Tight Interconnection and Multi-Level Control of Arabidopsis MYB44 in MAPK Cascade Signalling

Abiotic stress poses a huge, ever-increasing problem to plants and agriculture. The dissection of signalling pathways mediating stress tolerance is a prerequisite to develop more resistant plant species. Mitogen-activated protein kinase (MAPK) cascades are universal signalling modules. In Arabidopsis, the MAPK MPK3 and its upstream regulator MAPK kinase MKK4 initiate the adaptation response to numerous abiotic and biotic stresses. Yet, molecular steps directly linked with MKK4 – MPK3 activation are largely unknown. Starting with a yeast-two-hybrid screen for interacting partners of MKK4, we identified a transcription factor, MYB44. MYB44 is controlled at multiple levels by and strongly inter-connected with MAPK signalling. As we had shown earlier, stress-induced expression of the MYB44 gene is regulated by a MPK3-targeted bZIP transcription factor VIP1. At the protein level, MYB44 interacts with MPK3 in vivo. MYB44 is phosphorylated by MPK3 in vitro at a single residue, Ser145. Although replacement of Ser145 by a non-phosphorylatable (S145A) or phosphomimetic (S145D) residue did not alter MYB44 subcellular localisation, dimerization behaviour nor DNA-binding characteristics, abiotic stress tolerance tests in stable transgenic Arabidopsis plants clearly related S145 phosphorylation to MYB44 function: Compared to Arabidopsis wild type plants, MYB44 overexpressing lines exhibit an enhanced tolerance to osmotic stress and are slightly more sensitive to abscisic acid. Interestingly, overexpression of the S145A variant revealed that impaired phosphorylation does not render the MYB44 protein non-functional. Instead, S145A lines are highly sensitive to abiotic stress, and thereby remarkably similar to mpk3-deficient plants. Its in vivo interaction with the nuclear sub-pools of both MPK3 and MKK4 renders MYB44 the first plant transcription factor to have a second function as putative MAPK cascade scaffolding protein.


Introduction
Environmental stresses represent the most limiting factors for agricultural productivity worldwide. Climate change and the unpredictable onset of abiotic or biotic adversities are a major challenge to any living species. As plants cannot escape environmental threats, they need to adapt at the cellular and physiological level to withstand unfavourable cues. A major mechanism linking environmental stress perception to cellular responses involves signalling through Mitogen-activated protein kinase (MAPK) cascades [1,2].
MAPK cascades are eukaryotic signalling modules that consist as a minimum of a MAPK kinase kinase (MAPKKK), a MAPKK and a MAPK, which via a phospho-relay system serve both signal transduction and amplification. By altering phosphorylationdependent properties of their target proteins, activated MAPKs translate the information further, eventually leading to e.g. changes in gene expression, cellular redox state or cell integrity. In contrast to animals, where MAPK cascades have been studied extensively and several MAPK substrates been identified (reviewed in [3]), comparatively little is known about individual MAPK cascades and their direct targets in plants [4,5].

combinations and the role of scaffolding proteins
The Arabidopsis thaliana genome encodes for 60 MAPKKKs, 10 MAPKKs and 20 MAPKs [6]. Similar family sizes are found in other plant species. Specificity, i.e. which of the theoretically possible 60620610 = 1200 module combinations is formed and integer under a certain condition, can be conferred to by specific expression or subcellular localisation of MAPK module components, or through the action of scaffold proteins, shared docking domains, and adaptor or anchoring proteins [7][8][9]. Scaffolding proteins have been shown to prevent cross talks between MAPK pathways employing a common MAPKKK in yeast, and also in mammalian cells [10,11].

Transcription factors
MAPK-mediated stress responses to environmental cues will ultimately involve major changes at the transcript, protein and metabolic level. Since MAPK(KK)s themselves normally have no DNA-binding activity, re-programming of gene expression can be accomplished through phosphorylation of transcription factors, thereby altering protein properties such as subcellular localisation, protein stability, homo/heterodimerisation, trans-activating capacity, target DNA motif affinity and -specificity [18][19][20][21][22].
A small number of transcription factors associated with MAPK signalling have been identified and functionally characterised, including primarily members of the WRKY family [23,24]. MPK3 and MPK6 enhance the stability of AtEIN3 [25] and the DNAbinding affinity of AtERF104 [26], two transcription factors involved in ethylene signalling. Phosphorylation by MPK3 regulates the subcellular localisation of the bZIP protein VIP1 in a stress-dependent manner [18].
MYB proteins represent another large family of plant transcription factors, comprising over 120 members in Arabidopsis [27]. MYB proteins regulate transcription by binding to their target motifs as MYB/MYB homo-or heterodimers. Examples also exist for regulatory complexes formed by MYB proteins and other types of transcription factors [27,28]. Members of the MYB family have been characterised to various extent in a diversity of species, and a role in germination, development, and responses to stress and hormone treatment could be [27,28]. Interspecies comparisons revealed a good correlation between sequence homology and function in some cases. For example, involvement of MYBs in the regulation of anthocyan synthesis appears to be restricted to a fraction of closely related family members [27,29]. However, MYB function in stress responses cannot be ascribed to a particular subfamily [27,29]. PtMYB4, a MYB protein with a putative function in xylem development, is phosphorylated by PtMPK6 in vitro, and studies in yeast suggest that phosphorylation promotes its trans-activating capacity [21]. Although several MYB proteins were among putative MAPK targets, isolated from a phosphopeptide screen [30], in vivo evidence of individual MAPK -MYB relations is still very limited. Given the documented role of MAPK cascades as stress mediators and the involvement of plant MYB proteins in diverse abiotic [29] and biotic [27] stress responses, a MAPK-regulated control of MYB transcription factors, similar to those of the WRKY family, appears likely.
Our earlier studies had indirectly linked a multiple-stressresponsive Arabidopsis MYB gene to MAPK-mediated stress signalling [18,31].Specifically, upon stress treatment, MPK3 becomes rapidly activated and phosphorylates the bZIP transcription factor VIP1 (virE2-interacting protein 1). Subsequently, phosphorylated VIP1 translocates to the nucleus, where it induces the expression of early stress-responsive genes, including MYB44 via direct promoter binding [31].
Here, we characterise MYB44 as direct and immediate component in the MKK4 -MPK3 signalling pathway. We locate the MPK3-phosphorylated residue in MYB44 and assess the effect of phosphorylation on MYB44 protein properties and its impact on abiotic stress tolerance. The role of MPK3 as post-translational regulator of MYB44 adds to its previously documented involvement in transcriptional regulation of the MYB44 gene. In vivo interaction of the MYB44 transcription factor with the nuclear subpools of MAPK cascade components is indicative of a second function, as scaffolding protein of the MKK4 -MPK3 module.

Formation of MYB44 -MKK4 complexes and MYB44 homodimers in yeast
With a prime interest in signalling proteins associated with the MKK4 -MPK3 submodule, we employed the yeast-two-hybrid (Y2H) system to screen an Arabidopsis cDNA library, using MKK4 as bait.
MKK4 interaction was assessed by yeast growth on selective medium and subsequent liquid ß-galactosidase assays. One of the interacting candidates was identified as transcription factor MYB44. Targeted yeast-two-hybrid analyses confirmed the interaction of (activation domain) AD-MYB44 and (binding domain) BD-MKK4 ( figure 1A). The strength of interaction was comparable to that of a positive control, (BD-MKK4/AD-MPK6). No ß-galactosidase activity was found in extracts from yeast co-transformed with BD-MKK4 and the pGAD control; or AD-MYB44 and a BD-control construct. In contrast, the reporter gene was induced upon co-expression of BD-MYB44 / AD-MYB44, clearly indicative of MYB44 intraspecific complex formation ( figure 1B).
Next, we tested via Y2H whether MYB44 would also interact with MPK3, the specific target of MKK4 in MAPK signalling. However, there was no specific GAL4 reporter gene activity in cells co-expressing MYB44 and MPK3. Likewise, co-transformation of MYB44 with the closest MPK3 homolog, MPK6, or other representative members of Arabidopsis MAPK subfamilies, yielded no detectable interaction (figure S1). These findings might suggest that MYB44 and MPK3 are no direct interacting partners, or could be indicative of poor protein interaction caused by interfering yeast proteins or by conformational constrains of the AD/BD fusion peptides. We therefore initiated alternative approaches in which we tested for regulatory crosstalk between MPK3 and MYB44.

MPK3 phoshorylates MYB44 in vitro
Several MAPK interacting proteins from animals, yeast and also plants, carry a characteristic domain, termed '' KIM'' (kinase interaction motif; R/K x2-6 I/L6I/L) [32]. Mutation analyses have shown an intact KIM to be required for Arabidopsis MAPK / phosphatase interaction [33]. Presence of KIM is an indication of but no absolute prerequisite for MAPK interaction. For instance, the MPK3 target protein VIP1 lacks such motif [18] . In MYB44, a putative KIM is located outside the MYB domain, in a non-conserved region (residues 171 to 179). Also, the protein sequence contains several sites matching the minimal consensus motif for MAPK-mediated phosphorylation, Ser-Pro or Thr-Pro (figure 2).
In vitro kinase assays were performed to test whether MYB44 can be phosphorylated by MPK3. In order to circumvent possible inhibitory/artificial effects of fusion peptides on MYB44 accessibility for the kinase, we produced tag-free MYB44 recombinant protein, applying the intein-based strategy (NEB). MPK3 was produced as GST fusion protein, because strong kinase activity of recombinant GST-MPK3 is well-documented [18,34]. The general MAPK substrate myelin basic protein (MBP) served as positive control. Both MBP and MYB44 were phosphorylated by MPK3. These findings indicate that MYB44 and MPK3 interact in vitro, and that one or several residues in MYB44 are targeted by the kinase (figure 3). MPK3/MYB44 complex formation was further supported by in vitro pull-down experiments (figure S2) MYB44 is phosphorylated by MPK3 specifically at Ser145 MYB44 contains four putative MAPK phosphorylation sites, located at S10, S53, S142 and S145 (figure 2). S10 and S53 are positioned in and between the R2/R3 repeats of the MYB DNA binding domain, respectively, whereas S142, S145 reside in a region close to the putative KIM (R171-L179). In silico analysis (EXPASY peptide cutter) predicts trypsin to cut at MYB44 residue: The proximity of residues S142 / S145 and their expected occurrence on a common, comparatively large peptide (spanning residues 126-208), would impede reliable prediction of phosphorylated residue(s) by mass spectrometry. However, phylogenetic considerations helped to narrow down the number of candidate MYB44 phosphorylation target sites. MYB44, together with MYB70, MYB77 and MYB73 forms the S22 subfamily of R2R3MYB transcription factors [27,35]. MYB44 residues S10 and S53 lie within the highly conserved region comprising the R2/R3 DNAbinding domain. All members of the S22 subfamily contain a serine residue corresponding to MYB44 S53, whereas a S10-corresponding site is found in MYB44, MYB70 and MYB73, but absent in MYB77. Comparison between MYB44, MYB70 and MYB77 with respect to their phosphorylatability may therefore help to restrict the number of MYB44 candidate phosphorylation sites. In vitro kinase assays with MYB44, MYB77 and MYB70 revealed MPK3-mediated phosphorylation of MYB44 and MYB77 but not of MYB70. These findings suggest a) that MYB77 is also a MPK3 target and b) that MYB44 residues S142 and/or S145, rather than S10 or S53 are the most likely targets sites for MPK3 phosphorylation (figure S3). Consequently, MYB44 variants mutated at residues Ser142 and/or Ser145 were generated. The exchange of serine residues to alanine (A) or aspartate (D) is a common approach for creating non-phosphorylatable or phosphomimetic variants, respectively (e.g.. [18,36]. Subsequent in vitro kinase assays revealed strong MPK3-mediated P 32P incorporation in MYB44 wild type protein as well as in the S142A and S142D variants. Contrary, signals were drastically reduced in S145A and [S142A/1455A]; and not detectable in S145D and [S142D/1455D] variants, respectively (figure 4), revealing S145 as major target site. The presence of strong signals in S142A and S142D variants, but complete absence in the phosphomimetic S145D variant furthermore indicates that S145 phosphorylation is not a prerequisite for subsequent phosphory- lation at any of the remaining (S10, S53 or S142) residues. Thus, MPK3 targets only one of the four candidate sites; namely S145.

Analyses of interaction between MYB44, MPK3 and MKK4 in vivo
Many transcription factors are known to form functional complexes. For example, some NAC and MADS proteins engage in homo-or hetero-dimeric or tetrameric aggregates [37][38][39]. Until now, the dimerization behaviour of MYB44 is unknown. In Y2H experiments, we had observed MYB44 / MYB44 interaction (figure 1B). Using bimolecular fluorescence (BiFC; [40]), we tested whether MYB44 homodimers would also be formed in planta. Two systems, Agrobacterium-mediated transformation of tobacco leaves and PEG-mediated transformation of Arabidopsis mesophyll protoplasts, were employed for transgene expression. Furthermore, MYB44/MPK3 interaction appeared to be limited to the nucleus. Similar results were obtained from BiFC studies in Arabidopsis protoplasts upon co-transformation with the abovedescribed DNA constructs (figure 6B). Reciprocal BiFC analyses (i.e. of protoplasts co-transformed with MYB44-YC/MPK3-YN) further support the existence and nuclear distribution of MYB44 / MPK3 complexes (figure S5). To assess whether interaction with MYB44 was a general feature of MAPKs, a distinct MAPK unrelated to MKK4 signalling, MPK5, was tested in parallel. MYB44-YN / MPK5-YC co-transformation into tobacco leaves or Arabidopsis protoplasts did not give rise to detectable fluorescence. However, reconstituted fluorescence was found in the positive control, i.e. cells co-expressing MPK3-YC and its known partner MKK4-YN, demonstrating functionality of our experimental setup. A similar nuclear signal distribution was detected, when testing MYB44 and MKK4 in BiFC assays ( figure 6A,B), substantiating the results from Y2H experiments. In contrast, when testing for interaction between MPK3 and MKK4 we found a nucleo-cytoplasmic distribution in tobacco leaf epidermis cells and Arabidopsis protoplasts ( figure 6A,B).
Taken together, signal distributions that were observed in BiFC assays are consistent with the reported subcellular localization of the individual proteins [41][42][43]. Moreover our results indicate that MYB44 specifically interacts with the nuclear sub-pools of MPK3 and MKK4.

MYB44 -DNA binding properties and localization are independent of S145 phosphorylation
In order to examine the significance of MYB44 phosphorylation for its predicted function in transcriptional control, gel retardation The affinity of transcription factors for certain DNA elements is essentially determined by their DNA binding domains. An exchange of crucial residues within that domain can completely disrupt DNA binding or alter the proteins preference for a certain motif. In vitro studies with bZIP proteins in which serine residues within the bZIP domain had been systematically exchanged suggest that phosphorylation in the DNA-binding domain at selective positions could be a means of fine-tuning transcription factor activity [44]. Modifications outside the DNA-binding domain can also alter DNA binding properties, e.g. if protein folding and/ or dimerization are affected.
The DNA consensus sequence recognized by plant MYB proteins (CCT/AACC) [45] differs substantially from that bound by animal MYB proteins(C/TAACGG) [45,46]. MYB44 binds to the typical plant MYB binding site, MBSII from Petunia, but it also has affinity to the animal-like MBSI motif [47]. This might be due to specific features in the R2R3 recognition helices: The R2 helix is more plant-like; while the R3 helix is more animal-like [47].
To assess whether the DNA binding behaviour of MYB44 was (S145)-phosphorylation-dependent, electrophoretic mobility shift assays (EMSA) were conducted (figure 7). MBSII and MBSI sequences are based on Kirik et al., [47]. Recombinant (purified, tag-free) MYB44 wild type, S145A and S145D proteins were incubated with 59-IRDye-labelled MBSII double-stranded DNA    (lane 4). An excess of MBSI (59-AGCTTAATGTGTGTCAGT-TAGGGTGTCTCG-39) also reduced probe binding, although less severely than the specific competitor MBSII (lane 5). Thus, MYB44 specifically binds to MBS-containing elements, with a clear preference for the plant-like MBSII type. There were no discernible differences in terms of DNA affinity or specificity between MYB44 wild type and the S145A or S145D variants (lanes 6-13). Hence, these properties of the MYB44 protein appear to be S145 phosphorylation-independent.
BiFC studies had revealed an interaction of MYB44 with MPK3, and also with MKK4, in the nucleus (figure 6). Theoretically, these proteins might reside in a common complex with MYB44-bound DNA. Examples of nucleoprotein complexes containing MAPKs are known from yeast [48]. Using EMSA, we tested whether the presence of MPK3 or MKK4 would alter the DNA binding characteristics of MYB44. There were no discernible differences between samples containing labelled MBSII and MYB44 alone, or MYB44 plus MPK3 and/or MKK4. Also, preincubation of the proteins prior to adding MBSII and/or the addition of ATP plus MPK3 (to produce phosphorylated MYB44) had no visible effect ( figure 7B). Therefore, at least in vitro, MPK3 or MKK4 are apparently not contained in a common MYB44 -DNA complex, nor do these kinases assist or prevent MYB44 -DNA binding.
Transient ectopic expression in tobacco leaves was employed to investigate possible effects of phosphorylation on MYB44 subcellular localisation. The DNA constructs tested encoded for MYB44, MYB44 S145A and MYB44 S145D fusions to full-length yellow fluorescent protein. Overlapping with the specific distribution of MYB44 homodimers (figure 5), MYB44-YFP was detected exclusively in the nucleus (figure 8). The subcellular sorting of MYB44 is likely due to the presence of an arginine/lysine-enriched peptide region, representing a putative nuclear localisation signal (NLS). This NLS lies in the N-terminal part of the protein. Similar to MYB44-YFP, S145A and S145D-derived fluorescence signals were present only in the nucleus (figure 8).
To investigate possible effects of S145 modification on homodimerisation, MYB44 (de)phosphomimetic variants were examined by BiFC. Cells co-expressing S145A-YN/S145A-YC or S145D-YN/S145D-YC showed reconstituted fluorescence, indicative of nuclear homospecific protein complexes (figure 9). In conclusion, MYB44 nuclear import and/or retention as well as homodimerisation properties seem independent of S145 phosphorylation.

MYB44 phosphorylation regulates osmotic stress tolerance
MYB44 overexpression in Arabidopsis has been reported to improve stress tolerance towards osmotic (NaCl) and drought stress [42]. Moreover, a recent study bridged MKK4 to the MPK3-mediated NaCl response [49]. As was revealed by in-gelkinase assays, MKK4 regulates the activity of MPK3 upon NaCl exposure. mkk4 mutants, which are impaired in MPK3 activation, have a diminished stress tolerance, whereas MPK3 hyperactivation through MKK4 overexpression has the opposite effect [49] . These observations might suggest that MYB44 overexpressionconferred stress tolerance is directly related to MKK4-MPK3 signalling and -presumably-phosphorylation-dependent. To test this hypothesis, homozygous Arabidopsis plants overexpressing MYB44 or its (de)phosphomimetic variants, S145A and S145D, respectively, were generated. Transgene expression was confirmed by immunoblot analysis with an antibody directed against the Cterminal myc tag (figure 10). A single specific band was detected at approximately 50 kDa, which is close to the expected size (44.9 kDa). The migration of MYB44wt-, S145A-and S145D-myc fusion proteins was comparable.
Under control conditions, i.e. germination on K MS agar and transfer onto soil at 14 DAG, no significant differences between wild type and transgenic plants were observed. Transgene expression had no discernible effect on germination, development, seed yield or fertility, either.
Regarding the phosphomimetic variant, S145D survival was better (2d 71%, 3d 20%, 4d 3%) than in Col-0, but lower than in MYB44wt. This indicates that the S145D variant cannot fully replace MYB44wt as positive regulator of salt stress signalling. Comparison of stress survival in MYB44wt, S145D and S145A lines emphasizes the importance of that particular serine residue and indicates S145 phosphorylation to be a prerequisite for MYB44-mediated salt tolerance. Consistently, repeats of the salt stress assay with additional independent transgenic lines, disclosed the same trend of tolerance levels (highest.lowest) of MYB44wt . S145D . Col-O . S145D (figure S6).

MYB44 phosphorylation and abscisic acid
Many abiotic stresses, such as osmotic or drought stress, evoke the generation of the plant hormone abscisic acid (ABA) [50,51]. Mutant plants affected in ABA synthesis, perception or signalling display abnormal phenotypes under abiotic stress conditions [5,52,53]. Importantly, MPK3 is activated by diverse abiotic stresses that are related to ABA signalling, and also upon exogenous application of this hormone [54,55]. Besides regulating abiotic stress responses, ABA also acts in the promotion of seed dormancy and inhibition of seed germination [56]. MYB44 overexpression was shown to enhance sensitivity to ABA-induced post-germination growth arrest [42]. However, the cellular/ molecular processes underlying MYB44-mediated stress tolerance is still unknown. Based on the altered osmotic stress tolerance phenotypes found in MYB44, S145A and S145D lines and the implication of MPK3 in ABA-mediated responses, we suspected a relation between ABA-induced, MPK3-mediated processes and MYB44 phosphorylation.
To test this hypothesis, Col-0, mpk3, MYB44, Ser145A and S145D overexpressing lines were sown on medium supplemented with 1mM ABA, and germination was monitored over a 7-day period. On control medium lacking the hormone, seeds of all plant lines germinated equally well within two days. However, significant differences were observed in the presence of ABA. Compared to Col-0 wild type, germination in all tested lines was reduced, with mpk3 and S145A overexpressors showing the highest sensitivity ( figure 12). The hypersensitive phenotype of mpk3 mutants conforms to a recent study by [41], who found Arabidopsis pp2c5 mutants to be insensitive to germination inhibition by ABA. These mutants are deficient in the MAPKinactivating phosphatase PP2C5 and exhibit ABA-induced MPK3 hyper-activation [41] . Accordingly, pp2c5 and mpk3 have opposite germination potential in the presence of ABA.
Congruent with published data [42], we found MYB44wt overexpression to confer enhanced ABA sensitivity. Compared to MYB44wt, germination frequency of S145D lines was reduced up to day four, but reached similar levels later on. In contrast, overexpression of the non-phosphorylatable S145A variant drastically reduced germination through-out the study period. Interestingly, germination graphs of S145A and mpk3 largely overlapped -a similarity that we had noted already in the osmotic tolerance test ( figure 11). The differences in ABA sensitivity between S145A vs. MYB44wt type or S145D lines reiterates the importance of phosphorylation at this single residue, Ser145, to MYB44 function.

Discussion
We report the identification and characterisation of MYB44, an in vitro and in vivo interacting partner of the MAPK cascade components MKK4 and MPK3. Our study provides the first evidence of a direct link between MYB proteins and MAPKKs.
As documented by Y2H and BiFC experiments, MYB44 is an auto-interacting protein. It exclusively locates to the nucleus, where it also engages in interspecific complexes with a closely related member of the MYB S22 subfamily, MYB77. While MKK4 / MPK3 interaction occurs both in the cytoplasm and nucleus, MYB44 / MKK4 and MYB44 / MPK3 complexes are strictly nuclear (figures 5 and S6). The irreversibility of the reestablished fluorophore complex in BiFC assays prevents the analysis of dynamic interactions [57]. Alternative approaches will have to be taken to reveal the kinetics of complex formation. Results from BiFC studies in cells coexpressing MYB44 and MPK5 imply that the ability to bind to MYB44 is not a general feature of MAPKs. Importantly however, MPK6, the closest homolog of MPK3, was tested positive for interaction with MYB44, both in Arabidopsis protoplasts and tobacco cells ( figure  S7). Similar to MPK3, MPK6 binds to MYB44 exclusively in the nucleus, while MPK6 / MKK4 complexes show a nucleocytoplasmic distribution ( figure S7). MKK4 is a well-documented upstream regulator of MPK3 and MPK6, but not of MPK5 [58].  question. In our study we focused on the relation of MYB44 and MPK3.
MYB44 is phosphorylated by MPK3, and we could locate the phosphorylation site to a single residue, Ser145. The faint band seen by in vitro kinase assay (figure 4) in samples of S145A and [S142A/1455A] likely arises from a minor, presumably nonspecific phosphorylation at S10 and/or S53, as it is absent in samples of S145D or S142D/S145D. A possible explanation is that, if the true target site, i.e. S145 is blocked (S145A), MPK3 yet ''tries'' to transfer a phosphate residue to MYB44. In contrast, MYB44 S145D or [S142D/145D] proteins, for which no signal was detected, ''suggest the kinase'' that it has performed its function, i.e. phosphorylation, on them already. Whether any of the three remaining Ser-Pro motifs, that match the MAPK target consensus (S/T-P), are targeted by other members of the Arabidopsis MAPK family is an open question. Our single-site location of MPK3 phosphorylation in MYB44 partially contradicts a very recent report. Nguyen et al. [59] analysed recombinant GST-MYB44 fusion proteins after in vitro phosphorylation via MALDI-TOF gas chromatography and concluded that both S145 and S53 are MAPK-targeted residues. Congruent with our observations, in vitro kinase assays [59] revealed a drastic reduction of phosphorylation signals upon Ser145 exchange to alanine, but in fact a similarly weak signal remained in the (S53A/S145A) double-mutated version-which in our eyes does not support the concept of S53 phosphorylation. Compared to control plants, overexpression of MYB44-S53A/S145A in Arabidopsis did not alter ABA sensitivity [59] . We think that the S53A mutation in the N-terminal DNAbinding domain interferes with the formation of the R2/R3 helices and renders MYB44 non-functional. The authors did not include single-mutated MYB44 variants or further functional assays in their study.
We investigated the possible impact of Ser145 phosphorylation on MYB44 function. The replacement of Ser145 by amino acids mimicking a constitutively nonphosphorylated (S145A) or phosphomimetic state (S145D) had no discernible effect on the proteins subcellular distribution ( figure 8). Possibly, recognition of the NLS by the nuclear import machinery remains unaffected by modifications at the rather distantly-located S145. Thus, unlike the MPK3-regulated bZIP transcription factor VIP1 [18], nuclear localisation of MYB44 appears to be regulated in a phosphorylation-independent manner. The capacity of MYB44 to form intraspecific complexes also is apparently maintained in S145A and S145D ( figure 9).
We further examined the DNA binding properties of MYB44 and its S145 variants. Deletion of the C-terminal region (residues 213 onwards) was shown to enhance MYB44 -DNA binding [47], suggesting that protein features outside the R2R3 domain (spanning residues 6 -103, figure 2) might contribute to the proteins DNA affinity.   In gel retardation assays we observed similar DNA element affinity and specificity of MYB44 and its (de)phosphomimetic variants (figure 7A), implying that MYB44 -DNA binding is S145 phosphorylation independent. In vitro studies with typical animal and plant MYB proteins have demonstrated that the DNAbinding properties (strength and specificity) are determined by the corresponding MYB domains and not by other regions of the protein [60]. In contrast to the major deletion [47], a single amino acid exchange at S145, a position located in significant distance to the R2R3 domain, presumably does not alter the proteins overall conformation to an extent that would affect folding of the DNAbinding helices.
Consistent with S145A and S145D exchanges having no discernible effect on the DNA affinity or -specificity of MYB44, pre-or co-incubation with ATP and MPK3 (to produce phosphorylated MYB44) did also not alter these protein properties ( figure 7B). Gel retardation experiments further suggest that, at least in vitro, MYB44 binds its cognate DNA motif independently of MPK3 and/or MKK4. Of course, it cannot be excluded that (labile) MKK4/MPK3-MYB44-DNA complexes are destabilised under native gel electrophoresis conditions. Whether in vivo, the two functions of MYB44 -transcription factor and putative MKK4-MPK3 scaffolding protein -occur independently, synergistically or antagonistically is subject of our current research.
Stress tolerance tests with transgenic Arabidopsis plants clearly indicate that phosphorylation at S145 is highly relevant to MYB44 function. MYB44 properties other than its subcellular distribution, intra-specific complex formation or DNA binding characteristics (figures 7-9) are likely to be the cause for the altered stress performance exhibited by the MYB44 wild type, S145A or S145D overexpressing plant lines. Future studies involving transactivation capacity tests and transcriptome analyses of transgenic plants shall clarify this aspect. The affinity of MYB44 to other types of transcription factors may be controlled in a phosphorylationdependent manner. Co-regulation of gene expression has been shown for MYB77/ARF(auxin response factor 7) [61].
Compared to Col-0, plants overexpressing MYB44 or its phosphomimetic variant S145D display elevated tolerance to salt stress, while S145A lines are more sensitive (figures 11). Osmotic stress survival in S145D was lower than in MYB44wt lines, indicating that the S145D variant cannot fully replace MYB44wt as positive regulator of salt stress signalling. We think that in treated seedlings a sub-pool of MYB44 exists in the phosphorylated form, and replacement at Ser145 by aspartate may only partially mimic a truly phosphorylated serine residue. Despite their enhanced tolerance towards osmotic shock treatment, MYB44 and S145D transgenic lines withstand this stress for a period of only few days (figure 11). The absence or low abundance of further stress-related factors may limit prolonged seedling survival. To determine precisely up to which stress intensity and at which developmental stage(s) MYB44 plays a major role, substantial additional research is needed.
Intriguingly, osmotic stress tolerance is suppressed in mpk3 and S145A lines to similar extent. In mpk3, this reduced tolerance is likely due to the inability to phosphorylate proteins, including MYB44, that act downstream of MPK3 in stress signalling. The lowered tolerance in S145A overexpressing lines is indicative of a dominant-negative effect, which might by attributable to: i) S145A forming inactive complexes with endogenous MYB44, closely related MYB proteins or other types of transcription factors, e.g. bHLH proteins. In fact, the presence of nuclear MYB44 / MYB44 (figure 5) and MYB44 / MYB77 aggregates (figure S4) supports this assumption. Highly abundant S145A might subtract the pool of monomeric MYB proteins, preventing the formation of functional transcription factor complexes. ii) S145A competition with endogenous MYB44 or close homologs (e.g. MYB77), and/or other stress-related substrates for MPK3 binding sites and phosphorylation. iii) S145A blocking the formation of functional MKK4-MPK3 modules (subject of our future research).
The stress phenotypes caused by overexpression of MYB44wt, S145A or S145D are unlikely due to competition with endogenous MYB44, since loss of myb44 was shown to have no obvious effect on abiotic stress performance [42]. As already noted by Jung et al [42], the strong similarity in expression profiles indicates a certain redundancy among the four members of the R2R3 MYB S22 subfamily. Moreover, microarray data indicate a cross-regulation within the S22 family: MYB44 overexpression correlates with reduced expression of the three remaining S22 members [42] . A possible approach to circumvent redundancy effects would involve overexpression of MYB44, S145A or S145D in mutants deficient in multiple members of the MYB S22 subfamily, e.g. (myb44/ myb77/myb70/myb73). Whether such quadruple mutants would be viable and develop normally is not known. Phylogenetic considerations and in vitro kinase experiments (figures 2 and S3), revealing MPK3 phosphorylation of MYB77, but not of MYB70, suggest a particularly close relatedness between MYB44 and MYB77.
Interestingly, microarray data [42] point to a specific transcriptional auto-regulation within the MYB S22 subfamily. MYB70, MYB73 and MYB77 are among the (few) genes significantly repressed in MYB44 overexpressing plants. This might be a further indication of gene redundancy, but remains to be determined in further detail.
Our data strongly suggest that in planta interaction of MYB44 with MPK3 has an imminent effect: phosphorylation of the transcription factor. Yet, the meaning and outcome of the direct MYB44 -MKK4 aggregation remains hypothetical. ''Short cuts'' in MAPK(KK) signalling do exist, as exemplified by the direct phosphorylation of WRKY53 by the MAPKKK MEKK1 [20] . However, a direct MYB44 phosphorylation by the kinase MKK4 seems unlikely, since peptide motifs matching the minimal consensus of MAPKK-targeted proteins (dual phosphorylation at Thr-Asp-Tyr or Thr-Glu-Tyr) are not contained in the MYB44 sequence. Possible scenarios include: i) MYB44 acts as scaffolding protein, tethering the functional MKK4-MPK3 complex. By stabilising the MAPKK-MAPK sub-module, MYB44 would thus reinforce its phosphorylation by MPK3. Because MYB44 has a very low basal, but strongly stress-induced expression (Genevestigator), it would support MKK4-MPK3 association in a stress-dependent manner. ii) MYB44 binding to MKK4 may prevent MKK4 association with other MAPKs, thereby ensuring MAPKK-MAPK specificity: Such function would not be unique to MYB44, since myb44 mutants do not display developmental or stressrelated abnormalities [42]. iii) By binding to MKK4, MYB44 may alter the cyto-nuclear distribution of MKK4 in favour of the nuclear pool. This might be achieved by assisting MKK4 nuclear import and/ or preventing MKK4 nuclear export. The subcellular site(s) of MKK4-mediated phosphorylation of MPK3 are unknown. According to our BiFC studies, the two proteins interact in cytoplasm and nucleus. Whether the cytoplasmic or nuclear micro-environment differentially facilitates MKK4 activity is still elusive.
Intriguingly, heterologous overexpression of the Arabidopsis MYB44 gene confers an enhanced osmotic and drought stress tolerance to soybean [62]. This aspect raises several highly interesting questions, e.g.: Do soybean MAPK(s) recognize the heterologous substrate, or can MYB44 bind stress-related promoters independently of its phosphorylation status? Would the S145D phosphomimetic variant be even more effective? Is the tolerance-enhancing effect of MYB44 solely due to its function as transcription factor, or can it act as scaffold to assist specificity in MAPKKMAPK signalling similar to the (yet hypothetical) scenario in Arabidopsis? Can these functions be uncoupled? Analyses of MYB44wt and S145D alleles in the mpk3 mutant background shall help to address some of these aspects. Furthermore, mpk3 mutants may be employed for in-gel kinase assays to elucidate if MYB44 phosphorylation maintains correlation with stress exposure, and if MPK3 can be functionally replaced by MPK6 in phosphorylating MYB44 under these conditions.
In this study, possible roles of MYB44 phosphorylation in the biotic stress response were not investigated. Given the known role of MKK4-MPK3 during pathogen attack [63], and the implication of MYB44 in insect attack [64], harpin [65] and VIP1mediated signalling [31], an altered stress performance in transgenic MYB44 versus its (de)phosphomimetic variant lines appears very likely. The challenge and chance for research and agriculture is to develop a MYB44 derivate that is functional in diverse plant species and that incorporates a maximum of its stresstolerance-enhancing properties. Exploration of the potential biotechnological implications of the results represented here is part of our ongoing research.

Methods
Yeast-two-hybrid interaction assays S. Cerevisia strain L40 was transformed with plamids pGAD (for activation domain fusion) and pBTM (for binding domain fusion) and selected on -Leu/-Trp medium, as described in [66].The vectors contained coding sequences of Arabidopsis MYB44, MAPKs or MAPKKs. Interaction was assessed using the liquid ß-galactosidase activity assay as described in .In the yeast-twohybrid screen, strain L40 was cotransformed with MKK4-pBTM as bait, and an Arabiodopsis cDNA library contained in pAct-AD (Clontech) as prey. Cells were selected on -His/-Leu/-Trp medium.

Plant Material, germination assay on ABA and osmotic stress test
Arabidopsis thaliana ecotype Columbia (Col-0) plants were routinuously grown at 20uC, 60% humidity under long-day conditions (16/8 h light/dark). To break dormancy, seeds were incubated at 4uC for 2 days.
At least 2-3 individual T3 transgenic lines were used for the stress assays. Experiments were independently repeated at least three times.
Surface-sterilised seeds were incubated on half-strength MS medium/0.25% sucrose / 1% plant agar with or without 1mM ABA. After stratification for 2d at 4uC, plates were transferred to an incubator (8/16 h dark/night regime, 25uC). Germination percentage was monitored after 3 to 7 days. Seeds were considered germinated after the cotelydones had emerged.
To assess osmotic stress tolerance, seeds were plated on halfstrength MS medium (see above). After three days, seedlings were transferred to plates containing half-strength MS supplemented with 150 mM NaCl and grown for further 4 days. About 50 seeds were used for each treatment.

Plasmids and mutagenesis
The full-length MYB44 coding sequence was PCR-amplified from Columbia-0 cDNA with primers introducing a 59BspHI (gatattcatgactgataggatcaaaggtcc) and 39NotI site (atagcggccgccgattctcccaactccaatttgac).
All constructs used for stable and transient expression in Arabidopsis and tobacco are based on the binary vector pGreen0029 [67], containing the CaMV35S promoter and nos terminator. MYC or YFP tags were inserted at the candidate gene 39end as NotI-EagI fragments.

Recombinant protein isolation and in vitro kinase assay
Taq-free recombinant MYB proteins were isolated using the intein-splicing system (IMPACT TM Kit, Biolabs). Escherichia coli strain BL-21 DE3 was transformed with vector pTX3B, containing the full-length coding sequences of MYB44 or its S142/S145 variants. Recombinant proteins were expressed as maltose binding protein -intein fusion proteins. Isolation, purification and final release of cleaved purified proteins was performed according to the manufacturers instructions. Induction, purification and isolation were monitored by SDS-PAGE/Coomassie staining.
GST-MPK3 was expressed in BL-21 and isolated as described previously [34].
In vitro kinase assays with purified GST-MPK3 and MYB proteins were performed according to [34] . Kinase reactions were heat-denatured and resolved by 12% SDS-PAGE and visualized by Coomassie staining. Vacuum-dried gels were exposed to X-ray film (Amersham) for 1-3 days.

Arabidopsis protoplast isolation, transformation and subcellular localisation studies
Mesophyll protoplasts were isolated from leaves of 3 to 4-weekold Arabidopsis Columbia-0 plants. The protoplast isolation and transfection procedure was based on the method by Wu et al. [68], modified according to Pitzschke and Persak [69]. 12-16 h posttransformation, subcellular localisation studies were conducted at a UV microscope (Leica DM5500B), equipped with excitation/ emission filters: BP450-450 nm/LP515 nm. Confocal images were acquired with a TCS-SP5 microscope using excitation wavelength of 488 nm (argon laser) for YFP and 594 nm for autofluorescence, as previously described [69].

Transient expression in tobacco
Nicotiana tobaccum, leaves were infiltrated with agrobacteria (strain GV3101, carrying pSoup and the pGreen construct of interest) according to Hellens et al. [70]. 4-7 days post-infiltration, epidermal peels were investigated by microscopy as described above.

Generation of stable transgenic Arabidopsis plants
The floral dip method [71]was applied to transform 6 to 7week-old Arabidopsis Columbia-0 plants with Agrobacteria strain GV3101, carrying the helper plasmid pSoup [67] and wild type or S142/S145 variants of MYB44-myc_pGreen. Transformants were selected on kanamycin and tested for transgene expression by immunoblot analysis. Progeny of plants showing a 1:3 segregation (kanamycin resistance marker) were propagated in a homozygous state. All analyses were performed with homozygous lines.

Plant protein extraction and immunodetection of MYB44-myc
Proteins were extracted from Arabidopsis leaves or eight-dayold seedlings as described by Brock et al. [41]. 15 mg protein (quantified by Bradford assay) were separated by SDS-PAGE and blotted on a PVDF membrane (Porablot, Roth). For immunodetection of MYB-myc fusion proteins, rabbit anti-myc and IRDye800CW-coupled donkey-anti-rabbit (NEB Biolabs) were used as primary and secondary antibodies, respectively. Membranes were scanned with a LiCOR Odyssee machine, at 800 nm for the detection of bound antibodies and at 700 nm to document equal sample loading. Methods S1 (DOC)