Trithorax-group proteins ARABIDOPSIS TRITHORAX4 (ATX4) and ATX5 function in abscisic acid and dehydration stress responses

Transcription

1 Research Trithorax-group proteins ARABIDOPSIS TRITHORAX4 (ATX4) and ATX5 function in abscisic acid and dehydration stress responses Yutong Liu, Ai Zhang, Hao Yin, Qingxiang Meng, Xiaoming Yu, Shuangzhan Huang, Jie Wang, Rafiq Ahmad, Bao Liu and Zheng-Yi Xu Key Laboratory of Molecular Epigenetics of the Ministry of Education (MOE), Northeast Normal University, Changchun , China Authors for correspondence: Zheng-Yi Xu Tel: Bao Liu Tel: Received: 26 June 2017 Accepted: 2 November 2017 doi: /nph Key words: abiotic stress, abscisic acid, Arabidopsis thaliana, chromatin state, dehydration stress, histone methylation, transcriptional regulation. Summary Trithorax-group proteins (TrxGs) play essential regulatory roles in chromatin modification to activate transcription. Although TrxGs have been shown to be extensively involved in the activation of developmental genes, how the specific TrxGs function in the dehydration and abscisic acid (ABA)-mediated modulation of downstream gene expression remains unknown. Here, we report that two evolutionarily conserved Arabidopsis thaliana TrxGs, ARABIDOPSIS TRITHORAX4 (ATX4) and ATX5, play essential roles in the drought stress response. atx4 and atx5 single loss-of-function mutants showed drought stress-tolerant and ABA-hypersensitive phenotypes during seed germination and seedling development, while the atx4 atx5 double mutant displayed further exacerbation of the phenotypes. Genome-wide RNA-sequencing analyses showed that ATX4 and ATX5 regulate the expression of genes functioning in dehydration stress. Intriguingly, ABA-HYPERSENSITIVE GERMINATION 3 (AHG3), an essential negative regulator of ABA signaling, acts genetically downstream of ATX4 and ATX5 in response to ABA. ATX4 and ATX5 directly bind to the AHG3 locus and trimethylate histone H3 of Lys 4 (H3K4). Moreover, ATX4 and ATX5 occupancies at AHG3 are dramatically increased under ABA treatment, and are also essential for RNA polymerase II (RNAPII) occupancies. Our findings reveal novel molecular functions of A. thaliana TrxGs in dehydration stress and ABA responses. Introduction Epigenetic mechanisms play essential roles in the regulatory networks involved in plant developmental responses to environmental conditions. The evolutionarily conserved epigenetic components polycomb group (PcG) and trithorax group (TrxG) are involved in developmental processes that respond to environmental signals, and play important roles in plant plasticity (Paz Sanchez et al., 2015). These two components have been well characterized in metazoans, in which the PcG complex catalyzes trimethylation of histone H3 of Lys 27 (H3K27), while the TrxG complex catalyzes the trimethylation of H3K4 and histone H3 of Lys 36 (H3K36) (Schuettengruber et al., 2011). The TrxG activators are a large and functionally diverse group of regulatory proteins (Kingston & Tamkun, 2014), including the Su(var)3-9, Enhancer-of-zeste and Trithorax (SET) domain-containing proteins with H3K4 methyltransferase activity, ATP-dependent chromatinremodeling factors, and other associated proteins (Krajewski et al., 2005; Avramova, 2009). In Arabidopsis thaliana, five trithorax (ARABIDOPSIS TRITHORAX1 5 (ATX1 5 )) and seven trithoraxrelated (ATX-RELATED1 7 (ATXR1 7 )) genes have been identified (Baumbusch et al., 2001; Alvarez-Venegas & Avramova, 2002; Veerappan et al., 2008). Among the five trithorax genes, ATX1 and ATX2 form the first group, and ATX3, ATX4 and ATX5 form the second group, based on evolutionary and functional classifications (Baumbusch et al., 2001; Alvarez-Venegas & Avramova, 2002). The structural feature that separates the two groups is the presence of the Domain Associated with SET in Trithorax (DAST; the phenylalanine-tyrosine rich at the N- terminus or C-terminus (FYRN/FYRC) juxtaposed version) in the ATX1/2 sister group; ATX3, ATX4 and ATX5 lack DAST but harbor the Plant homeodomain (PHD) finger. ATX1 and ATX2 share 75% similarity at the amino acid level, with similar architectural motifs (Saleh et al., 2008; Avramova, 2009). However, atx1 and atx2 mutants have totally distinct phenotypes; while atx1 displays early bolting and numerous flower-organ aberrations, atx2 does not show apparent phenotypic differences from wild-type (WT) plants (Alvarez-Venegas et al., 2003; Saleh et al., 2008). Further biochemical assays revealed that ATX1 is a histone H3 lysine 4 trimethylation (H3K4me3) methyltransferase while ATX2 is an H3K4me2 methyltransferase. It has been reported that ATX1 harboring H3K4 trimethylation activity functions in 1582

2 New Phytologist Research 1583 floral organ development, organogenesis, and abiotic and biotic stress responses (Alvarez-Venegas et al., 2003; Ding et al., 2011a). ATX2 mutant lines do not display obvious phenotypes, and ATX2 has H3K4 dimethylation activity (Saleh et al., 2008). Of note, ATX1 and ATX2 influence the expression of largely nonoverlapping gene sets (Saleh et al., 2008). A recent study showed that the atx3 atx4 atx5 triple mutant displayed defects in vegetative and reproductive stages but single mutants for each of the three genes or double mutants of their pairwise combinations did not display noticeable growth and developmental phenotypes (Chen et al., 2017). Genome-wide chromatin immunoprecipitation (ChIP)-sequencing analysis revealed that H3K4me2 and H3K4me3 levels were reduced in the atx3 atx4 atx5 triple mutant, indicating that these three genes have redundant functions during vegetative and reproductive stages by modulating H3K4me2 and H3K4me3 levels (Chen et al., 2017). Abscisic acid (ABA) plays essential roles in various biological processes including maintenance of seed dormancy, inhibition of seed germination, acceleration of senescence and stress tolerance (Borsani et al., 2002; Finkelstein et al., 2002; Xiong & Zhu, 2002; Nambara & Marion-Poll, 2005; Xu et al., 2012, 2013; Park et al., 2015; Zhu, 2016). It has been shown that several protein phosphatases are deeply involved in ABA signaling (Rodriguez et al., 1998; Schweighofer et al., 2004). ABA INSENSITIVE1 (ABI1) and ABA INSENSITIVE2 (ABI2 ) encode homologous PROTEIN PHOSPHATASES 2C which have been identified in studies of two dominant ABA-insensitive mutants, abi1-1 and abi2-1 (Leung et al., 1994, 1997; Allen et al., 1999). AtPP2CA, another A. thaliana PP2C, was shown to block ABA signal transduction when transiently expressed in Zea mays (Sheen, 1998). Knockdown of AtPP2CA by an antisense gene led to accelerated plant development and increased freezing tolerance (T ahtiharju & Palva, 2001). These results suggest the essential role of AtPP2CA as a negative regulator of the ABA response (Park et al., 2009). An ABA-hypersensitive mutant, ABA-hypersensitive germination 3 (ahg3), was isolated by screening with an ABA analog (Nishimura et al., 2004; Yoshida et al., 2006). One ABA-HYPERSENSITIVE GERMINATION 3 (AHG3) loss-of-function mutant, ahg3-1, has been shown to display strong ABA-hypersensitive phenotypes (Nishimura et al., 2004; Yoshida et al., 2006). Further studies revealed that AHG3 functions as a negative regulator in the ABA signaling pathway during seed germination and early seedling growth (Schweighofer et al., 2004; Hubbard et al., 2010). H3K4me3 conferred by the trithorax proteins and their homologs in different eukaryotic systems has been linked to the transcriptional activation of regulated genes, which suggests that the molecular basis of the trithorax effects lies in its ability to covalently modify histones (Allen et al., 1999; Bernstein et al., 2006; Schuettengruber et al., 2007; Berr et al., 2010; Howe et al., 2017). Genome-wide ChIP coupled with next-generation sequencing (NGS) analysis in A. thaliana revealed that H3K4me3 shows a broad distribution in gene body regions of dehydration stress and ABA-responsive genes (Barski et al., 2007; Van Dijk et al., 2010). In addition to the essential roles of H3K4me3 in transcriptional initiation and elongation, H3K4me3 is an important signature of gene transcription memory under drought stress, which is facilitated by ATX1 (Oh et al., 2008; Zhang et al., 2009; Ding et al., 2011b, 2012; Kim et al., 2012, 2015). Despite these advances, many important issues remain to be addressed. For example, in addition to ATX1, do other ATXs and ATXRs function during the dehydration stress and ABA responses and, if so, how? What are the specific physiological functions of those ATXs and ATXRs? What are the specific underlying molecular mechanisms of these ATXs and ATXRs in the regulation of gene expression? In this study, we report that two evolutionarily conserved A. thaliana TrxGs, ATX4 and ATX5, play pivotal roles in the drought stress response. Both atx4 and atx5 single loss-offunction mutants showed clear ABA-hypersensitive and drought stress-tolerant phenotypes, which were further exacerbated in the atx4 atx5 double mutant. Biochemical analysis revealed that ATX4 and ATX5 specifically di- and trimethylate histone H3K4. Genome-wide RNA-sequencing analysis indicated that ATX4 and ATX5 play essential roles in the transcriptional regulation of a specific subset of functionally relevant genes. We further identified that AHG3, an essential negative regulator in ABA signaling, acts genetically downstream of ATX4 and ATX5 in response to ABA. Intriguingly, ATX4 and ATX5 directly bind to the AHG3 locus and control H3K4me3 levels in the presence of ABA. Our findings reveal novel molecular functions of A. thaliana TrxGs in drought stress and ABA responses. Materials and Methods Plant growth conditions and genotyping Arabidopsis thaliana (L.) Heynh. Columbia-0 (Col-0) plants were grown either in a glasshouse or on Murashige and Skoog (MS) plates at 23 C in an incubator maintaining 60% relative humidity conditions with 80 lmol m 2 s 1 or 160 lmol m 2 s 1 under a 16 h : 8 h, light : dark photoperiod (Moore et al., 2003). The atx4 mutants, containing homozygous atx4-1 (SAIL_1257_ E09) and atx4-2 (SALK_060156), and atx5 mutants, containing homozygous atx5-1 (SAIL_705_H05) and atx5-2 (WiscDsLoxHs127_10D), were isolated using ATX4-1LP/RP and ATX4-2LP/RP, and ATX5-1LP/RP and ATX5-2LP/RP, respectively. To generate atx4 atx5 double mutants, atx4-1 was crossed with atx5-1 mutants, and double mutants were screened using ATX4-1LP/RP and ATX5-1LP/RP PCR primers. To generate atx4-1 ahg3-1 and atx5-1 ahg3-1 double mutants and the atx4-1 atx5-1 ahg3-1 triple mutant, atx4-1, atx5-1 and atx4-1 atx5-1 mutants, respectively, were crossed with the ahg3-1 mutant (Nishimura et al., 2004; Jiang et al., 2015), and mutants were screened using primers ATX4-1LP/RP, ATX5-1LP/RP and AHG3-LP/RP, respectively. All PCR primers used for genotyping are listed in Supporting Information Table S1. Construction of plasmids and generation of transgenic plants Gene-specific primers ATX4-F/R and ATX5-F/R were used to isolate ATX4 and ATX5 cdna, respectively, from a cdna

3 1584 Research New Phytologist library by PCR. To generate the pcsv1300-atx4-2xflag and pcsv1300-atx5-2xflag constructs, full-length ATX4 and ATX5, respectively, were amplified and cloned into the pcsv1300 vector with a 2xFLAG tag using the BamHI site (Xu et al., 2012). To generate ATX4-GFP and ATX5-GFP constructs to transfect A. thaliana protoplasts, ATX4 and ATX5 DNA fragments were inserted into 326-GFP using XbaI and BamHI sites and the BamHI site (Jin et al., 2001), respectively. To generate patx4::gus and patx5::gus constructs, kb and kb fragments upstream of ATX4 and ATX5, respectively, were amplified by PCR using primers ATX4p-F/R and ATX5p-F/R and inserted into the binary vector ppzp211 using SacI and PstI sites (Tian et al., 2015). To generate patx4::atx4-2xflag and patx5::atx5-2xflag constructs, the genomic sequences of ATX4 and ATX5 containing promoter regions were amplified using ATX4G-F/R (2xFLAG/KnpI/Sal I) and ATX5G-F/R (2xFLAG/KnpI/SalI) primers, respectively. Subsequently, the constructs were inserted into the binary vector pcambia1302 (Invitrogen). All primers are listed in Table S1. Plants transformed with ppzp211-patx4::gus, ppzp211-patx5::gus, pcambia1302-patx4::atx4-2xflag and pcambia1302- patx5::atx5-2xflag binary constructs were grown on B5 plates treated with kanamycin (50 mg l 1 ) or hygromycin (50 mg l 1 ) (Clough & Bent, 1998). Transient expression in protoplasts and b-glucuronidase (GUS) staining Rosette leaves of 4-wk-old A. thaliana plants grown under shortday conditions were used for the isolation of protoplasts (Jin et al., 2001; Hyunjong et al., 2006). The relevant vectors ATX4 green fluorescent protein (GFP), ATX5-GFP and nuclear localization signal red fluorescent protein (NLS-RFP) were used for protoplast transformation. A fluorescence microscope was used to observe GFP and RFP signals (Kim et al., 2001; Bae et al., 2008). GUS assays were performed on different tissue extracts of transgenic plants expressing patx4::gus and patx5::gus as previously described by Xu et al. (2012). To test for GUS expression before and after ABA and dehydration stress, plants were treated with 100 lm ABA for 4 h and dehydration for 2 h, respectively (Taylor et al., 1995; Ren et al., 2010). Stomatal aperture size, water loss, and survival rate measurements The stomatal aperture sizes were measured as previously described (Jeon et al., 2008) and conducted using the software IMAGEJ (National Institutes of Health, Bethesda, MD, USA). The experiment was repeated three times, for each 100 pairs of guard cells from three different plants were measured. For measurement of leaf water loss, the above-ground parts of the plants (the above-ground parts of the 14-d-old plants were excised, placed on weighing paper under 40% relative humidity and weighed at three different time-points) were excised, placed on weighing paper and weighed at different time-points, with three replicates per time-point (Xu et al., 2012). To test for drought tolerance, plants were grown on soil in the same glasshouse with 160 lmol m 2 s 1 under a 16 h : 8 h, light : dark photoperiod (23 C; 60% relative humidity) for 2 wk; water was withheld from 14-d-old plants for 14 d, and the survival rates of plants were determined 2 d after re-watering (rehydration) (Kang et al., 2002; Xu et al., 2012). Three independent experiments were performed. Histone H3K4 methylation assay Nuclei were isolated from WT, atx4-1 and atx5-1, and their histone modification status was determined (Serino & Deng, 2007). The same amount of proteins from WT, atx4-1 and atx5-1 or WT, ATX4OX-1, -2, ATX5OX-1 and -2 were used for western blotting assays. The protein expression levels were analyzed by western blotting using different antibodies including H3K4me3 (Abcam ab8580; Cambridge, MA, USA), H3K4me1 (Millipore ; Darmstadt, Germany), H3K4me2 (Millipore ), H3K27me3 (Millipore ), H3K36me3 (Abcam ab9050), H3K9me3 (Millipore ) and H3 (Active Motif 39763, Carlsbad, CA, USA). Bioinformatics analyses of RNA-sequencing (RNA-seq) data Arabidopsis thaliana plants were grown either in the glasshouse at 23 C or in the incubator maintaining 60% relative humidity with 160 lmol m 2 s 1 light intensity under a 16 h : 8 h, light : dark photoperiod (Moore et al., 2003). Total RNA was extracted with Trizol (Invitrogen) and 3 lg of RNA for each sample was used for library construction and subsequent RNA deep sequencing on the Illumina Hi-seq platform (Illumina, San Diego, CA, USA) with three biological replications. The Agilent 2100 Bioanalyzer (Agilent Technologies, Waldbronn, Germany) was used to determine the quality and concentration of RNA. Sequencing was performed in paired-end mode with a read length of 150 nucleotides. Next, low-quality (< Q20) reads were excluded from raw data using FASTX-TOOLKIT v ( The clean reads were mapped to the Arabidopsis reference genome (TAIR10) using TOPHAT v (Trapnell et al., 2009) with TAIR10 gene annotation as the transcript index. Gene quantification was performed using CUFFLINKS ( shtml; with genomic annotation from the TAIR10 genome release. The differentially expressed genes were filtered according to the fold change ( log 2 FC > 1) and an adjusted P-value (< 0.05), calculated with CUFFDIFF, a subpackage of CUFFLINKS ( re/tophat/index.shtml; links/). The gene ontology (GO) grouping of differentially expressed genes (DEGs) was performed by hypergeometric distribution in R v ( with an adjusted P-value < 0.05 as a cutoff to determine significantly enriched GO terms.

4 New Phytologist Research 1585 Chromatin immunoprecipitation assays The chromatin immunoprecipitation (ChIP) assay was performed according to the previously described method (Haring et al., 2007) with a slight modification. Two-week-old plants that were mock-treated or treated with ABA were selected for ChIPqPCR. Specific antibodies, including anti-h3k4me3 (Abcam ab8580), anti-flag (a fusion tag, called FLAG and consisting of eight amino acids (AspTyrLysAspAspAspAspLys); Einhauer & Jungbauer, 2001) (F1804; Sigma) and anti-rna polymerase II (RNAPII) (at-300; Santa Cruz Biotechnology, Santa Cruz, CA, USA), were added to the chromatin, which was isolated and sheared to bp with an FB120 Sonic Dismembrator (Fisher Scientific, Waltham, MA, USA) for an overnight incubation at 4 C. The antibody protein complexes were isolated by binding to protein A beads. The DNA fragments in the immunoprecipitated complexes were released by reversing the crosslinking at 65 C for 8 h, then extracted with phenol/chloroform and precipitated in ethanol and re-suspended in water. The specific primers used for real-time PCR are listed in Table S1. ACTIN7 (ACT7 ) was used as a negative control. Accession number Data generated in this study are deposited in the National Center for Biotechnology Information Sequence Read Archive (accession no. SRP123762). Results Isolating ABA-hypersensitive ATX4 and ATX5 loss-offunction mutants During the measurement of seed germination rates (cotyledon greening) in homozygous mutant lines for the ATX and ATXR genes treated with 0.5 lm ABA, we identified two T-DNA Fig. 1 ARABIDOPSIS TRITHORAX4 (ATX4) and ATX5 play negative roles in abscisic acid (ABA) responses in Arabidopsis thaliana. (a d) Measurement of seed germination rates (cotyledon greening) of wild-type (WT), atx4-1, atx4-2, atx5-1 and atx5-2 as well as complementation lines (Com4#1, Com4#2, Com5#1 and Com5#2) after ABA treatment. Plants were germinated in halfstrength Murashige and Skoog (MS) medium supplemented with dimethyl sulfoxide (DMSO) or different concentrations of ABA. Images were taken after incubation for 17 d. Error bars indicate SD (n = 3). Statistical analyses were performed comparing WT and atx4-1, WT and atx4-2, WT and atx5-1, and WT and atx5-2. **, P-value < 0.01 (Student s t-test). (e h) Measurement of root growth of WT, atx4-1, atx4-2, atx5-1 and atx5-2 as well as complementation lines (Com4#1, Com4#2, Com5#1 and Com5#2) with or without ABA treatment. Plants grown on half-strength MS plates for 3 d were transferred to medium containing DMSO or 10 lm ABA for 14 d, and images were then taken. Bars, 2 cm. Three independent experiments were performed using 20 plants per experiment. Error bars indicate SD (n = 3). Statistical analyses were performed comparing WT and atx4-1, WT and atx4-2, WT and atx5-1 and WT and atx5-2. **, P-value < 0.01 (Student s t-test).

5 1586 Research New Phytologist insertion mutants, atx4-1 (SALK_060156) and atx5-1 (SAIL_705_H05), which displayed ABA-hypersensitive phenotypes compared with WT (Fig. 1a d). ATX4 and ATX5 belong to the sister group of the ATX1/2 TrxG genes, which share 85% amino acid similarity and have very similar protein domain structures, such as PWWP (proline tryptophan tryptophan proline), FY-rich, PHD and SET domains (Fig. S1a). Using PCR analysis, we found that T-DNAs were inserted in the first exons of both ATX4 and ATX5, generating atx4-1 and atx5-1, respectively (Fig. S1b). We further found that the ATX4 and ATX5 transcripts were totally abolished in atx4-1 and atx5-1 based on semiquantitative reverse transcription PCR (RT-PCR) and quantitative real-time PCR (qrt-pcr) assays (Fig. S1c,d). To further confirm the ABA-related phenotypes of atx4-1 and atx5-1, we isolated another two independent T-DNA insertion mutants, atx4-2 (SAIL_1247_E09) and atx5-2 (WiscDsLoxHs127_10D). T-DNAs were inserted into the first exon of ATX4 (atx4-2) and the 16th exon of ATX5 (atx5-2) (Fig. S1b). As shown in Fig. 1(a d), atx4-2 and atx5-2 also displayed ABA-hypersensitive phenotypes during seed germination. Of note, another two ATX1/2 sister group protein ATX3 loss-of-function mutants, atx3-1 and atx3-2, did not display noticeable ABA phenotypes during seed germination (Fig. S1b f). To further investigate whether the ABA-hypersensitive phenotypes of atx4 and atx5 are causally linked to loss of ATX4 or ATX5, we generated complementation lines by transfecting constructs harboring ATX4 and ATX5 genomic DNA, separately, with two repeated FLAG epitopes driven by their native promoters (patx4::atx4-2xflag and patx5::atx5-2xflag), which were named Com4 and Com5, respectively. As shown in Fig. 1(a d), the two independent complementation lines, Com4 (Com4#1 and Com4#2) and Com5 (Com5#1 and Com5#2), displayed similar ABA sensitivities to WT. Next, we evaluated whether ATX4 and ATX5 played a role during seedling development in the presence of ABA. The 3-dold seedlings grown on 1/2 MS medium were transferred to 10 lm ABA-containing medium for 14 d. As shown in Fig. 1(e h), root growth of the atx4-1, atx4-2, atx5-1 and atx5-2 mutants was dramatically retarded under ABA conditions compared with that of WT and the complementation lines. Taken together, these results indicate that ATX4 and ATX5 play negative roles during seed germination and seedling development in response to exogenous ABA. ATX4 and ATX5 function in plant growth and development During the vegetative stage, we found that atx4-1, atx4-2, atx5-1 and atx5-2 grew slightly smaller than WT under conditions of 160 lmol m 2 s 1 light intensity with a 16 h : 8 h, light : dark regime (Fig. S2a,b). Intriguingly, these differences became exaggerated under conditions of 80 lmol m 2 s 1 light intensity with a 16 h : 8 h, light : dark regime (Fig. S2a,b). At later reproductive stages, apart from the reduced stem growth, atx4-1 and atx4-2 did not display additional phenotypic differences compared with the WT (Fig. S2c,e). However, atx5-1 and atx5-2 showed developmental abnormalities such as reduced stem growth, curling cauline leaves, and abnormal inflorescence architecture including a short pedicle and compact inflorescence (Fig. S2d,f). Together, these data indicate that ATX4 and ATX5 probably play redundant roles in leaf development during vegetative stages, although ATX5 may play a specific role in cauline leaf and inflorescence development during reproductive stages. ATX4 and ATX5 negatively regulate drought stress responses ABA plays a critical role in inducing stomatal closure under drought stress. Thus, we measured the width and length and further calculated the length to width ratio of stomata from epidermal peels of WT, atx4-1, atx4-2, atx5-1 and atx5-2. As shown in Fig. 2(a,b), atx4 and atx5 mutants displayed accelerated ABA-mediated stomatal closure under treatment with ABA. However, the ATX4 and ATX5 complementation lines showed a similar stomatal aperture to WT, indicating that ATX4 and ATX5 negatively control ABA-mediated stomatal closure. Next, we tested whether ATX4 and ATX5 play a role in the drought stress response. For this, aerial parts of WT and independent atx4 and atx5 mutant lines as well as the complementation lines were excised and water-loss rates were measured over time. As shown in Fig. 2(c,d), the atx4 and atx5 mutant lines displayed reduced water-loss rates compared with WT, while the complementation lines exhibited similar water-loss rates to WT, suggesting that the mutation in ATX4 or ATX5 is causal to the reduced transpiration. To further examine the longer term effects of ATX4 and ATX5 on drought stress, WT, atx4, atx5, and complementation lines grown for 14 d under normal growth conditions were exposed to dehydration stress by withholding water for 14 d. Survival rates were examined 2 d after re-watering. As shown in Fig. 2(e,f), the survival rates of atx4 and atx5 mutants were significantly higher than those of WT and the complementation lines. Together, these results indicate that ATX4 and ATX5 negatively regulate drought stress response in WT A. thaliana. Tissue-specific expression patterns of ATX4 and ATX5, and subcellular localizations of ATX4 and ATX5 To examine the spatial and temporal expression patterns of ATX4 and ATX5, we assayed the promoter activities of ATX4 and ATX5 in different tissues at different developmental stages. Transgenic plants were generated using patx4::gus or patx5:: GUS constructs containing intergenic DNA regions from the upstream region of the ATX4 or ATX5 start codon to drive the GUS coding region. As shown in Fig. S3(a), GUS was detected in leaves, primary and secondary roots, veins of different plant tissues, hypocotyls and floral tissues. Subsequently, we confirmed these results using qrt-pcr analysis. As shown in Fig. S3(b), ATX4 and ATX5 transcript levels were in accordance with the GUS expression patterns indicated by GUS signals in different tissues. Taken together, these results indicate that ATX4 and ATX5 displayed similar spatial and temporal expression patterns during vegetative and reproductive developmental stages. Next, we examined the responsiveness of ATX4 and ATX5 expression

6 New Phytologist Research 1587 Fig. 2 ARABIDOPSIS TRITHORAX4 (ATX4) and ATX5 negatively regulate drought stress responses in Arabidopsis thaliana. (a, b) The role of ATX4 and ATX5 in abscisic acid (ABA)-induced stomatal closure. Stomata of wild-type (WT), atx4-1, atx4-2, atx5-1 and atx5-2 as well as complementation lines (Com4#1, Com4#2, Com5#1 and Com5#2) were photographed with or without treatment with 5 lm ABA. (b) To quantify stomatal closure, the length to width ratios of stomata were measured in a triplicate experiment with 100 pairs of guard cells per experiment. Error bars indicate SD (n = 3). Statistical analyses were performed comparing WT and atx4-1, WT and atx4-2, WT and atx5-1, and WT and atx5-2. **, P-value < 0.01 (Student s t-test). (c, d) Measurement of water loss. The aerial parts of WT, atx4-1, atx4-2, atx5-1 and atx5-2 as well as complementation lines (Com4#1, Com4#2, Com5#1 and Com5#2) were excised and placed in conditions of 40% relative humidity. (c) Images of normal (23 C, 160 mmol m 2 s 1 light intensity) (upper panel) and detached (lower panel) leaves exposed to dehydration stress for 3 h. Bars, 2 cm. (d) To quantify water loss rates, fresh weights of detached leaves were measured at the indicated time-points. Data are mean values of three independent experiments (n = 12). (e, f) Measurement of survival rates. WT, atx4-1, atx4-2, atx5-1 and atx5-2 as well as complementation lines (Com4#1, Com4#2, Com5#1 and Com5#2) were subjected to treatment consisting of withholding water for 14 d. The images of surviving plants were taken at 2 d after rewatering. Bars, 1.5 cm. Five independent experiments were performed. Error bars indicate SD (n = 5). Statistical analyses were performed comparing WT and atx4-1, WT and atx4-2, WT and atx5-1, and WT and atx5-2. **, P-value < 0.01 (Student s t-test). in the presence of ABA and drought stress. For this purpose, we took advantage of transgenic plants expressing patx4::gus and patx5::gus. As show in Fig. 3(a), after dehydration stress and ABA treatments, GUS signals were increased in cotyledons, leaves and roots as well as guard cells. To determine the subcellular localization of ATX4 and ATX5, we generated constructs that introduced the GFP sequence at the C-terminus of ATX4 and ATX5. The ATX4-GFP and ATX5- GFP constructs were used to transfect Arabidopsis thaliana protoplasts together with NLS-RFP. We observed that GFP and RFP signals were co-localized at the nucleus, indicating that ATX4 and ATX5 are localized to the nucleus (Fig. 3b). ATX4 and ATX5 are specific H3K4 methyltransferases It was shown recently that concurrent disruption of the ATX3, ATX4 and ATX5 genes in atx3 atx4 atx5 triple mutants caused marked reduction in H3K4me2 and H3K4me3 levels genomewide, suggesting that these three genes probably encode H3K4 methyltransferases (Chen et al., 2017). However, it remains

7 1588 Research New Phytologist Fig. 3 Tissue-specific expression patterns of Arabidopsis thaliana ARABIDOPSIS TRITHORAX4 (ATX4) and ATX5 and subcellular localizations of ATX4 and ATX5. (a) b-glucuronidase (GUS) staining for spatial and temporal expression patterns of ATX4 and ATX5. Transgenic plants expressing patx4::gus and patx5::gus at different developmental stages were stained with 5- bromo-4-chloro-3-indolyl b-d-glucuronide (X-Gluc). GUS expression was examined in cotyledons, leaves, roots and guard cells before and after 100 lm abscisic acid (ABA) treatment for 4 h and dehydration for 2 h. (b) Subcellular localizations of ATX4-GFP and ATX5-GFP. Protoplasts from wild-type (WT) plants were transformed with ATX4-GFP together with NLS-RFP or ATX5-GFP together with NLS-RFP. The signals were observed under a fluorescence microscope. NLS-RFP was used as the nucleus marker. GFP, green fluorescent protein; RFP, red fluorescent protein; NLS, nuclear localization signal. Cell images were also taken under bright field as a control. Bars, 20 lm. unclear whether mutating these genes individually would alter H3K4me2 and H3K4me3 levels. To examine the specific methyltransferase activities of ATX4 and ATX5 separately, we isolated nuclei from WT, atx4-1 and atx5-1, and determined their histone modification status. As shown in Fig. 4(a,b), H3K4me3 and H3K4me2 levels were reduced in atx4-1 and atx5-1 compared with those of WT, but H3K4me1, H3K27me3, H3K36me3 and H3K9me2 were not altered. As an alternative approach to confirm this result, we ectopically expressed ATX4 and ATX5 separately in the WT background (Fig. S4a,c). As shown in Fig. S4, the transcript levels of ATX4 and ATX5 were significantly increased in two independent ATX4 or ATX5 overexpression transgenic lines (ATX4OX-1, ATX4OX-2, ATX5OX-1 and ATX5OX-2), which were detected using qrt-pcr analysis (Fig. S4b,d). As shown in Fig. 4(c,d), H3K4me3 and H3K4me2 levels were specifically increased in the overexpression transgenic lines, but H3K4me1, H3K27me3, H3K36me3 and H3K9me2 remained unchanged. Taken together, these results suggest that ATX4 and ATX5 specifically dimethylate and trimethylate H3K4. ATX4 and ATX5 direct the landscape of transcriptional regulation To explore the genome-wide transcriptional landscape dictated by ATX4 and ATX5, we conducted RNA-seq analysis using WT, atx4-1 and atx5-1 seedlings, each with three biological replicates. Using stringent statistical and filtering criteria (see the Materials

8 New Phytologist Research 1589 Fig. 4 ARABIDOPSIS TRITHORAX4 (ATX4) and ATX5 are Arabidopsis thaliana-specific H3K4 methyltransferases. (a, c) Examination of H3K4 methyltransferase activity. Equal amounts of total proteins extracted from the nuclei of (a) wild-type (WT), atx4-1 and atx5-1 or (c) WT and ATX4 and ATX5 overexpression lines (ATX4OX-1, ATX4OX- 2, ATX5OX-1 and ATX5OX-2) were used for western blotting assays using different histone methylation antibodies, with H3 antibody as a loading control. (b, d) Relative protein levels in (b) WT, atx4-1 and atx5-1 or (d) WT and ATX4 and ATX5 overexpression lines. The amount of protein was determined by first normalizing the band intensity of specific antibodies using the software IMAGEJ. and Methods section for details), we defined 363 upregulated genes (URGs) and 415 downregulated genes (DRGs) in atx4-1 versus WT, and 248 URGs and 324 DRGs in atx5-1 versus WT (Fig. S5a; Tables S2, S3). Of note, atx4-1 and atx5-1 shared 172 URGs and 208 DRGs, respectively (Fig. S5a), indicating that, to a great extent, ATX4 and ATX5 share similar downstream transcriptional targets (Fig. 5a). We further defined these up- or downregulated genes as commonly regulated by ATX4 and ATX5 (termed common ) (Table S4), or specifically regulated by ATX4 or ATX5 (termed ATX4-specific or ATX5-specific ) (Tables S5, S6). When we performed a gene enrichment analysis using common or ATX4- or ATX5-specific gene sets (Pvalue < 0.05), we found that they were stratified into different sets of biological processes (Figs 5b, S5b). In the case of common DRGs, we identified that they were stratified into response to abiotic stimulus, response to hormone stimulus, secondary metabolic processes, response to oxidative stress, response to ABA stimulus and response to water deprivation (Fig. 5b; Table S7). In the case of common URGs, we identified that they were enriched for defense response, response to carbohydrate stimulus, nitrogen biosynthetic process, and response to oxidative stress and response to light (Fig. 5b; Table S7). These results indicate that ATX4 and ATX5 play largely redundant roles during these different biological processes. Of note, for ATX4- specific DRGs, different biological processes including lipid transport, lipid transport, oxidative reduction and responses to jasmonic acid stimulus were enriched (Fig. S5b; Table S8). In contrast, biological processes including response to abiotic stimulus, response to hormone stimulus, hormone-mediated signaling and different compound metabolic processes were overrepresented for the ATX4-specific URGs (Fig. S5b; Table S8). In the case of ATX5-specific DRGs, they were stratified into sexual reproduction, lipid localization, cell wall organization, secondary metabolic process and lipid storage, and, in the case of ATX5- specific URGs, response to abiotic stimulus, response to hormone stimulus, oxidation reduction, photosynthesis and response to light reaction GO terms were overrepresented (Fig. S5b; Table S8). These results suggest that, to an extent, ATX4 and ATX5 may also have distinct roles, probably as a result of their differentiated functions. To further confirm the RNA-sequencing results, we selected eight genes that were stratified into response to abiotic stimulus, response to abscisic acid stimulus or response to water deprivation, which were commonly regulated in atx4-1 and atx5-1, and assessed their expression by qrt-pcr analysis (Tables S9, S10). Results indicated that the expression levels of all eight genes showed the same trends in both mutants (atx4-1 and atx5-1), consistent with the RNA-seq data. Specifically, four genes, ABA-HYPERSENSITIVE GERMINATION 3 (AHG3), CYTOCHROME P450, FAMILY 707, SUBFAMILY A, POLYPEPTIDE 2 (CYP707A2), HOMOBOX 7 (HB-7) and RESPONSIVE TO DESICCATION 20 (RD20), were downregulated in atx4-1 and atx5-1 relative to WT, while four other genes,

9 1590 Research New Phytologist Fig. 5 ARABIDOPSIS TRITHORAX4 (ATX4) and ATX5 direct the landscape of transcriptional regulation in Arabidopsis thaliana. (a) Hierarchical clustering analyses of differentially expressed genes (DEGs): downregulated genes (DRGs) and upregulated genes (URGs) between wild-type (WT), atx4-1 and atx5-1. Heat color gradation in red and blue denotes an increase and decrease, respectively (log 2 (fold change)). (b) Gene ontology (GO) analyses were performed to categorize functions of DEGs using commonly down- or upregulated genes (termed common ). (c) Quantitative RT-PCR analysis of genes in the categories response to abscisic acid stimulus, response to abiotic stimulus and response to water deprivation. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) was used as an internal control. Error bars indicate SD (n = 3).

10 New Phytologist Research 1591 Fig. 6 ARABIDOPSIS TRITHORAX4 (ATX4) and ATX5 play partially redundant roles in Arabidopsis thaliana plant growth, development and abscisic acid (ABA) responses. (a) Phenotypes of wild-type (WT), atx4-1, atx5-1 and atx4-1 atx5-1 double knock-out mutants with 80 lmol m 2 s 1 light intensity. Bars: (left and middle panels) 2 cm; (right panel) 5 cm. (b d) Relative root length, fresh weight and height were measured. Error bars indicate SD (n = 3). Statistical analyses were performed comparing atx4-1 or atx5-1 and atx4-1 atx5-1. **, P-value < 0.01 (Student s t-test). (e, f) Redundant negative roles of ATX4 and ATX5 in ABA responses. WT, atx4-1, atx5-1, and the atx4-1 atx5-1 double mutant were planted on half-strength Murashige and Skoog plates supplemented with dimethyl sulfoxide (DMSO) (control) or 0.5 lm ABA and seed germination rates were measured. Error bars indicate SD (n = 3). Statistical analyses were performed comparing atx4-1 or atx5-1 and atx4-1 atx5-1. **, P-value < 0.01 (Student s t-test). (g) Quantitative RT-PCR analysis of ABA stimulus-responsive genes. GAPDH was used as an internal control. Error bars indicate SD (n = 3). Statistical analyses were performed comparing atx4-1 or atx5-1 and atx4-1 atx5-1. **, P-value < 0.01; *, P-value < 0.05 (Student s t-test).

11 1592 Research New Phytologist C-REPEAT/DRE BINDING FACTOR 1 (CBF1), GLUTAMINE-DEPENDENT ASPARAGINE SYNTHASE 1 (ASN1), SENESCENCE ASSOCIATED GENE 14 (SAG14) and PLANT U-BOX 23 (PUB23), were upregulated in atx4-1 and atx5-1 relative to WT (Fig. 5c). Taken together, these results indicate that ATX4 and ATX5 play largely redundant but also distinctly differential roles in transcriptional orchestration of genes related to stress responses. ATX4 and ATX5 play partially redundant roles in plant growth and ABA responses To examine whether ATX4 and ATX5 have redundant functions in plant development and ABA responses, we crossed atx4-1 and atx5-1 single knock-out mutants and generated an atx4-1 atx5-1 double knock-out mutant (Fig. S6). At the vegetative developmental stage, the atx4-1 atx5-1 double mutant showed dramatically reduced primary root growth and retarded leaf growth rates compared with WT as well as atx4-1 and atx5-1 single mutants (Fig. 6a d). At reproductive stages, the atx4-1 atx5-1 double mutant showed more reduced stem growth and more curling cauline leaves than WT and both single mutants (Fig. 6a d). Next, we tested whether they play functionally redundant roles in response to ABA. As shown in Fig. 6(e,f), the atx4-1 atx5-1 double mutant displayed more hypersensitive phenotypes compared with those of atx4-1 and atx5-1. We also examined the expression of genes stratified into response to abiotic stimulus which were commonly regulated by ATX4 and ATX5 (Table S9). The expression of four stress response genes was more reduced in the atx4-1 atx5-1 double mutant compared with the atx4-1 and atx5-1 single mutants (Fig. 6g), indicating that ATX4 and ATX5 play largely redundant roles during ABA responses and ABAmediated regulation of gene expression. ATX4 and ATX5 regulate ABA responses via regulation of the H3K4me3 status at the AHG3 locus Among the essential ABA signaling components, we identified that AHG3 was specifically reduced in both atx4-1 and atx5-1 compared with WT (Fig. 6g). Thus, we reasoned that the hypersensitive phenotypes of the atx4 and atx5 mutants in response to ABA might be attributable to a reduction of AHG3 transcript levels in the mutants. To test this idea, we crossed the ahg3-1 mutant (Nishimura et al., 2004; Yoshida et al., 2006; Jiang et al., 2015) with the atx4-1, atx5-1 or atx4-1 atx5-1 mutant line and generated atx4-1 ahg3-1, atx5-1 ahg3-1, and atx4-1 atx5-1 ahg3-1 mutant lines, respectively, and further tested the ABA sensitivity in terms of seed germination. As shown in Fig. 7(a,b), in the 0.5 lm ABA treatment, atx4-1 and atx5-1 displayed more ABA-hypersensitive phenotypes than WT. Intriguingly, ahg3-1, double mutants atx4-1 atx5-1, atx4-1 ahg3-1 and atx5-1 ahg3-1, and the triple mutant atx4-1 atx5-1 ahg3-1 all displayed increased sensitivities compared with atx4-1 and atx5-1. As the phenotypes of the double knock-out mutants are similar to those of ahg3-1, we suspect that AHG3 is a downstream factor of ATX4 and ATX5. As ATX4 and ATX5 are the specific H3K4me3 methyltransferases, we speculated that H3K4me3 levels at the AHG3 locus might be controlled by ATX4 and ATX5. ChIP analysis of the complementation lines expressing FLAG-tagged ATX sequences showed a significant accumulation of ATX4 and ATX5 at the AHG3 locus following ABA treatment (Fig. 7c). Correspondingly, an analysis of H3K4 trimethylation at this locus in WT and mutant lines identified that atx mutants significantly reduced the extent of ABA-dependent H3K4 trimethylation of AHG3, and that the extent of this reduction was greater in the atx4 atx5 double mutant than in either of the single mutants (Fig. 7e). This correspondence was further extended to concomitant reductions in the ABA-dependent RNA polymerase occupancy of this locus by RNA polymerase (Fig. 7g). Comparable changes were not seen in chromatin at the ACT7 locus, a negative control (Fig. 7d,f,h). This indicates that ATX4 and ATX5 redundantly regulate H3K4me3 levels of the AHG3 locus in response to ABA. Taken together, these results indicate that ATX4 and ATX5 modulate H3K4me3 levels at the AHG3 locus, which further affects RNAPII occupancies which are essential for transcriptional regulation. Previously, it was reported that the RNAPII- Ser5P association marks the transition from preinitiation complex formation in the promoter region to transcription initiation as well early transcription elongation in 5 0 gene regions (Buratowski, 2009), while the RNAPII-Ser2P association is indicative of later stages of elongation and termination. Furthermore, co-immunoprecipitation (Co-IP) experiments were performed using patx4::atx4-2xflag and patx5::atx5-2xflag transgenic lines, which showed that ATX4 and ATX5 can directly interact with total RNAPII and RNAPII-Ser5P but not RNAPII-Ser2P in vivo (Fig. S7). This may be a prerequisite for the proper occupancy of RNAPII in the promoter and gene body regions to regulate transcriptional initiation and/or elongation. Discussion H3K4me3 mediated by the trithorax proteins in different eukaryotic systems has been linked to the transcriptional activation of target genes, suggesting that the molecular basis of the trithorax effects lies in their capability to modify histones, thereby affecting chromatin structure (Vakoc et al., 2006; Howe et al., 2017). ATX4 and ATX5 share 85% homology and harbor highly conserved domains such as the PWWP, PHD and SET domains. We showed here that atx4 and atx5 mutant lines displayed reduced growth rates compared with WT; moreover, atx4-1 atx5-1 double mutants displayed enhanced growth retardation compared with the atx4-1 and atx5-1 single mutants. ATX4 and ATX5 also play partially redundant roles in stem growth during reproductive stages. A recent study reported that the atx3, atx4 and atx5 single mutants as well as different combinations of atx3 atx4, atx3 atx5 and atx4 atx5 double mutants did not display noticeable growth and developmental phenotypes under normal conditions, although the specific conditions were not specified (Chen et al., 2017). They found, however, that the atx3 atx4 atx5

12 New Phytologist Research 1593 Fig. 7 ARABIDOPSIS TRITHORAX4 (ATX4) and ATX5 regulate abscisic acid (ABA) responses via regulation of the histone H3 lysine 4 trimethylation (H3K4me3) status of the ABA-HYPERSENSITIVE GERMINATION 3 (AHG3) locus in Arabidopsis thaliana. (a, b) AHG3 acts downstream of ATX4 and ATX5 in response to ABA. Wild-type (WT), atx4-1, atx5-1, ahg3-1, atx4-1 atx5-1, atx4-1 ahg3-1, atx5-1 ahg3-1, and atx4-1 atx5-1 ahg3-1 were planted on half-strength Murashige and Skoog plates supplemented with dimethyl sulfoxide (DMSO) (control) or 0.5 lm ABA. (a) Plant images were taken 9 d after planting. (b) Germination rates were measured 9 d after planting. Error bars indicate SD (n = 3). (c, d) ATX4 and ATX5 directly associate with the AHG3 locus but not ACTIN 7 (ACT7). Chromatin immunoprecipitation (ChIP)-qPCR was performed using FLAG antibody. The x-axis denotes different genetic regions of (c) AHG3 and (d) ACT7. C, control conditions; A, ABA treatment conditions. no Ab, without antibody; Ab, with antibody. (e, f) ATX4 and ATX5 redundantly regulate H3K4me3 levels under treatment with ABA at the AHG3 locus but not ACT7. ChIP-qPCR was performed using H3K4me3 antibody. The x-axis denotes different genetic regions of (e) AHG3 and (f) ACT7. C, control conditions; A, ABA treatment conditions. (g, h) ATX4 and ATX5 affect RNA polymerase II (RNAPII) occupancies at the AHG3 locus but not ACT7. ChIP-qPCR was performed using RNAPII antibody. The x-axis denotes different genetic regions of (g) AHG3 and (h) ACT7. C, control conditions; A, ABA treatment conditions.

13 1594 Research New Phytologist triple mutant displayed growth and developmental phenotypes (Chen et al., 2017). In the present study, we used very strict physiological conditions (see the Materials and Methods section) to conduct phenotyping, and we observed phenotypes in both the atx4 and atx5 single mutants. We also observed that the retarded vegetative growth of the atx4 and atx5 mutants and abnormal reproductive developmental phenotypes of the atx5 mutant were exaggerated under relatively low light conditions. It is possible that the phenotypic manifestations in these mutants are highly sensitive to the conditions. Of note, we found that atx4 and atx5 but not atx3 displayed ABA-hypersensitive phenotypes. It is therefore likely that ATX4 and ATX5 play specific roles in the response to ABA. We further identified that ATX4 and ATX5 function in H3K4 di- and trimethylation. This is broadly consistent with findings of a recent study (Chen et al., 2017), which showed that the levels of H3K4me2 and H3K4me3 were substantially reduced in the atx3 atx4 atx5 triple mutant compared with WT based on both western blotting and genome-wide ChIP-seq profiling. In terms of the separate effect of the individual genes on H3K4me2 and H3K4me3, the Chen et al. (2017) study found that only atx5 as a single or double mutant showed evidence of reduction of the marks. By contrast, we found that both atx4 and atx5 single mutants showed significant reduction of H3K4me2 and H3K4me3 using the protein extract from nuclei. Regardless of some discrepancies between our study and that by Chen et al. (2017), it is clear that both ATX4 and ATX5 govern H3K4me2 and H3K4me3 levels in A. thaliana. It has been shown that ATX1 is involved in the dehydration stress response in both ABA-dependent and -independent pathways (Agarwal & Jha, 2010; Ding et al., 2011a). The atx1 loss-of-function mutants display decreased germination rates, larger stomatal apertures, more rapid transpiration and decreased tolerance to dehydration stress. Further investigation revealed that this deficiency is attributable to reduced ABA biosynthesis resulting from reduced expression of NINE-CIS- EPOXYCAROTENOID DIOXYGENASE 3 (NCED3) encoding an ABA de novo synthetic rate-limiting enzyme (Ding et al., 2011b). Moreover, ATX1 also affects the expression of ABA-independent genes, such as COLD-REFULATED 15A (COR15A), ALCOHOL DEHYDROGENASE 1 (ADH1) and C-REPEAT-BINDING FACTOR 4 (CBF4), suggesting that ATX1 plays essential positive roles in dehydration stress response mechanisms in A. thaliana (Ding et al., 2011a). Among ATX1 5, only ATX1, ATX4 and ATX5 displayed abiotic stress and ABA-responsive phenotypes (Ding et al., 2011a). We therefore compared the gene expression profiles regulated by these genes. Previously, Saleh et al. (2008) performed a microarray analysis using 4-wk-old WT and atx1, with both conditions and developmental stage similar to those we used here. Thus, we directly compared the differentially up- or downregulated genes (UGRs or DGRs) between atx1 and atx4-1, between atx1 and atx5-1 and between atx4-1 and atx5-1 (Tables S11, S12). As shown in Fig. S8(a), atx4-1 and atx5-1 share nearly 17% of DGRs and 13.8% of UGRs, while atx1 and atx4-1 share only 3.5% of DGRs and 2.7% of UGRs, and atx1 and atx5-1 share 3.9% of DGRs and 2.5% of UGRs. We further performed GO analysis for the atx1 versus atx4-1-specific, atx1 versus atx5-1-specific and atx1 versus atx4-1 versus atx5-1 UGRs and DGRs (Table S13). Intriguingly, we did not find the GO terms response to abiotic stimulus and response to abscisic acid in atx1 versus atx4-1-specific and atx1 versus atx5-1-specific UGRs and DGRs. In atx1 versus atx4-1 versus atx5-1 UGRs and DGRs, we identified different GO terms including response to abiotic stimulus, response to abscisic acid, response to salt stress and response to osmotic stress (Fig. S8b). To further specifically dissect these genes, we compared the UGRs or DGRs that were stratified into these functional categories (Fig. S9; Table S14). As shown in Fig. S9, we found that most UGRs and DGRs displayed highly similar patterns in atx4-1 and atx5-1. Although a few UGRs and DGRs which were identified in atx4 or atx5 could also be detected in atx1, itis likely that ATX1 regulates different subsets of stress and ABAresponsive genes. These results indicate that different TrxG genes exert distinctive physiological readouts, although they possess the same enzymatic activities that trimethylate histone H3K4. At the molecular level, it is possible that harboring different domain structures somehow conferred their targeting specificities to different genomic loci. It has been reported that ATX1 modulates transcription through two distinct roles, at promoters and at transcribed sequences (Ding et al., 2011b). At promoter regions, ATX1 is required for TATA-binding protein and RNAPII recruitment (Ding et al., 2011b). After preinitiation complex (PIC) formation and phosphorylation of RNAPII C-TERMINAL DOMAIN (CTD) at Ser5P, in which step there is a transition of transcription from initiation to the elongation phase (Kwak et al., 2013), RNAPII recruits ATX1 to this position and forms occupancy peaks at c nt downstream of the TRANSCRIPTION START SITE (TSS) (Ding et al., 2011b). We observed that ATX4 and ATX5 also generated an occupancy peak c nt downstream of the AHG3 locus, and atx4-1 and atx5-1 loss-of-function mutants displayed reduced RNAPII. This indicates that ATX4 and ATX5 might have similar functions during transcriptional regulation at the molecular level. We also identified that ATX4 and ATX5 can directly interact with total RNAPII and RNAPII-Ser5P but not with RNAPII-Ser2P in vivo. These results are similar to the observation that the SET domain of ATX1 can bind to total RNAPII and RNAPII-Ser5P but not RNAPII-Ser2P in vivo (Ding et al., 2011b). It has been reported that ATX1 has dual functions in both facilitating PIC assembly and generating H3K4me3 as an activating mark for transcriptional elongation. Our data support a model in which ATX4 and ATX5 are recruited first to the promoter of AHG3 and then to a c nt position in AHG3 by direct binding to the Ser5 phosphorylated CTD of Polymerase (Pol II). The inability of ATX4 and ATX5 to bind to Ser2P-modified CTD of the Pol II tail might indicate a mechanism for ATX4 and ATX5 dissociation from the elongation transcription complex. ATX4 and ATX5 remain at the end of the genes c nt downstream of the TSSs of AHG3, establishing the characteristic H3K4me3 peaks at