A cytoplasm-specific activity encoded by the Trithorax-like ATX1 gene.

Eukaryotes produce multiple products from a single gene locus by alternative splicing, translation or promoter usage as mechanisms expanding the complexity of their proteome. Trithorax proteins, including the Arabidopsis Trithorax-like protein ATX1, are histone modifiers regulating gene activity. Here, we report that a novel member of the Trithorax family has a role unrelated to chromatin. It is encoded from an internal promoter in the ATX1 locus as an isoform containing only the SET domain (soloSET). It is located exclusively in the cytoplasm and its substrate is the elongation factor 1A (EF1A). Loss of SET, but not of the histone modifying ATX1-SET activity, affects cytoskeletal actin bundling illustrating that the two isoforms have distinct functions in Arabidopsis cells.

INTRODUCTION

The highly conserved (∼150 amino acids) SET [Su(var)3-9, E(z), Trithorax] domain is found in chromatin proteins involved in both gene repressing and gene activating complexes. The SET domains have an intrinsic preference for specific histone lysine-residues (1). A methylation sign at a particular histone lysine defines the transcriptional state of the involved gene (2). SET domain-containing proteins of the TRITHORAX family, including ATX1, specifically methylate lysine 4 of histone H3 (H3K4me) (3,4). Loss of ATX1-function has pleiotropic effects in Arabidopsis affecting development, organogenesis (5), and ability to respond to biotic (6) and abiotic (7) stresses. ATX1 is not involved in a genome-wide H3K4 methylation but, instead, targets specific genes (8). Increased levels in nucleosomal H3K4me3 marks of ATX1-regulated genes correlates with elevated transcript levels (3,8) defining ATX1 as a histone modifier and as a positive effector of gene expression.

Trithorax proteins carry signature structural domains [FYRN-FYRC (DAST), PHD and SET] plus additional clade-specific domains (9,10). A combinatorial assembly of various peptide domains generates possibilities for diversification and precision of function (11). The best studied SET domain proteins function in the nucleus. An earlier study has shown that a Polycomb group protein, Exh2, could assemble a methyltransferase complex in the cytoplasm of T cells (12) but Ezh2 was not characterized as a cytoplasm-specific isoform and its substrate in the cytoplasm is unknown. Here, we describe a cytoplasmic isoform of ATX1 containing only the SET domain (soloSET). We identified the elongation factor 1A (EF1A), its Lys396 (K396) in particular, as the cytoplasmic substrate for soloSET. Methylated lysines at this position are also found in maize and yeast (13,14) suggesting an evolutionary conservation of this modification in eukaryotic EF1A. However, whether/which lysines of the Arabidopsis EF1A protein are methylated has not been reported.

In addition to its roles in polypeptide chain elongation, eEF1A appears to be unique among all the translation factors by the diverse functions it has been ascribed outside of protein synthesis [reviewed in ref. (15)]. Due to its ability to bind and bundle actin, EF1A is considered a key factor in cytoskeleton organization (16–18). In plant cells actin bundling is used to build, position, and stabilize the main routes for organelle transport over long distances (19). Although methylated EF1A lysines have been connected to actin organization and bundling (13,14), the methyltransferases involved and the contribution of any particular methylated lysine are still unknown. Here, we show that methylation of the EF1A K396, located in the structural domain involved in actin bundling, is dependent on the activity of the SET domain encoded by the chromatin-modifying gene ATX1. This finding reveals an unsuspected role for a gene known so far to encode only an epigenetic factor.

MATERIALS AND METHODS

Arabidopsis thaliana ecotype Ws-2 and the earlier described atx1 mutant line (5) were used. All primers used in this study are summarized in Supplementary Table S1.

Construction of GUS—expressing vectors and transgenic Arabidopsis lines

The Promoter-SET/GUS (uidA) (PSET::GUS) reporter was constructed by cloning 476bp fragment upstream of the SET domain sequence using the PstI containing forward and the NcoI containing reverse primers, respectively (Supplementary Table S1 and Figure S1). The product of the Taq polymerase PCR reaction (Invitrogen, http://www.invitrogen.com) was ligated into the pGEM-T vector system (PROMEGA, http://www.promega.com) and after sequence verification was sub-cloned into the pCambia1303 vector within the PstI and NcoI sites, upstream of the β-Glucuronidase (Gus) gene. The PDIST::GUS construct was cloned as above, except that the forward and reverse primers (Supplementary Table S1) amplified a 1026bp fragment further upstream (Supplementary Figure S1) to be tested as a promoter. The PATX1::GUS construct is described in ref. (20). Transgenic Arabidopsis lines expressing each of the constructs were generated by Agrobacteria mediated transformation (21) and transformed lines were selected for hygromycin resistance.

Construction of GFP-soloSET, GFP-EF1A, RFP-soloSET expression vectors and tobacco transient expression assays

GFP-soloSET and RFP-soloSET used in the transient expression assays were generated by recombining the pDONRT-SET vector with pB7FWG2.0 and pB7WGR2.0 expression vectors, respectively (22). Agrobacteria colonies containing binary plasmids for plant transformation were grown overnight in 10 ml of media with antibiotics. The cells were collected and re-suspended in an equal volume of induction medium (60 mM K2HPO4, 33 mM KH2PO4, (NH4)SO4, 1.7 mMNa Citrate·2H2O, 10 mM MES, 1 mM MgSO4, 0.2% glucose, 0.5% glycerol, antibiotics and 50 µg/ml of acetosyringone), and incubated with shaking for 6 h at 30°C. The cells were re-suspended to an OD600 of 0.8 in infiltration medium [0.5× MS, 10 mM N-moropholino-ethanesulfonic acid (MES), 150 µM Acetosyringone] and used for infiltrating the abaxial surface of Nicotiana benthamiana leaves. After 40 h, 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. Protein expression was confirmed by western blotting where necessary. In co-transformation experiments for GFP-soloSET with RNAi-SET or of GFP-EF1A with RFP-soloSET, equal volumes of Agrobacteria in infiltration medium, at OD600 of 0.8 were mixed and used for infiltration. Separate plants were used for control infiltrations.

Generation of transgenic plants expressing RNAi-SET

The SET domain of ATX1 was amplified using the gateway primers attB1SETFWD and attB2-ATX1. The PCR product was cloned by using the BP recombination reaction (Invitrogen) into the entry vector pDONRT207. After verifying the sequence integrity, the pDONRT-SET vector was recombined with the destination vector pFGC5941 (GenBank accession AY310901) by the LR reaction (Invitrogen) to generate the RNAi binary vector. The vector structure is illustrated in Supplementary Figure S2a. RNAi-SET transgenic A. thaliana WS-2 lines were produced by Agrobacteria tumefaciens C58C1 mediated transformation (21). Transgenic lines were selected for basta resistance. The RNAi-SET effects were ATX1 specific as mRNAs from two other SET domain genes, ATX2 and CLF, were present in cells where ATX1 transcripts were not detected (Supplementary Figure S2b).

Total RNA was extracted from the examined Arabidopsis plants using the TRIzol method (23). First-strand cDNA synthesis was preformed on 500 ng of RNA using the M-MLV System for RT–PCR (Invitrogen).

Two-dimensional gel electrophoresis and isoelectric focusing

Arabidopsis rosette leaves (0.3 g) were pulverized to a fine powder with liquid nitrogen with a mortar and pestle. Total protein extracts were prepared from the same samples that were used for mRNA purification by the TRIzol reagent (23). The proteins were re-suspended in 0.3 ml of isoelectric focusing (IEF) sample extraction media (8 M urea, 2 M thiourea, 2% [w/v] CHAPS, 2% [v/v] Triton X-100, 50 mM dithiothreitol) and were resolved in IEF and SDS-PAGE according to an established protocol (24) using the Bio-Rad system. Following SDS–PAGE, gels were washed in deionized water and stained with Coomassie or reacted with specific antibodies. Protein concentrations were determined using a protein assay Bio-Rad kit. Antibodies against ATX1 (raised in rabbit sera, GenScript, SC1031) and anti-trimethyl-H3K4 (Abcam, ab8580-100) were used. A Q-TOF Ultima tandem mass spectrometer (Waters) with electrospray ionization was used to analyze the eluting peptides.

Visualization of filamentous actin

Actin was labeled with rhodamine phalloidin following a recent protocol (25) with modifications. Leaves collected from 3-week-old A. thaliana plants were fixed with 4% paraformaldehyde freshly prepared in PEM buffer (100 mM PIPES, 10 mM EGTA, 5 mM MgSO4, pH 6.9) for 1 h at room temperature, rinsed with wash buffer (0.1% Triton X-100 in PEM) and three times with PEM. The samples were treated with 1% cellulase (Sigma) and 1% driserase (Sigma) in PEM for 20 min at room temperature, rinsed with wash buffer and treated with 1% glycerol in wash buffer for 2 h. After rinsing with PEM, samples were stained with 100 nM rhodamine-phalloidin (Invitrogen, Molecular Probes, USA) in PEM for 30 min. After rinsing with wash buffer, samples were mounted on a glass slide in PEM and observed under a confocal microscope (model LSM510; Carl Zeiss).

For analyses of transcription factor binding motifs, information available in the database PLACE was used: http://www.dna.affrc.go.jp/PLACE/signalscan.html.

Mass spectrometry and data analysis

The stained bands were excised and subjected to LC/MS as described (26). Briefly, gel pieces were digested by trypsin (no.V5111, Promega, Madison, WI, USA) and digested peptides were extracted in 5% formic acid/50% acetonitrile and separated using C18 reversed phase LC column (75 micron × 15cm, Pepmap 300, 5 micron particle size) (Dionex, Sunnyvale, CA, USA). A Q-TOF Ultima tandem mass spectrometer (Waters) with electrospray ionization was used to analyze the eluting peptides. The system was user-controlled employing MassLynx software (v 4.1, Waters) in data-dependant acquisition mode with the following parameters: 0.9-sec survey scan (380–1900 Da) followed by up to three 1.4-s MS/MS acquisitions (60–1900 Da). The instrument was operated at a mass resolution of 8000. The instrument was calibrated using the fragment ion masses of doubly protonated Glu-fibrinopeptide. The peak lists of MS/MS data were generated using Distiller (Matrix Science, London, UK) using charge state recognition and de-isotoping with the other default parameters for Q-TOF data. Data base searches of the acquired MS/MS spectra were performed using Mascot (Matrix Science, v1.9.0, London, UK). The NCBI non-redundant database (2 010 130–10 386 837 sequences 3 543 419 944 residues) was used restricted to A. thaliana. Search parameters used were: no restrictions on protein molecular weight or pI, enzymatic specificity was set to trypsin with up to three missed cleavage sites, carbamidomethylation of C was selected as a fixed modification, and methionine oxidation along with mono-, di- and trimethylation of lysine were allowed as variable peptide modifications. Mass accuracy settings were 0.15 Da for peptide mass and 0.12 Da for fragment ion masses.

RESULTS

A soloSET domain isoform is generated from an internal promoter in the ATX1 locus

In addition to the full-size ATX1 protein, immunoblot assays of total protein extracts from wild-type Arabidopsis leaf cells with anti ATX1-antibodies consistently revealed presence of a lower-size ∼22 kDa protein band (Figure 1a). N-terminal Edman sequencing identified the sequence MNTPSNIL, which matches the 5′-end of the SET domain, indicating that the 22 kDa protein band represents the SET domain of ATX1. Unexpectedly, this band was present also in blots from the atx1 mutant cells (Figure 1b). As the T-DNA insertion in the atx1 mutant line disrupts the ATX1 mRNA (Figure 1c), the 22 kDa protein from the C-terminal gene region in atx1 cells apparently originated from a message generated downstream of the insertion (Figure 1b–d). To test for the existence of a promoter that could account for such a message, we mapped various transcripts from the ATX1 region.

Figure 1.

SoloSET isoform produced from an internal ATX1 promoter. (a) Western blots of total cellular extracts from WT cells with the antiATX1 antibody. The 116 kDa corresponding to full-size ATX1 is indicated by an arrow. The 22 kDa (soloSET) protein band is indicated by the arrowhead. Sample dilutions (0-, 2-, 4-fold) are indicated on top of lanes. Two nonspecific signals are marked with asterisks. (b) Western blot analysis of total cellular extracts from atx1 leaf cells with antiATX1 antibody. The ∼63 kDa protein band (indicated by arrow) apparently represents a truncated protein. The 22 kDa soloSET protein is indicated by the arrowhead. Other annotations are as in (a). (c) Schematic representation of the ATX1 locus. The site of Ti-DNA insertion on atx1 mutants is shown by the black triangle. The empty boxes under the gene indicate the size of protein products generated from the gene. (d) Tested regions amplified with specific primers. The boxes represent the size of the expected bands amplified with the primers indicated in the boxes (Supplementary Table S1).

Primers overlapping regions upstream of the T-DNA insertion amplified the expected bands from both wild-type and atx1 mutant cells illustrating presence of messages from the sequences in the N-terminal half of the gene (Figure 1c and d panels Tu-DAST, Tu-PWWP and DAST). However, sequences flanking the T-DNA insertion site were amplified only from WT mRNA but not from the atx1 template (panel DAST-PHD). Similarly, specific primers overlapping sequences downstream of the insertion site amplified a band only from the wild type (panel PHD-SET). However, primers specific for the SET domain revealed presence of transcripts from both the WT and the atx1 samples (panel SET) supporting the existence of a promoter that drives transcription from SET sequences separately from their transcription in the context of the full ATX1 mRNA.

To validate the existence of such a promoter by another approach, a DNA region upstream of the MNTPSNILS sequence was tested as a promoter for the GUS gene (PSET::GUS in Supplementary Figure S1a and b). A 1024 bp DNA sequence from a region further upstream, carrying a putative TATA-box (the distal promoter) was also tested for promoter activity (PDIST::GUS). Transgenic lines stably expressing each of these constructs revealed that PSET drove GUS expression in all twelve independently transformed lines indicating that the 0.5 kb DNA upstream of SET does function as a promoter (Supplementary Figure S3a and b). In contrast, the distal ∼1 kb sequence did not activate GUS expression (11 PDIST::GUS lines were analyzed) despite the presence of a TATA-box like sequence. Comparing lines expressing GUS under the native promoter for the full-size ATX1 gene (PATX1::GUS) revealed partially overlapping expression domains and some differences as well (Supplementary Figure S3a and c; 15). PSET::GUS was strongly expressed in young seedlings, while PATX1::GUS staining was strong in cotyledons, but weak in the first true leaves, except at the hydothodes. Later in development, PSET::GUS narrowed its expression domains but remained prominent in cells at attachment sites of organ to stems; ATX1 is ubiquitously expressed, particularly strong in the vasculature of leaves and stems (Supplementary Figure S3b and c; 15).Collectively, our results indicate that a bona fide promotor exists upstream of the SET-ATG codon driving strong SET mRNA expression in cells, particularly at the sites of organ attachment.

Comparative analysis of the putative transcription factor-binding sites at the ATX1 and soloSET promoters using the PLACE database (‘Materials and Methods’ section) revealed a larger number of putative TF-binding sites upstream of ATX1 suggesting that ATX1 could be subjected to a broader array of regulatory signals than soloSET. However, shared putative binding sites for factors regulated by abiotic stress and hormonal signals were also recognized.

Subcellular localization of soloSET and effects of RNAi-SET expression

Transient expression of a soloSET-GFP fusion protein in tobacco leaf cells revealed that it was present only in the cytoplasm (Figure 2a). This is in contrast to ATX1-GFP, which localizes in the nucleus (Figure 2b) but dynamically shifts between the nucleus and the cytoplasm in response to signals (27). To asses a possible role of soloSET, we constructed transgenic lines expressing RNAi-SET. Transgenic plants displayed strong phenotypes: precocious flowering (occurring at a stage with only four true leaves), asymmetric rosettes (note different sizes of leaves 3 and 4), aberrant flowers and chlorosis (Figure 2c). Aberrant flowers and precocious flowering were displayed also by atx1 mutants (5), although never as early as at a four-leaf stage. Also, we have not observed chlorosis under normal watering and long day conditions in atx1 mutants. Despite these perceived differences, however, we concluded that the visible phenotypes, alone, could not be reliable indicates of soloSET function. Consequently, we focused on determining its role at the cellular and molecular levels.

Figure 2.

Cytoplasmic localization of soloSET and transgenic plants expressing RNAi-SET. (a) Transiently expressed GFP-soloSET fusion protein in tobacco leaf cells is detected only in cytoplasm. Arrows point to perinuclearly localized green signal. (b) Nuclearly localized full-size ATX1-GFP fusion protein is shown for comparison. Chloroplasts are shown in red (chlorophyll autoflorescence). Bars are 20 µm. (c) transgenic Arabidopsis plants expressing RNAi-SET showing precocious flowering (at the stage of only four true leaves), chlorosis in sepals and inflorescences, and a flower with abnormal short stamen (arrowhead) and asymmetric petals (arrows).

First, we asked whether RNAi-SET affected the levels of ATX1 and/or soloSET transcripts. Using primers overlapping various regions of the ATX1 gene, we show the results from six independently transformed lines analyzed by RT–PCR (Figure 3a and b). As expected, the primers failed to amplify all regions in the ATX1 mRNA, except at the 3′-end. An exception was line 4 where weak bands were detected from amplification of regions upstream of the Ti-insertion. Most unexpected, however, was the recovery of the SET-domain sequences from all transgenic samples indicating that SET domain transcripts were produced in all RNAi-SET lines (Figure 3a). These results indicated that production of ATX1 transcripts was knocked down in the presence of RNAi-SET, while transcripts from its SET-domain region (soloSET) were not substantially decreased.

Figure 3.

Transcripts produced from the ATX1 locus in RNAi- SET transformed plants. (a) RT–PCR analysis of ATX1 specific transcripts. The peptide domains encoded by the tested regions are shown to the right of the panels. Numbers of top of lanes indicate independently transformed lines tested for expression of ATX1 and its derivatives. As controls, all primers were tested for the specific fragment amplification in wild-type leaf (L) and flower (F) tissues. SET domain transcripts detected in all transgenic lines are shown in the panel indicated as SET. Transcripts from upstream ATX1 regions were not detected, except for weak bands corresponding to the Tu-PWWP and DAST regions in line 4. (b) Positioning of the primers used to amplify specific regions and of the produced transcripts are illustrated.

To pursue further the fate of the soloSET messages we examined the RNAi-SET effects on the production of soloSET protein. To this end, total protein extracts from WT cells were resolved by two-dimensional (2D) gel electrophoresis and reacted with antiATX1 antibodies. The full-size ATX1 and the soloSET proteins produced in wild-type cells are shown (Figure 4a). The nature of the spots indicated by the arrowhead and the arrow was confirmed by mass spectroscopy (MS) (Supplementary Figure S4a and b). In western blots of extracts from RNAi-SET expressing cells both spots were missing indicating that neither ATX1 nor soloSET were produced (Figure 4a). A few other proteins reacting non-specifically with the ATX1 antibody retained their signals in the SET-RNAi sample illustrating that the RNAi-SET effects upon ATX1 and soloSET were specific. The results indicate that, ultimately, in the presence of RNAi-SET, production of both ATX1 and soloSET is suppressed but suppression is achieved by different mechanisms: by knocking down ATX1 transcripts but by disrupting the production of the soloSET protein from its transcripts. What determines the different modes of ATX1 and soloSET suppression by SET-RNAi remains to be elucidated but it is tempting to speculate that they reflect different mechanisms regulating the production of the full-size ATX1 and of its shorter soloSET version.

Figure 4.

RNAi- SET expressing cells inhibit synthesis of ATX1 and soloSET proteins. (a) Two-dimensional-electrophoretically resolved proteins from WT and RNAi-SET expressing cells followed by immunostaining with the antiATX1 antibody. Full-size ATX1 (arrowhead), soloSET (arrow) and an apparently degradation product (white arrow) are present only in the WT. Proteins non-specifically binding the antiATX1 antibody were not detectably affected. The two bottom panels are Coomassie stained gels run in parallel with the top pair. MW scale is shown on the vertical coordinate. Horizontal arrows show the pH gradient. (b) GFP-soloSET fusion protein transiently expressed in tobacco leaf cells. Chloroplasts are shown in red as a background for transformed cells (green). (c) RNAi-SET co-transformed with GFP-soloSET eliminated the GFP-signal consistent with a suppressed production of soloSET. Bars are 50 µM.

The ability of RNAi-SET to suppress soloSET was confirmed also by an independent approach. A soloSET-GFP fusion construct produced a strong fluorescent signal when expressed alone in tobacco cells (Figure 4b). However, the signal was completely eliminated when soloSET-GFP and RNAi-SET were co-transformed (Figure 4c). We conclude that soloSET-GFP production was effectively suppressed by RNAi-SET.

SoloSET interacts with EF1A

In an earlier screening of a yeast-two-hybrid expression library with the SET-domain sequence as bait, we have identified the cDNA from the At5g60390 gene as a SET-domain binding candidate. At5g60390 encodes the elongation factor 1A (EF1A) and, here, we tested the interaction of the two proteins by the pull-down assay. A GST-SET fusion protein was produced, immobilized on a column and reacted with protein extracts from wild-type leaf cells. A band of ∼50 kDa, corresponding to the molecular size of the Arabidopsis EF1A protein (Figure 5a), was detected by immunoblotting with antibodies against the maize protein. The band was positively identified as EF1A by MS (Supplementary Figure S5). We note that the SET-peptide selectively picked EF1A amongst the proteins present in whole cellular extracts confirming, thus, the interaction of EF1A and SET-ATX1 and their strong preference for each other as binding partners.

Figure 5.

Interaction of the SET domain with eEF1A. (a) GST-SET pull-down of total protein extract from WT Arabidopsis leaves followed by immunoreaction with the maize antiEF1A antibody. The 50 kDa protein corresponds to the Arabidopsis EF1A (arrow). (b) Transiently co-expressed GFP-EF1A (green) and RFP-soloSET (red) fusion proteins in tobacco leaf cells. Overlapping signals appear in yellow. Arrows point to nuclei. (c) Two-dimensonal-distribution of proteins from WT and from RNAi-SET (RNAi) expressing cellular extracts (Coomassie blue staining). Arrows point to the EF1A spot, isoelectric point 9.20. (d) Immunostaining of the 2D-gel blots with the antibody against H3K4me3. (e) Western blots of protein extracts from SET-RNAi, WT and atx1 cells with the antiH3K4me3 antibody. Arrow indicates position of the 50 kDa protein band. Coomassie-stained proteins are shown as loading controls in the lower panel.

Next, we asked whether the two proteins interact in plant cells. We co-expressed RFP-tagged soloSET and GFP-tagged EF1A in tobacco cells and detected that they co-localize in the cytoplasm (Figure 5b). As also shown in earlier reports (28,29), the Arabidopsis EF1A localized in the nuclei as well. However, the soloSET co-localized with EF1A only on the nuclear periphery but not in the nuclei (Figure 5b, inset) supporting the conclusion that that the two proteins interact only in the cytoplasm.

SoloSET is required and sufficient for methylation of EF1A

Known bacterial EF-Tu (30) and eukaryotic eEF1A (yeast, fungi, animals and plants) are posttranslationally methylated (13,31–33). However, the methyltransferases involved have not been identified. It is also unknown which, if any, of the Arabidopsis EF1A lysine residues are methylated. The binding of soloSET to EF1A hinted at the intriguing possibility that soloSET might play a role in methylating EF1A.

Using C14-labeled SAM as donor and recombinantly produced soloSET and EF1A proteins, we were unable to detect a methylated EF1A product. Likewise, recombinantly expressed ATX1 is unable to effectively methylate its histone target in vitro (5). Consequently, we examined whether soloSET was involved in EF1A methylation in vivo. Our approach was based on the serendipitous finding that antibodies against tri-methylated lysine 4 of histone H3 (H3K4me3), but not against H3K4me2 or H3K9me2, detected methylated EF1A. We note that the H3K4me3 antibodies are used here only as a tool to detect EF1A methylation and not to imply specificity for any methylated lysine; it is noted also, that the H3K4me3 antibodies reacted with a few other unidentified proteins.

Cellular extracts from WT and from RNAi-SET transgenic plants were resolved by 2D gel electrophoresis and the spot corresponding to EF1A (Figure 5c, arrow in left-hand panel) was excised and positively identified as EF1A by MS. Importantly, EF1A was produced in RNAi-SET cells (Figure 5c, arrow in right-hand panel). However, there was a notable difference in the EF1A methylation states in RNAi-SET versus wild-type cells. The signal from EF1A in transgenic cells was absent (Figure 5d, arrow) indicating loss of EF1A methylation. The signals from the other non-specific proteins remained largely unchanged providing evidence that depletion of SET domain activity affected specifically the methylation of EF1A. We note also that ATX1 disruption in atx1 cells did not eliminate methylation of EF1A (Figure 5e) although it abolished ATX1 function at chromatin (8,20). As soloSET is produced in the atx1 background, the result indicates that the SET domain, alone, is necessary and sufficient for EF1A methylation.

Mass spectroscopy was used to identify the methylated residues. Analyses of the spots corresponding to EF1A from WT cells detected mono-methylated K261, di-methylated K55 and tri-methylated K79, K187 K306, and K396 (Supplementary Figure S6a). The same modified lysines were identified again in the RNAi-SET expressing sample, with the notable exception of K396 (Supplementary Figure S6b). This result defines Lys396 as the specific target. The loss of the three-methyl groups suggested that the ATX1-derived SET was involved in modifying K396 at all three positions of the amino group, unlike ATX1, which mainly affects tri-methylation of H3K4 in chromatin (8). These results, reproduced in three independent experiments, confirm that EF1A–K396 is specifically modified by the SET domain.

The validity of our MS results for the Arabidopsis EF1A is further supported by the pattern of lysine methylations reported for the maize EF1A: tri-methylated K79, K187, K306 and monomethylated K396 (13). The results indicate that lysine methylations at these positions are conserved in the monocot and dicot lineages. Even more significant is the finding of a tri-methylated K390 in the yeast EF1A, corresponding to the K396 of plants’ EF1A (14), as it suggested an evolutionarily conserved role for the methylation of this residue in EF1A. K396 is located in the domain involved in the EF1A actin bundling, the best studied non-canonical function of elongation factors (14–18). In yeast, substitution of the post-translationally modified lysines with arginines did not affect translation but altered actin cytoskeleton organization (14). Next, we examined whether loss of K396 methylation resulting from soloSET suppression had any effect on actin filament patterns in Arabidopsis cells.

Actin cytoskeleton in RNAi-SET and in atx1 mutant cells

Consistent with available data, actin in wild-type leaf cells was observed at the plasma membrane (cortical actin) and as fine and thick bundled actin networks (Figure 6a–c). Longitudinal, transverse, and randomly oriented actin bundles, considered general features of interphase plant cells (19), are clearly displayed. The transvacuolar cytoplasmic strands (TVSs) providing direct connections between different regions in the cytoplasm are also prominent in wild-type cells. In contrast, RNAi-SET expressing cells displayed a dramatically different pattern of reduced actin bundles and absent TVSs (Figure 6e–j). The fluorescent signal is associated mostly with particulate structures at the vacuolar periphery suggesting a collapse of the bundled structures. These changes were displayed by all RNAi-SET expressing cells implicating SET-domain activity in the actin bundling in Arabidopsis cells. In sharp contrast, actin bundles and TVSs similar to the wild type were prominently displayed in atx1 mutant cells (Figure 6k–m).

Figure 6.

The actin cytoskeleton and RNAi-SET. Actin filaments stained with phalloidin in wild-type leaf cells exhibiting actin bundles of various thicknesses (a–c). A contrast image (d) of the cells shown in (c) illustrating the TVS (arrows). Bars are 20 µM. Actin filaments in RNAi-SET expressing cells in (e–j). Granular particles on the vacuolar periphery are indicated by arrows in e. (g) Bars are 20 µM. Closer look at vacuolar-associated actin (h–j). Arrows indicate the cell wall. Bar is 10 µM. Actin bundles in atx1 mutant cells (k–m). Bars are 20 µM.

These results are important because they indicate that loss of ATX1 function does not alter the actin cytoskeleton. Thereby, the cytoskeletal phenotype is dependent on presence of functional SET, but not ATX1, illustrating a critical distinction of the roles played by the two isoforms. As soloSET is produced in the atx1 background (Figure 1b), it is logical to conclude that actin bundling is dependent on the activity carried by soloSET, most likely, through its role in methylating EF1A K396. This residue is located in the actin-binding domain (III) of EF1A and its methylation state might be critical for the interaction. Observations that amino acid mutations in the yeast EF1A domain III resulted in disappearance of actin cables and appearance of granular actin patches (17), while substitution of the methylated lysines with arginines altered actin cytoskeleton organization in yeast (14) support this possibility.

DISCUSSION

Collectively, our results provide answers to two currently open and, apparently, unrelated questions: whether chromatin modifying (epigenetic) genes encode factors that modify proteins other than the histones and what methyltransferases are involved in modifying the EF1A lysines.

We provide an example of an isoform of an epigenetic factor with a role unrelated to chromatin. It is also the first example of a Trithorax family member produced from an internal gene promoter.

Three independent types of evidence distinguish SET function as part of ATX1 and as a solo domain: first, the two products of the ATX1 gene are localized in different cellular compartments; second, a disrupted ATX1 (in atx1 cells, where soloSET is produced) did not eliminate methylation of EF1A (Figure 5e) but H3K4 methylation at ATX1 chromatin targets was erased (3,8,20). As presence of SET domain alone it is not sufficient for the methylation of H3K4 but is sufficient for the methylation of EF1A, it indicates that the SET domain peptide plays different roles when alone and as part of the ATX1 protein; third, loss of SET, but not of the histone modifying ATX1-SET activity (in the atx1 background) affects cytoskeletal actin bundling and the cytoplasm-specific phenotype. Our results do not exclude the possibility that SET within ATX1 could also modify K396, a function that would be redundant with soloSET. However, given that deletion of SET from ATX1 (in atx1 mutants) does not affect the EF1A associated functions, indicates that the role of ATX1 is not essential for these cytoplasmic functions. We consider these results among the most significant findings of this study as it has been known for a number of years that some EF1A lysines are methylated, that methylation of specific lysines is evolutionarily conserved, and that these modifications are involved in cytoskeletal functions (actin bundling) rather than in protein synthesis. However, the methyltransferases modifying the EF1A lysines and whether distinct enzymes modify specific lysines have been hitherto unknown. Here, we have identified the first methyltransferases involved in the methylation of EF1A and have shown that it modifies specifically one, among several, methylated lysines. Thereby, it is logical to expect that different methyltransferases modify specific lysines, which regulate (or reflect) the broad spectrum of roles of the multitasking EF1A. It is tempting to speculate that lysine-specific methylation in EF1A play roles similar to the roles played by the specific histone marks on chromatin. Furthermore, the methylated lysines of the Arabidopsis EF1A identified in our study and, particularly, the conserved K396 methylation, were also found in a monocot and in yeast providing a strong argument for an evolutionarily conserved importance for this modification.

Lastly, we note that soloSET provides an example of a eukaryotic SET domain protein structurally similar to the SET domain proteins found in bacteria. Despite a belief that SET domain genes have occurred with eukaryotes (34,35), phylogenetic analysis has suggested that the genes are ancient and do not result from a lateral gene transfer (36). Protein methylating activity was reported for a M. mazei SET domain protein (37) but the role of the bacterial SET domain proteins remains unknown. Demonstrating that SET domain proteins have roles outside the nucleus might trigger exploration of their function in prokaryotes and stimulate discovery of novel cytoplasmic functions produced by genes known currently to encode only chromatin modifiers.

SUPPLEMENTARY DATA

Supplementary Data are available at NAR Online.

FUNDING

National Institutes of Health award P20 RR15635 (to MS facility), National Institutes of Health award RR015468-01; NSF-EPS0701892, partial (to Z.A.); NSF-MCB0749504, partial (to Z.A.). Funding for open access charge: NSF-MCB0749504.

Conflict of interest statement. None declared.