Phosphatidylinositol 5-Phosphate Links Dehydration Stress to the Activity of ARABIDOPSIS TRITHORAX-LIKE Factor ATX1

Background Changes in gene expression enable organisms to respond to environmental stress. Levels of cellular lipid second messengers, such as the phosphoinositide PtdIns5P, change in response to a variety of stresses and can modulate the localization, conformation and activity of a number of intracellular proteins. The plant trithorax factor (ATX1) tri-methylates the lysine 4 residue of histone H3 (H3K4me3) at gene coding sequences, which positively correlates with gene transcription. Microarray analysis has identified a target gene (WRKY70) that is regulated by both ATX1 and by the exogenous addition of PtdIns5P in Arabidopsis. Interestingly, ATX1 contains a PtdIns5P interaction domain (PHD finger) and thus, phosphoinositide signaling, may link environmental stress to changes in gene transcription. Principal Findings Using the plant Arabidopsis as a model system, we demonstrate a link between PtdIns5P and the activity of the chromatin modifier ATX1 in response to dehydration stress. We show for the first time that dehydration leads to an increase in cellular PtdIns5P in Arabidopsis. The Arabidopsis homolog of myotubularin (AtMTM1) is capable of generating PtdIns5P and here, we show that AtMTM1 is essential for the induced increase in PtdIns5P upon dehydration. Furthermore, we demonstrate that the ATX1-dependent gene, WRKY70, is downregulated during dehydration and that lowered transcript levels are accompanied by a drastic reduction in H3K4me3 of its nucleosomes. We follow changes in WRKY70 nucleosomal K4 methylation as a model to study ATX1 activity at chromatin during dehydration stress. We found that during dehydration stress, the physical presence of ATX1 at the WRKY70 locus was diminished and that ATX1 depletion resulted from it being retained in the cytoplasm when PtdIns5P was elevated. The PHD of ATX1 and catalytically active AtMTM1 are required for the cytoplasmic localization of ATX1. Conclusions/Significance The novelty of the manuscript is in the discovery of a mechanistic link between a chromatin modifying activity (ATX1) and a lipid (PtdIns5P) synthesis in a signaling pathway that ultimately results in altered expression of ATX1 dependent genes downregulated in response to dehydration stress.


Introduction
Phosphoinositides can mediate differentiation, cellular growth, and responses to biotic and abiotic stress [1]. Differential phosphorylation of phosphatidylinositol results in phosphoinositides, which may function as precursors of second messengers or to modulate the localization, conformation, and activity of bound proteins. PtdIns5P is a minor component of the cellular lipid pool implicated in the cell osmoprotective response pathway [2], in the etiology of muscular and neuronal pathologies [3], and in the hostcell response to infection by S. flexneri [4]. In the nucleus, PtdIns5P can interact with ING2, an adaptor protein regulating p53 and histone acetylation in response to cellular stress [5,6]. In plants, PtdInsPs function in responses to salinity, drought, and temper-ature stresses [7][8][9]. PtdIns5P accumulates in response to hyperosmotic stress in Chlamydomonas, in plant tissues from tomato, pea, alfalfa, and in cultured carrot cells [9]. PtdIns5P has not been formally identified in Arabidopsis [10] but Arabidopsis proteins Patellin1 and ATX1 can bind PtdIns5P in vitro [11,12].
ATX1 is a chromatin modifier that tri-methylates lysine 4 of histone H3 of associated nucleosomes [13]. ATX1 binds specifically PtdIns5P in vitro and microarray analysis revealed a set of co-regulated genes [12,14] suggesting the existence of a shared pathway in which PtdIns5P acts as a ligand modifying the activity of ATX1. However, evidence for the existence of PtdIns5P in Arabidopsis, for a change in the lipid level under stress, and for role of PtdIns5P in ATX1 activity is lacking. Here, we investigated whether PtdIns5P existed in Arabidopsis, whether the PtdIns5P level changed upon dehydration stress, whether endogenous PtdIns5P could affect the activity of ATX1, and how this might occur. We tested the hypothesis that dehydration stress increases cellular PtdIns5P, which acting as a ligand modulates ATX1 activity in vivo. Without excluding a role for PtdIns5P in chromatin [15], we elucidate a PtdIns5P-driven mechanism that sequesters ATX1 in the cytoplasm suppressing its function in the nucleus. The cytoplasmic sequestration is dependent on the ePHD finger, which interacted with PtdIns5P in vitro, and on factors that affect the cellular level of PtdIns5P. The model suggests a pathway mediated by PtdIns5P that translates environmental stress into altered activity of a chromatin modifier.

PtdIns5P in Arabidopsis thaliana
A high cellular level of PtdIns4P, a predicted low level of PtdIns5P, and the overlap of PtdIns4P and PtdIns5P peaks in high performance liquid chromatography (HPLC) impede identification of PtdIns5P in Arabidopsis [10] (Figure 1).
To determine the presence of endogenous PtdIns5P in Arabidopsis, we used a mass assay based on the specificity of PIP4Ka to phosphorylate PtdIns5P at the 49-position of the inositol head group [16] allowing a quantitative determination of intracellular PtdIns5P [17].
Total phospholipids extracted from four different tissues were used as starting material for the PtdIns5P mass assay. The two labeled PtdInsP 2 isomers formed in all the reactions were putatively identified as PtdIns(4,5)P 2 (spot 1) and PtdIns(3,4)P 2 (spot 2) by thin layer chromatography (TLC) (Figure 2a). According to the known substrate specificity of PIP4Ka [16], these reaction products reflected the endogenous PtdIns5P and PtdIns3P, respectively. To further identify the products they were deacylated, subjected to HPLC analysis and identified as PtdIns(4,5)P 2 and PtdIns(3,4)P 2 (Figure 2b). In all subsequent analyses we focused exclusively on determining the mass of endogenous PtdIns5P.
Both TLC and HPLC analyses indicated presence of PtdIns5P in the four tested plant tissues: rosette leaves, stems, inflorescences, and siliques. However, a potential caveat was that although PtdIns4P is not a substrate for the PIP4Ka, it was still possible that the enzyme had different substrate specificity for PtdInsPs of plant origin, particularly in cases when PtdIns4P is more abundant than PtdIns5P. To confirm that PIP4Ka phosphorylated PtdIns5P on the 49 position of the inositol ring, the two radiolabeled products (spots 1 and 2) were together incubated in the absence or presence of yeast 59-phosphatase (Figure 2c).
A decrease in the radioactivity in the PtdIns(4,5)P 2 spot coinciding with the appearance of radioactivity in PtdIns4P (containing an equal amount of radiolabel lost from PtdIns(4,5)P 2 (spot 1) indicated that the radiolabel in the PtdIns(4,5)P 2 was exclusively on the 49-position. The 59-phosphatase did not affect the intensity of the PtdIns(3,4)P 2 (spot 2). We conclude that the PtdIns(4,5)P 2 produced by PIP4Ka originates from endogenous PtdIns5P present in Arabidopsis.

Cellular PtdIns5P levels upon osmotic and hypotonic stresses
Osmotic stress increases PtdIns5P in yeast, animal, and plant cells [2,9,10,18]. Dry air, drought and floods are environmental factors that stimulated our interest to examine changes in cellular PtdIns5P upon exposure of Arabidopsis to these stresses. In concurrent unrelated studies we have established that 9-12 days exposure to water withdrawal conditions (in-soil dehydration stress) followed by 3 day-watering resulted in cellular water loss (,60% residual water levels) and cellular water level recovery (as measured by the restoration of cell turgidity) that were similar to the water-loss and recovery levels achieved in detached leaves when exposed for 2 hours at room temperature conditions (20uC) followed by an overnight submergence in water [19; a submitted manuscript]. In addition, air-dried detached leaves showed little variation in residual cellular water levels (between 58-62%) making it the preferred choice for inducing dehydration stress in Figure 1. Failure to detect PtdIns5P by HPLC. HPLC analysis of deacylated lipid products isolated from rosette leaves of in vivo radiolabeled whole plants with orthophosphate (300 mCi/ml) for 16 hours at room temperature. Optimal conditions for our analyses were established by harvesting rosette leaves and exposing them to ambient air (20uC). Water-loss was determined as the residual tissue mass measured up to four hours of air-exposure ( Figure 3a). PtdIns5P was measured in total lipids extracted after 30 minutes and after 2 hours exposure to air. Leaves collected from the same plants subjected immediately to lipid extraction were used as controls (0 min). In fresh (non-treated) Arabidopsis leaves, the amount of total phospholipid and of PtdIns5P measured in six independent experiments were: 2.08 nmoles/mg fresh tissue (s.d. 0.22) and 3.6 pmoles/mg fresh tissue (s.d. 0.096), respectively (Figure 3b, see also Figure 4a). To examine the effect of water stress, detached leaves were submerged in water for 2 hours before measuring lipid content. Leaves collected from the same plants and immediately subjected to lipid extraction were used as controls ( Figure 3c). The results suggested that leaf submergence in water also increased the cellular PtdIns5P, although to a lesser degree than dehydration for the same length of time. Because a 2 hour exposure to air showed the highest level of PtdIns5P accumulation, we used this condition (from here on referred to as dehydration-stress) to experimentally trigger elevated cellular PtdIns5P.

PtdIns5P levels in AtMTM1 over-expressing and knockout cells
A family of phosphoinositide 39-phosphatases (myotubularins, MTM1s) generates cellular PtdIns5P through their activity  Detached (column 1), airexposed (column 2) and water-submerged leaves (column 3) were treated in parallel for 120 minutes. Cellular PtdIns5P is indicated as pmoles PtdIns5P/mg initial fresh tissue mass. Bars [20][21][22]. Inactivation or constitutive expression of MTM1 is a strategy used to study the roles of its substrates (PtdIns3P and PtdIns3,5P 2 ) and respective products (PI and PtdIns5P) in cellular homeostasis.
Recently, we found that the Arabidopsis gene At3g10550 encodes a protein 59% similar to the human MTMR2 (called AtMTM11) and, like mammalian myotubularins, has a phosphoinositide 39phosphatase activity against both PtdIns3P and PtdIns3,5P 2 [19]. Whether constitutive expression of AtMTM1 could increase the level of cellular PtdIns5P levels in plant cells is a key question, which was addressed by over-expressing AtMTM1 (OX-AtMTM1) in transgenic plants. The cellular PtdIns5P level was determined in transgenic and wild type plants before and after stress. Differences in PtdIns5P levels were evaluated by the Mann-Whitney (U)-test [23].
Upon dehydration, PtdIns5P increased by ,30% in wild type leaves (U-test p,0.006, Figure 4a). A further increase was seen upon dehydration of OX-AtMTM1 cells producing ,70% more PtdIns5P than non-stressed wild type cells (U-test p,0.002) implicating AtMTM1 in elevating cellular PtdIns5P upon stress. Importantly, solely the over-expression of AtMTM1 led to an increase in PtdIns5P that was similar in magnitude to dehydration stress stimulated wild type plants. To assess if AtMTM1 links dehydration stress with increased PtdIns5P, we measured the endogenous level of PtdIns5P in homozygous knockout mtm1 mutant plants. In non-stressed mtm1 leaves, PtdIns5P was not significantly different from that in non-stressed wild type leaves (Figure 4a, U-test p.0.2). However, the increased PtdIns5P seen in wild type plants in response to stress was completely blunted in mtm1 mutant leaves producing significantly less PtdIns5P (,35% less, U-test p,0.01) than wild type leaves under stress. The difference was even more pronounced when dehydration-stressed mtm1 mutant leaves were compared with dehydration stressed OX-AtMTM1 leaves (,75% less PtdIns5P, U-test p,0.001) ( Figure 4a). These data underscore the role of MTM1 activity in increasing PtdIns5P upon stress but not of the basal PtdIns5P level. As both PtdIns5P and ATX1 regulate WRKY70 and ATX1 interacts with PtdIns5P, we sought to determine a mechanistic link between these observations in response to dehydration stress.

ATX1 activity in stressed plants
As previously shown [24] the transcript level of WRKY70 is regulated by ATX1 as it decreased in homozygous ATX1 deleted (atx1) plants ( Figure 4b). Chip analysis with antibodies specific to tri-methylated histone 3 lysine4 showed a dramatic decrease in H3K4me3 signal from the 59 end of the WRKY70 gene in atx1 plants (Figure 4c), showing that the activity of ATX1 is largely responsible for controlling the levels of H3K4me3 at the WRKY70 promoter. Furthermore, and important to this study, the level of WRKY70 transcript was decreased in response to dehydration stress ( Figure 5a). Previous microarray analyses have suggested that PtdIns5P negatively influences ATX1 activity [12]. Wholegenome analyses, however, could not provide answers to two critical questions: whether increased PtdIns5P affected the H3K4methylation profiles of ATX1 targets and whether association of ATX1 with target nucleosomes changed under elevated PtdIns5P. Here, we use the WRKY70 gene as a model to investigate ATX1 activity in vivo under dehydration-stress or under experimental conditions where we manipulated the levels of endogenous PtdIns5P. Selection of the WRKY70 gene was determined by three important factors: the WRKY70 nucleosomes are directly modified by ATX1 activity [24], its transcript levels are subject to regulation by dehydration stress (Figure 5a), and WRKY70 transcript levels were diminished upon addition of exogenous PtdIns5P [12].
WRKY70 expression and H3K4me3 profiles in OX-AtMTM1 and in mtm1 mutant cells Similar to the results with dehydration-stressed wild type leaves, lower WRKY70 transcripts were found in stressed OX-AtMTM1 leaves ( Figure 5a). However, the levels of WRKY70 transcripts in basal conditions were already significantly decreased and similar to the levels measured in wild type plants after dehydration stress, correlating with the increased levels of PtdIns5P ( Figure 5a). By contrast, in non-stressed mtm1 cells, the WRKY70 transcripts were not significantly different than in the wild type (p,0.12) but, importantly, the decrease in WRKY70 transcripts in response to the stress was suppressed (Figure 5a). Consistent with the level of the WRKY70 transcripts there was less H3K4me3 at the WRKY70 nucleosomes in basal AtMTM1-overexpressing plants, similar to the levels seen in dehydration stressed wild type plants (Figure 5b). In mtm1 plants where the increased PtdIns5P is abrogated in response to dehydration stress, the H3K4me3 levels were not significantly changed after stress (Figure 5b). These data show that the transcript levels of WRKY70 and the H3K4me3 patterns of its nucleosomes correlate with manipulated cellular PtdIns5P levels.
To mechanistically link changes in the H3K4me3 patterns, we assessed the localization of ATX1 at the WRKY70 promoter using the ChIP assay with anti-ATX1 antibodies. ATX1 was localized at the WRKY70 59-end but its presence decreased in response to dehydration stress of wild type plants. In non-stressed plants that over-express AtMTM1 the levels of ATX1 were already significantly decreased and reflected levels seen in stressed wild type plants. In contrast, in mtm1 knock out plants, ATX1 localized at the WRKY70 but was unaltered in response to dehydration stress ( Figure 5c). These data suggest that transcription of WRKY70 is regulated in response to dehydration stress in a PtdIns5P dependent manner. Changes in PtdIns5P appear to control the localization and the activity (H3K4me3) of ATX1 at the promoter. These changes are not a reflection of changes in ATX1 expression as it was not altered by stress nor was it altered in the various genotypes ( Figure 5d). These results illustrated that the ATX1 level at the WRKY70 gene locus changed in response to dehydration and depended on the presence of a functional AtMTM1. Decreased presence of ATX1 at the WRKY70 gene correlated with decreased transcript and H3K4me3 levels (Figure 5a-c). By contrast, transcript levels from two other ATX1 direct targets (AG and FLC) not involved in the dehydration-stress response and not found in the PtdIns5P-ATX1 co-regulated set [12] did not change upon stress in the OX-AtMTM1 or mtm1 backgrounds ( Figure 6).

Cellular distribution of ATX1
ATX1 can bind PtdIns5P [12] and here we analyze the cellular localization of ATX1 under conditions that elevated cellular PtdIns5P. Dehydration raises PtdIns5P but causes partial collapse of internal structures (not shown) complicating cytological analysis. As an alternative, distribution of GFP-tagged ATX1 (wild-type or various deletion forms) was followed in the presence of AtMTM1-RFP-fusion protein co-expressed in tobacco leaf-epithelial cells. ATX1-GFP is seen in the nuclei, in the cytoplasm, and at the plasma membrane (Figure 7a). However, in cells co-expressing ATX1 and AtMTM1 most of the ATX1 was absent from the nuclei (Figure 7b). Quantitation showed that ATX was present in the nuclei when expressed alone but when co-expressed with AtMTM1, its presence in nuclei decreased to about 20% ( Figure 8).
Next, we determined which of the ATX1 structural domains was involved in the re-localization. The N-terminal portion of ATX1 ( Figure 9) localizes in the nucleus (Figure 7a). Overexpression of AtMTM1 did not relocate the nuclear ATX1 (Figure 7b) suggesting that the N-terminal protein half was not involved. When the C-terminal portion of ATX1 was expressed alone, it localized in both the nuclei and in the cytoplasm; however, the nuclear signal decreased in cells co-expressing AtMTM1 (Figures 7a, b and Figure 9) implicating it in the relocalization. The ATX1-C terminal portion contains the ePHD and the SET structural domains (Figure 9), which were expressed separately to determine their individual roles. The SET domain expressed alone, or co-expressed with AtMTM1, remained in the cytoplasm (Figure 7a, b) implying that the SET domain was not involved in the ATX1 nuclear localization. However, the ePHD-GFP, which was present in the nucleus when expressed alone, became cytoplasmic when co-expressed with AtMTM1-RFP (Figure 7a, b) suggesting that the ePHD is involved in the nucleus-to-cytoplasm re-localization of ATX1 in cells overexpressing AtMTM1.
The role of the ePHD in the cytoplasmic retention of ATX1 was proven further by generating an ATX1 protein with deleted ePHD. When expressed alone, ATX1-DPHD-GFP localized in the nucleus; however, in cells co-expressing MYO-RFP, the GFP signal remained nuclear contrasting the redistribution seen with wild-type ATX1-GFP (Figure 7a, b). We conclude that the ATX1-ePHD is involved in the subcellular localization of ATX1 likely through binding the ligand PtdIns5P.

PtdIns5P and the role of ePHD
Two mutually non-exclusive mechanisms may account for the cytoplasmic retention of ePHD in OX-AtMTM1 cells: elevated PtdIns5P binds and keeps the protein in the cytoplasm and/or the AtMTM1 protein binds the ePHD. The ability of ePHD ATX1 to specifically bind PtdIns5P in vitro supports the first model [12]. To provide evidence in vivo, we generated AtMTM1 isoforms with a mutant active site to test whether production of PtdIns5P by AtMTM1 is required for the cellular localization of the ePHD. A protein with a mutation in the critically important Cys in the active site (mut1) displayed low (,5%) phosphatase activity against its substrate PtdIns3,5P 2 compared to the wild type. A second protein with a mutation in the adjacent Ser (mut2) retained ,30% of wild type activity (Figure 10c, d). Co-expression of ePHD with mut1-RFP failed to shift ePHD from the nucleus to the cytoplasm (Figure 10a, e) in contrast to experiments with the wild-type AtMTM1-RFP (Figure 10a, b). Importantly, co-expression of ePHD with mut2-RFP resulted in a distribution similar to the one observed with wild-type AtMTM1-RFP (Figure 10b, f). Because the effects of the AtMTM1 mutants on the cellular distribution of ePHD correlate with their lipid 39-phosphatase activity we conclude that production of PtdIns5P is required for retaining ePHD in the cytoplasm. These results do not preclude binding of ePHD and AtMTM1, a possibility that remains to be explored.

Discussion
By analyzing the H3K4me3 pattern of the WRKY70 nucleosomes, we followed changes in ATX1 activity in vivo under stress or in the presence of overexpressed AtMTM1. The statistically significant decrease in WRKY70 nucleosomal H3K4me3 in dehydrated wild type cells, in parallel with an increase in cellular PtdIns5P, reflected diminished ATX1 activity. Increasing cellular PtdIns5P by overexpressing AtMTM1 also led to a diminution of  (Figure 5c), suggesting that PtdIns5P may regulate ATX1 localization. The cytoplasmic localization of ATX1 observed under increased cellular PtdIns5P was in accordance. The endogenous PtdIns5P level was not significantly lower than in the wild type suggesting that PtdIns5P/ PtdIns3,5P 2 homeostasis was not severely disrupted in AtMTM1depleted cells. Under stress, however, the mtm1 mutant cells failed to elevate PtdIns5P underscoring the role of AtMTM1 in producing PtdIns5P. These data are consistent with a causal role for a dehydration-induced PtdIns5P regulating the localization of ATX1. Furthermore, H3K4me3 and transcripts from WRKY70 did not change significantly in mtm1 cells indicating a requirement for PtdIns5P. These results, however, do not preclude an interaction of AtMTM1 with ATX1-SET. An intriguing feature of all known active and inactive myotubularins, including the two Arabidopsis proteins, is the presence of a conserved domain, SID (Set interacting domain) that specifically binds the SET [25] domain. The SID domain is conserved in all known active and inactive members of the myotubularin family [25,26]. The biological significance of this interaction, however, is under debate, as crystallographic studies have shown that the SID peptide is buried inside the tertiary structure of the protein arguing against its binding with other proteins [27]. However, mutations in SID impeding its binding to SET result in abnormal growth and leukemia [25,28] and the inactive myotubularin Sbf1 has been isolated as a component of the Trx complex, physically and functionally linked with the trithorax protein in Drosophila [29]. An important implication of these results is that myotubularins, active or inactive, may have nuclear localization and could participate in chromatin remodeling [30].
Lack of a visible phenotype in mtm mutants and the level of endogenous PtdIns5P not significantly lower than in fresh wild type cells suggest that the AtMTM1 function is not essential for the cellular homeostasis in non-stressed cells, in stark contrast with humans where myotubularin deficiency is implicated in severe  . Sub-cellular distribution of ATX1 co-expressed with AtMTM1. ATX1 and its derivatives were expressed as ATX1-GFP fusion proteins (green signal) while AtMTM1 was expressed as an RFP fusion protein (red signal). a) all images in this column illustrate cells expressing ATX1 or derivatives alone; b) all images in this column illustrate cells co-expressing ATX1 (or ATX1 derivatives) with AtMTM1. Arrows point to nuclei in cells expressing a GFP-fusion protein. ATX-N is the N-terminal ATX1 portion, ATX1-C is the C-terminal ATX1 portion containing the ePHD and the SET domains (SF7). Arrows point to nuclei; Bars are 50 mm. doi:10.1371/journal.pone.0013396.g007 muscle dystrophy, neuronal diseases, and leukemia [1,3,31]. Despite highly conserved in evolution, molecular tools are adapted and used in kingdom and species-specific manners [32]. Two phosphatases that can convert PtdIns(4,5)P 2 to PtdIns5P were discovered in human cells [33] but whether a similar route for generating PtdIns5P exists in plants is unknown.
As a substrate of AtMTM1, it is possible that PtdIns3P and PtdIns(3,5)P 2 cellular levels were also affected. PtdIns3P is associated with late endosomes, multivesicular bodies and prevacuolar membranes [8] suggesting a possible indirect involvement of PtdIns3P in the reported events. However, the ability of the ePHD finger to bind PtdIns5P suggests that PtdIns5P is likely to be the endogenous regulator of ATX1 localization.
Most likely, AtMTM1 has a broader role than just regulating ATX1 consistent with our results from whole-genome transcrip-tion profiling of PtdIns5P-treated [12] and of OX-MTM1 genomes [19]. Our data suggest that ATX1 and AtMTM1 (PtdIns5P, respectively) function in distinct pathways, which partially overlap only along branches that target a shared set of genes. In agreement, ATX1 regulated stress-response genes that are not involved in the PtdIns5P/MTM1-pathway (i.e. the NCED3 gene, not shown); the developmentally regulated AG and FLC genes (Figure 6a, b) are efficiently transcribed during dehydration stress, illustrating different patterns of ATX1 activity, and perhaps, regulation. The dynamic subcellular localization of ATX1 in nonstressed root cells [12] suggested that, most likely, ATX1 was responding to a variety of cellular signals. Furthermore, the FLC gene, which is under the control of complex genetic, epigenetic, and growth-signaling mechanisms, re-establishes anew the H3K4me3 modification pattern at the transcription start site in each developmental cycle. H3K4me3 levels correlate with reactivated FLC transcription, which persists until transition to flowering [34]. It is plausible, then, that the FLC H3K4me3 and transcript levels established early in development might not be affected by ATX1 presence at the FLC locus later in life. On the other hand, the WRKY70 gene displaying dynamic H3K4me3 changes in responses to biotic [24] and dehydration stresses (this study) showed a clear dependence on ATX1 presence.
Production of PtdIns5P is required for retention of ePHD in the cytoplasm (Figure 10). The ePHD found in all animal and plant members of the Trithorax family belongs to a family distinct from the families of the PHD domains of the chromatin proteins ING, NURF, CHD, ORC1, or ACF. Some PHDs bind PtdIns5P [5,6] as well as H3K4me3 [35,36]. Whether ePHD binds histones is unknown and a role for ePHD at chromatin is not excluded. Here, we report a novel function for the ePHD domain of a chromatinassociated protein as a mechanism regulating ATX1 activity. The shared ATX1-AtMTM1 dehydration-responding fraction identified by microarray analysis provided further evidence for a pathway involving an epigenetic regulator and a lipid phosphatase producing PtdIns5P (microarray expression data are deposited in the NCBI series GSE15577). The genes co-regulated by ATX1 and by exogenously supplied PtdIns5P [12] further support the biological relevance of the ATX1 binding to this lipid ligand. A bound lipid may activate or deactivate the protein by sequestering it in a cellular sub-compartment, by inducing a conformational change, or by triggering phosphorylation of another downstream target [37]. It is possible that PtdIns5P can regulate ATX1 activity in both nuclear and cytoplasmic contexts and that different mechanisms may operate in the two cellular compartments.

Plant material
Wild type and mutant plants were grown in soil under the same controlled daylight environmental conditions. Leaves were detached from plants and immediately weighed to determine their fresh weight. Recovery upon dehydration was established by measuring turgidity as described in [38]. Briefly, dehydration stressed leaves were submerged in deionized water for 24 h, blotted dry and weighed to determine their TW (turgid weight).
Transgenic plants were generated by transformation with binary vectors. Binary plasmids were transformed into chemically competent Agrobacterium tumefaciens strain C58C1 by incubating DNA with Agrobacteria on ice for 5 minutes, freezing in liquid nitrogen for 5 minutes and heat shock at 37uC for 5 minutes. The cells were allowed to recover in growth medium with shaking for 2 hours at room temperature and plated on selection medium containing rifampicin, gentamycin and a third antibiotic for plasmid selection. Agrobacteria selected for transformation were used to transform Col-O plants using a floral dipping method as described [39].
To generate stably AtMTM-over-expressing lines the full-length sequence of At3g10550 was amplified using PCR-specific primers containing the attB1 and attB2 sequences and recombined by the attB x attP (BP) reaction into the pAlligator vector [40] to generate an N-terminal HA-tagged expression clone for stable expression in plants. The entry clone was also recombined into the pB7WGR2.0 vector [41] to generate an N-terminal RFP fusion expression clone for expression in plants. The binary vector pAlligator2-AtMTM construct containing the coding sequence of the full length AtMTM gene cloned in-frame with an N-terminal HA tag and driven by the 35S promoter was used for transformation in Arabidopsis. Transgenic plants were selected by GFP-fluorescent seeds where expression of the GFP marker was driven by a seed-specific promoter. The T-DNA insertion line SALK_029185 was obtained from the SALK institute. Homozygous (mtm) lines were screened by segregation, PCR, and RT-PCR analyses; atx1 mutant lines were as described [13].
The ePHD domain of ATX1 was PCR amplified using the forward primer phd-attB1 and the reverse primer atx1-attB2.
The SET domain of ATX1 was PCR amplified using the forward primer set-attB1 59-GGGGACAAGTTTGTACAAAAAAGCAGGCTTAAT-GAATACTCCAAGCAAC-39 and the reverse primer atx1-attB2.
The gateway PCR products were cloned into the entry vector pDONR221 and the entry clones recombined into expression vectors pB7FWG2,0 vector [41] for C-terminal GFP-fusion and expression in plants; pDESTT17 vector (Invitrogen) for N-terminal fusion to a 66 his tag and inducible expression in bacteria.
Generation of coding sequences of AtMTM1 with mutations in the active site was carried out by a site directed mutagenesis approach using an overlap extension PCR approach [42]. The following primers overlapping by 42 bp in reverse orientation were designed to contain point mutations of residues of the conserved CX5R phosphatase domain active site: (PTP-fwd 59-CTTGTGCACAGCAGTGATGGATGGGTCA-GAAC-39 and PTP-rev 59-GTTCTGACCCATCCATCACAGCAGTGCA-CAAG-39). In a first step the rid-attB1 primer 59-GGGGACAAGTTTGTACAAAAAAGCAGGCTTAAT-GACGCCGCCGAGACCACCG-39 and the PTP-rev primer were used to obtain a 757 bp PCR product A.
The PTP-fwd primer 59-CTTGTGCACAGCAGTGATG-GATGGGTCAGAAC-39 and a AtMTM-attB2 primer 59-GGGGACCACTTTGTACAAGAAAGCTGGGTATTTAGG-TTGGAAATAGCTATCG-39 were used to amplify a 1228 bp PCR product B. The overlapping PCR products A and B were used as templates for a second step overlapping PCR, in the presence of primers rid-attB1 and AtMTM-attB2 to generate a 1953 bp gateway PCR product. After recombination cloning into the pDONR221 vector, several entry clones were sequenced and the two active site mutations of interest defined as mut1 and mut2 were selected for further studies.

Phosphatase activity
Phosphoinositide 3-phosphatase assays were performed using the malachite green assay [43,44] Clones obtained as described above were introduced into the pGEX4T-1 vector (GE Healthcare), expressed as GST-fusions in E. coli BL21 cells, and used to assay the enzyme activity. Recombinant affinity-column purified GST-tagged proteins were incubated with substrates (2.5 mmoles per experimental point) in 50 ml assay buffer (50 mM TrisHCl, pH 8.0, 100 mM KCl and 2 mM DTT). Di-C8 phosphoinositides (Echelon Biosciences Inc. catalogue numbers P-3008, P-5008, and P-3508) were used as substrates and PTEN lipid phosphatase (Echelon Biosciences Inc. catalogue number E-3000) was used as a positive control. Following incubation at 37uC for 30 minutes the reactions were quenched by the addition of 20 ml of 0.1 M Nethylmaleimide and spun at 18,000 g for 10 minutes. Twenty-five microliters of the supernatant was added to 100 ml of malachite green reagent and vortexed. Samples were allowed to incubate for 40 minutes for color development before measuring absorbance at 620 nm. Inorganic phosphate release was measured by a standard curve of KH 2 PO 4 in distilled water.

Tobacco transient assays
Agrobacteria colonies containing binary plasmids for plant transformation were grown overnight in 10 mls of media with antibiotics. The cells were collected and re-suspended in an equal volume of induction medium (60 mM K 2 HPO 4 , 33 mM KH 2 PO 4 , (NH 4 )SO 4 , 1.7 mM Na Citrate.2H 2 O, 10 mM MES, 1 mM MgSO 4 , 0.2% Glucose, 0.5% Glycerol, antibiotics and 50 mg/ml of acetosyringone), and incubated with shaking for 6 hours at 30uC. The cells were re-suspended to an OD of 0.8 in infiltration medium (0.56 MS, 10 mM N-moropholino-ethanesulfonic acid, 150 uM acetosyringone) and used to infiltrate the abaxial surface of N. benthamiana leaves. After 40 hours detection of expression was conducted by laser scanning confocal microscopy using 488-and 633-nm excitation and two-channel measurement of emission, 522 nm (green/GFP) and 680 nm (red/chlorophyll). RFP was detected by excitation at 540 nm and emission at 590 nm.

Real time quantitative RT-PCR analysis
RNA for qRT-PCR (quantitative real-time reverse transcription-polymerase chain reaction) was isolated with TRIZOL (Invitrogen) and purified with a RNeasy plant mini kit (Qiagen, catalogue number 74903). For the first strand cDNA synthesis 8 mg total RNA was treated with DNase I, extracted with phenol and chloroform, precipitated with ethanol, followed by the addition of oligo (dT) and superscript III reverse transcriptase (Invitrogen). Real-time PCR analysis was performed using the iCycleriQ real-time PCR instrument (Bio-Rad) and iQ SYBR Green Supermix (BioRad). The relative expressions of specific genes were quantitated using 2 2DDCt calculation, where DCt is the difference in the threshold cycles of the test and housekeeping gene ACTIN7. DCt is the threshold cycle of the target gene subtracted from the threshold cycle of the housekeeping gene. The mean threshold cycle values for the genes of interest were calculated from three experiments. The following primers were used: Chromatin Immunoprecipitation (ChIP) assays A protocol previously described [45] was used. Anti-H3K4me3 antibodies (Millipore, catalogue number 07-473) were used. Negative control samples were treated in the same way except that antibodies were not added. Each immunoprecipitation was performed in at least three separate experiments. Calibration curves were constructed to determine optimal amounts of chromatin to be used in each experiment and to ensure equivalent amounts of starting material. The PCR products were amplified using the primer sequences for WRKY70: F (59-agcaactcctctctcaacccg-39); R: (59-ccattgacgtaactggcctga-39); ACTIN7: F (59-ggtgaggatattcagccacttgtctg-39); R: (59-tgtgagatcccgacccgcaagatc-39). Mass assay; TLC; HPLC; PIP4K assay; 59-phosphatase assay Leaves were carefully excised from growing plants and immediately weighed. Total lipids were extracted from leaves either immediately or after dehydration stress (left on the bench for 2 hours at room temperature resulting in a weight loss of 30-40%) using ice-cold chloroform (250 ml) and methanol (500 ml) in a UCD-200TM bioruptor (Tosho Denki Co., Japan) operating at full power for 2 minutes. After storage at 220uC overnight another sonication step ensured complete leaf tissue disruption. Water (250 ml), 2.4 M HCl (200 ml) and CHCl 3 (250 ml) were subsequently added to induce phase separation. After thorough mixing and centrifugation (5,000 g for 2 minutes at room temperature), the lower chloroform phase was washed once with theoretical upper phase and then dried in a speedvac at room temperature. PtdIns5P was quantitated in a neomycin affinity chromatography enriched phosphoinositide fraction using a radioenzymatic assay [6,46]. The mass of PtdIns5P was normalized according to the starting leaf mass. Deacylated radiolabeled phospholipids were also analyzed by HPLC [17]. Purified yeast YNK-5 phosphatase was used to de-phosphorylate the isolated PtdInsP 2 formed from the phosphorylation of endogenous leaf PtdInsP by PIP4Ka [17]. The TLC plate was developed once in chloroform/methanol/25% ammonia/water (ratio: 45/35/2/8, v/v/v/v). After air-drying for 30 minutes the plate was exposed to a phosphorimaging screen or to X ray film. HPLC analysis used a gradient of 0-1 M ammonium phosphate pH 3.85 with a flow rate of 1 ml/minute. Direct on-line radioactivity detection every 15 seconds avoided the requirement for fraction collecting. Correct peak identification was achieved using in-house radioactively labelled standards. The method has been described previously [46].

Statistical analysis
A Wilcoxon signed t-rank test was applied for determining significant differences [23] employing significance threshold of P,0.05.
As biological material was taken from different plants, the measurements represented statistically relevant averages of PtdIns5P content in all the tested samples.