable in the reaction products of SAPK10, which exhibited the highest expression and activity. When the SAPK10 reaction was carried out with unlabelled ATP, the Coomassie-stained product bands derived from the WT substrate gave two major bands, both of which exhibited mobilities slower than the substrate, and an additional minor band with a further slower mobility (Figure 6c). These mobility shifts were considered to be because of phosphorylation. In contrast, the same reaction with the S102A mutant substrate resulted in the production of a single major band with a mobility that corresponded to that of the faster migrating major product from the WT substrate (Figure 6c). The lack of the slower migrating product band, when the S102 mutant substrate was used, was probably because of the decrease in the number of phosphorylation sites, which again supported Ser102 phosphorylation by SAPK10.

Figure 6.

Identification of serine/threonine residues phosphorylated by SAPK10 in vitro.
(a) WT and S102A mutant (S102A) forms of recombinant TRX-TRAB1 (57–171) separated on an SDS–polyacrylamide gel and stained with Coomassie blue (CBB).
(b) Immunocomplex kinase assay was performed with ABA-activated SAPK8, SAPK9 or SAPK10 as in Figure 5(c) using the WT or S102A mutant (S102A) form of recombinant TRX-TRAB1 (57–171) shown in (a) as a substrate.
(c) Immunocomplex kinase assay with (+) or without (−) ABA-activated SAPK10 as in (b), except that unlabelled ATP was used. The reaction products were separated by SDS–PAGE and stained with CBB. Band patterns are illustrated schematically at the bottom. Asterisks indicate the band subjected to mass spectrometric analysis.
(d) MALDI-TOF mass spectra showing the ion peaks corresponding to the calculated masses of singly phosphorylated forms of tryptic fragments 1 and 2 of TRAB1, shown in (e).

Additional evidence for this conclusion was obtained by mass spectrometric analysis. The slower migrating product band from the products generated from the WT substrate shown in Figure 6c was excised, in-gel digested with trypsin, and subjected to matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometric analysis. The analysis detected two mass values, 951.6 and 1513.1, which correspond to a singly phosphorylated fragment 92QGSLTLPR (theoretical mass = 951.50) and 100TLSVKTVDEVWR (theoretical mass = 1512.78), respectively (Figure 6d,e). The latter peptide contained Ser102. No other mass peaks corresponding to possible phosphorylated tryptic fragments were predicted from the amino acid sequence.


In the present study we showed that transient overexpression of ABA-activated SnRK2 protein kinases, SAPK8, SAPK9 and SAPK10, resulted in the enhancement of transcription from ABA-responsive promoters. We also demonstrated that these protein kinases phosphorylate an ABRE-binding factor, TRAB1, at specific serine/threonine residues in vitro. One of the TRAB1 SAPK-phosphorylated residues, Ser102, is phosphorylated in vivo in response to ABA, and is critical to TRAB1 activation. Furthermore, the activation of SAPK8, 9 and 10 slightly preceded TRAB1 phosphorylation in response to ABA. These data strongly indicate that the ABA-activated SnRK2 members are involved in the gene-expression pathway of ABA-signal transduction, and directly phosphorylate ABFs to activate ABA-responsive genes. Thus phosphorylation of TRAB1 by SAPKs can be considered as the final step in the gene-regulation pathway of the primary ABA signal transduction if a simple linear cascade is assumed (see below). In addition, ABA-activated SnRK2 protein kinases appear as the branching node of the stomatal regulation pathway and the gene-regulation pathway, as these kinases have been shown to function in the regulation of the stomatal aperture.

Two reports describe ABA-activated ABF kinase activities in Arabidopsis. Uno et al. (2000) detected a 42-kDa protein kinase activity that phosphorylated an N-terminal fragment of 59 or 50 amino acids from AREB1 or AREB2, respectively, but not the fragments of the bZIP regions, in the extract from T87 cultured cells. This kinase activity became detectable within 2 min after ABA treatment (Uno et al., 2000). Because one of the 42-kDa ABA-activated protein kinases detected with the in-gel kinase assay using MBP substrate has been identified as SRK2E/OST1, the AREB kinase activity is likely to be that of OST1/SnRK2E-related kinases (Mustilli et al., 2002; Yoshida et al., 2002). Furthermore, the N-terminal fragment of AREB1 or AREB2 phosphorylated by this kinase activity contains the serine residues corresponding to those residues identified or proposed to be phosphorylated by SAPK10 in this study (Uno et al., 2000). Therefore these results are consistent with our present findings.

Lu et al. (2002) also detected 42- and 46-kDa kinase activities that were rapidly activated in response to ABA by in-gel kinase assays, using recombinant ABI5 or MBP as a substrate. However, they suggested the 42-kDa activity to be that of a MAP kinase, AtMPK3, based on several observations. Unfortunately, ABI5 kinase activity was not tested with recombinant AtMPK3 expressed in plants, therefore its identity as AtMPK3 does not appear conclusive. Nevertheless, their data demonstrated the importance of AtMPK3 for ABI5 functioning during early seedling growth. If AtMPK3 is responsible for ABA-induced phosphorylation of ABI5, its role could be different from the phosphorylation of other ABFs by SnRK2 protein kinases. ABI5 has been shown to be regulated by a proteasome-mediated process, and to have special developmental and physiological functions at very early seedling stages (Lopez-Molina et al., 2001). ABI5 phosphorylation by AtMPK3 might be involved in such a regulatory process.

Johnson et al. (2002) have recently reported that wheat PKABA1, which is orthologous to rice SAPK1, interacts with a wheat ABRE-binding bZIP factor family protein, TaABF, in the yeast two-hybrid assay. As this interaction was abolished by deletion of the ATP-binding site from PKABA1, it was suggested that TaABF1 could be an in vivo substrate of PKABA1. Although this report appears to be in line with our results on one hand, it is controversial in several aspects. PKABA1 expressed and recovered from insect cells was shown to phosphorylate oligopeptides with sequences derived from TaABF in vitro. Although the data appear suggestive of TaABF as a substrate of PKABA1, this was not conclusive. The sequence of the TaABF peptide that they determined to be specifically phosphorylated by PKABA1 did not contain serine residue equivalent to either Ser86 or Ser102. Furthermore, the rice orthologue SAPK1, as well as other rice SAPKs expressed in plant cells under unstressed conditions, is inactive. Unless PKABA1 was misactivated in insect cells, the kinase is considered to be in an unactivated state. We reported that SAPK1 is not activated by ABA signal (Kobayashi et al., 2004), and thus is expected to be the orthologous PKABA1. How PKABA1 or SAPK1 participates in ABA-responsive transcription via ABFs remains to be solved.

In addition to the functionally critical Ser102 of TRAB1, the ABA-activated form of SAPK10 phosphorylated a nearby serine or threonine residue in peptide QGS94LT96LPR. Although the possibility cannot be excluded that MALDI-TOF mass spectrometric analysis failed to detect the peptide in which both Ser94 and Thr96 were phosphorylated, Ser94 is more likely to be phosphorylated by SAPK10, considering the similarity in the flanking amino acid sequences of Ser94 and Ser102. RTLSV, containing Ser102, has similarity to the RQGSL motif, with a consensus sequence of R-(hydrophilic)-(hydrophobic)-S-(hydrophobic). In fact, such a sequence motif is found repeatedly in MBP and histone H3, both of which are good in vitro SAPK substrates. Kelner et al. (2004) recently suggested a substrate sequence of a tobacco SnRK2 protein kinase NtOSAK as [KR]-[QMTAS]-X-[ST]-[VILMF]-[SQN]-[FLIRK], using the computer program predikin, which was developed for prediction of the substrate specificity of protein Ser/Thr kinases (Brinkworth et al., 2003) (http://smms.uq.edu.au/kinsub). The program predikin predicted the same sequence for SAPK10. This predicted sequence has strong similarity to the sequence surrounding Ser102 and Ser94. Although the Ser94 flanking sequence has more homology than Ser102 to the prediction, the flanking sequences of the serine residues corresponding to Ser102 in other ABFs show similar levels of homology in most cases. Unlike Ser102, however, mutating Ser94 to an alanine did not lead to a loss of the activation function of TRAB1 (Kagaya et al., 2002). The functional requirement of Ser102 or Ser94 was determined with an N-terminally truncated form of TRAB1 (Kagaya et al., 2002). The RQGS94L sequence or similar Ser-containing pentapeptide sequences are found twice in the truncated region. The motif is conserved in other ABFs, and repeated more in some cases (Choi et al., 2000; Lopez-Molina and Chua, 2000; Uno et al., 2000). While the analyses with the truncated form might have assisted in finding the important residue separate from putative redundant Ser/Thr residues present in the truncated portion, it could have failed to detect other important phosphorylated Ser/Thr residues. We suspect that phosphorylation of Ser94 may contribute to the full activation of TRAB1 together with other serine residues in the RQGSL motif, which could be redundantly functional in vivo in the full-length context.

Hyperosmotic stress or drought-induced activation of some ABA-induced genes have been considered to be mediated by the increase in ABA levels because induction is strongly impaired in ABA-deficient mutants (Lang and Palva, 1992; Lang et al., 1994; Xiong et al., 2002a). However, the induction of some other genes has been reported to be inhibited modestly or only partially in ABA-deficient mutants (Lang et al., 1994; Welin et al., 1994; Xiong et al., 2002a; Yoshiba et al., 1999). The simplest interpretation for the partial impairment is that the ABA-independent-signalling systems are operating for the hyperosmotic stress/drought induction of these genes, such as that mediated by drought-response elements (Shinozaki and Yamaguchi-Shinozaki, 2000). From these views, the SnRK2-mediated rapid activation of an ABF in response to hyperosmotic stress revealed in our study is somewhat puzzling. However, the direct activation of transcription via ABREs by hyperosmotic signals could be significant at an early phase of response, which would become less significant or negligible after the ABA levels have increased. This idea is consistent with the observation that the hyperosmotically induced Ser102 phosphorylation was rapid and somewhat transient.

SAPK10 activation was detectable within 1 min of ABA treatment, soon after which TRAB1 phosphorylation was detectable. Because activation of SAPKs by ABA is suggested to occur by phosphorylation (Kobayashi et al., 2004), activation of the upstream SAPK kinase should happen earlier. Within such a short period, not many signalling events would be expected to occur. It is known that ABA signalling in stomatal regulation involves H2O2 as an intermediate messenger (Pei et al., 2000). Because ost1 mutation did not affect H2O2-induced stomatal closure, OST1 is proposed to function upstream of H2O2 generation (Mustilli et al., 2002). If so, it is suggested that H2O2 is not involved in the initial pathway to ABA-induced transcription.

Within 5 min of ABA treatment, TRAB1 phosphorylation was observed, which is expected to be immediately followed by the activation of ABRE-regulated genes. Such signalling events probably represent part of the most direct cascade in the gene-regulation pathways of ABA signalling. Global ABA-induced gene expression appears to involve induction of signalling factors which include ABFs (Choi et al., 2000; Finkelstein and Lynch, 2000; Kang et al., 2002; Uno et al., 2000). ABA also induces the expression of negative regulators of ABA signalling, such as ABI1/2 family protein phosphatases 2C (PP2C) (Leung et al., 1997; Pedro et al., 1998; Saez et al., 2004). Genetic analyses have revealed other negative as well as positive regulators that control or modulate ABA signalling, not only in the stomata but also in the gene-regulation pathway (Busk and Pages, 1998; Finkelstein et al., 2002). Such factors include those that function in RNA metabolism (Himmelbach et al., 2003; Kuhn and Schroeder, 2003). A recent study indicates the importance of ABA-induced gene expression, even in the regulation of stomatal openings (Leonhardt et al., 2004). Therefore, after initial delivery of ABA signal to the transcription system via the phosphorylation of ABFs by ABA-activated SnRK2 protein kinases, the signalling system would be remodelled into further complex networks. In other words, a signalling pathway would change with time. The roles of SnRK2 protein kinases appear more complex. For example, a V. faba ABA-activated SnRK2 protein kinase, AAPK, has been shown to interact with and phosphorylate a nuclear speckle-localized RNA-binding protein, AKIP1, which has the ability to bind transcripts of an ABA-induced gene, dehydrin (Li et al., 2002; Ng et al., 2004). Therefore, SnRK2 protein kinase may also be involved in gene regulation at a post-transcriptional level. Nevertheless, our study here, which revealed one of the critical events in ABA signalling that leads to gene expression, provides an important clue to the overall understanding of the complex ABA-signalling networks.

Experimental procedures

Plasmid construction

p35S-ShΔ-TRAB1-dHA/His (Kagaya et al., 2002) was digested with EcoRI and NotI. The resulting insert was ligated into the corresponding sites of pET32a(+) (Novagen, Madison, WI, USA) to produce PET32a(+)–TRAB1 (57–171), which allowed the expression of TRX-TRAB1 (57–171) fusion protein in E. coli. Similarly, the expression plasmid of the S102A mutant version of TRX-TRAB1 (57–171) fusion protein was constructed using the S102A mutant of p35S-ShΔ-TRAB1-dHA/His (Kagaya et al., 2002). To produce PET32a(+)–TRAB1 (57–90), the expression plasmid for TRX-TRAB1 (57–90), PET32-a(+)–TRAB1 (57–171), was digested with PstI and NotI, treated with T4 DNA polymerase to generate blunt ends, and self-ligated.

To produce p35S-ShΔ-TRAB1-myc for the expression of myc-tagged TRAB1 (TRAB1-myc) in protoplasts, the EcoRISalI TRAB1 fragment from p35S-ShΔ-TRAB1-dHA/His was cloned into the corresponding sites of p35S-ShΔ-myc-stop, which had been constructed by replacing the dHA-His tag sequence of p35S-ShΔ-dHA/His-stop (Kagaya et al., 2002) with a double-stranded synthetic oligonucleotide encoding a double myc-tag sequence (top strand, 5′-GAACAAAAATTAATTTCTGAAGAAGATTTAGGTGAACAAAAATTAATTTCTGAAGAAGATTTAAGATCTATGAATCGTAGATAC-3′; corresponding amino acid sequence, EQKLISEEDLGEQKLISEEDLRSMNRRY).

Transient expression and analysis of expressed protein

Plasmids were transfected into protoplasts prepared from rice-cultured cells (line Oc) by electroporation, as described previously (Kagaya et al., 2002). Co-transfection experiments using the 55-bp TATA–GUS reporter plasmid were performed as described previously (Hobo et al., 1999b).The expression plasmids used were as described by Kagaya et al. (2002); Kobayashi et al. (2004), or as described above. For analysis of protein kinases, transfected protoplasts were cultured, treated and extracted as described previously (Kobayashi et al., 2004), or as follows. For immunoprecipitation or detection of TRAB1-dHA-His by immunoblot analysis with anti-HA antibody (Kagaya et al., 2002), cultured and treated protoplasts were recovered by centrifugation at 8000 g for 30 sec and extracted with an appropriate buffer.

Affinity recovery with Ni-ATA-agarose and the following in-gel protein kinase assay or immunoblot analysis of the expressed dHA-His-tagged SAPKs were performed as described previously (Kobayashi et al., 2004). For the immunoprecipitate kinase assay, the affinity-recovered protein solution (10 μl) was diluted with 90 μl antibody-binding buffer (25 mm Tris–HCl pH 7.5, 150 mm NaCl, 0.1% Tween 20), pretreated with 5 μl (beads volume) protein-G-agarose (ImmunoPure Plus Immobilized Protein G; Pierce, Rockford, IL, USA) at 4°C for 30 min, and subjected to centrifugation to remove the beads. Following the pretreatment, the solution was incubated with 10 μl (beads volume) protein-G-agarose and 5 μl anti-HA antibody (Covance, Richmond, CA, USA) at 4°C for 2 h with continuous gentle shaking. After incubation with the antibody, the protein-G-agarose beads were washed with a washing buffer (50 mm Tris–HCl pH 7.5, 250 mm NaCl, 5 mm EDTA, 0.1% Tween 20) three times, then washed once with a kinase assay buffer (50 mm Tris–HCl pH 7.5, 1 mm DTT, 10 mm MgCl2, 0.1 mm ATP), and incubated with 10 μl of the kinase assay buffer, 2 μg substrate protein (1 μl), and 1 μCi [γ-32P] ATP at 37°C for 2 h. After the reaction, an appropriate amount of SDS–PAGE sample buffer was added to the reaction mixture and resolved by SDS–PAGE.

Mass spectrometry

Coomassie blue (CBB)-stained bands were excised with a razor blade, rinsed in water, and subjected to in-gel tryptic digestion. Peptide digests were concentrated using ZipTip C18 micropipette tips (Millipore, Bedford, MA, USA) and analysed using an AXIMA-CFR mass spectrometer (Shimadzu Biotech, Kyoto, Japan). The analysis was performed as a custom service provided by Shimadzu Biotech.


We thank Ms Tomiko Chikada and Yoko Tomita, and Dr Hirokazu Kato for excellent technical assistance. This work was funded in part by Grants-in-Aid for Scientific Research on Priority Areas (Grant 12138204) and the Center of Excellence from the Japanese Ministry of Education, Culture, Sports, Science and Technology, and by grants from the Ministry of Agriculture, Forestry and Fisheries of Japan (Rice Genome Project IP-5002) and the Japan Society for the Promotion of Sciences (Research for the Future Grant JSPS-ooL1603).