Arabidopsis O‐GlcNAc transferase SEC activates histone methyltransferase ATX1 to regulate flowering

Abstract

Post‐translational modification of proteins by O‐linked β‐N‐acetylglucosamine (O‐GlcNAc) is catalyzed by O‐GlcNAc transferases (OGTs). O‐GlcNAc modification of proteins regulates multiple important biological processes in metazoans. However, whether protein O‐GlcNAcylation is involved in epigenetic processes during plant development is largely unknown. Here, we show that loss of function of SECRET AGENT (SEC), an OGT in Arabidopsis, leads to an early flowering phenotype. This results from reduced histone H3 lysine 4 trimethylation (H3K4me3) of FLOWERING LOCUS C (FLC) locus, which encodes a key negative regulator of flowering. SEC activates ARABIDOPSIS HOMOLOG OF TRITHORAX1 (ATX1), a histone lysine methyltransferase (HKMT), through O‐GlcNAc modification to augment ATX1‐mediated H3K4me3 histone modification at FLC locus. SEC transfers an O‐GlcNAc group on Ser947 of ATX1, which resides in the SET domain, thereby activating ATX1. Taken together, these results uncover a novel post‐translational O‐GlcNAc modification‐mediated mechanism for regulation of HKMT activity and establish the function of O‐GlcNAc signaling in epigenetic processes in plants.

Introduction

The O‐linked β‐N‐acetylglucosamine (O‐GlcNAc) modification of nuclear and cytoplasmic proteins is ubiquitous and essential for a number of biological processes. The dynamic addition and removal of O‐GlcNAc at serine and threonine residues are catalyzed by two highly conserved types of enzymes, O‐GlcNAc transferase (OGT) and O‐GlcNAcase (OGA), respectively (Haltiwanger et al, 1990; Dong & Hart, 1994; Hart et al, 2007). Proteins modified with O‐GlcNAc include transcription factors, polymerases, proteasomes, and RNA‐processing proteins (Hart et al, 2007), indicating that protein O‐GlcNAcylation has significant biological functions. O‐GlcNAc modification of core histones has also been identified and is considered as part of the histone code (Sakabe et al, 2010). Recent studies revealed that histone O‐GlcNAcylation is coordinated with other post‐translational modifications. For example, H2B O‐GlcNAcylation at Ser112 facilitates ubiquitination at Lys120 and thus gene transcription (Fujiki et al, 2011). In addition to histones, some proteins associated with chromatin remodeling are modified with O‐GlcNAc. In the MCF7 cell line, OGT associates with the enhancer of zeste homolog 2 (EZH2) in the polycomb repressive complex 2 (PRC2), and O‐GlcNAc modification of EZH2 at Ser75 plays a role in maintaining EZH2 stability to catalyze H3K27me3 (Chu et al, 2014). O‐GlcNAcylation of MLL5β, an isoform of the mammalian trithorax histone lysine methyltransferase (HKMT) MLL5, is critical for recruitment and assembly of the MLL5β‐AP‐1 transcription activation complex (Nin et al, 2015).

In contrast to the extensive work in animals, the effect of O‐GlcNAcylation on epigenetic modulation in plants has been largely unexplored. Unlike for mammalian cells, in which OGT is encoded by a single‐copy gene (Kreppel et al, 1997; Shafi et al, 2000), two OGTs, SECRET AGENT (SEC) and SPINDLY (SPY), have been predicted in Arabidopsis (Hartweck et al, 2002, 2006). Recently, it was reported that the DELLA protein RGA is O‐GlcNAcylated. O‐GlcNAcylation inhibits RGA binding to PHYTOCHROME‐INTERACTING FACTOR 3 (PIF3), PIF4, JASMONATE ZIM‐domain 1 (JAZ1), and BRASSINAZOLE‐RESISTANT 1 (BZR1), which in turn coordinately regulate multiple signal pathways in Arabidopsis. The endogenous O‐GlcNAc modification of RGA is catalyzed by SEC but not SPY (Zentella et al, 2016). SPY was later determined to be a novel O‐fucosyltransferase that catalyzes O‐fucosylation of DELLA (Zentella et al, 2017). In tobacco (Nicotiana tabacum), core histones are modified by O‐GlcNAc and interact with the lectin Nictaba, which is speculated to have a role in chromatin folding (Schouppe et al, 2011).

SET‐domain proteins function primarily as histone methyltransferases to catalyze histone methylation modification. Most SET proteins function as HKMTs; in addition, some members catalyze methylation of arginine on both histone and non‐histone proteins (Ng et al, 2007; Niu et al, 2007; Wang et al, 2007). Trithorax group (TrxG) proteins can catalyze H3K4 methylation. In Arabidopsis, five Trithorax homologs (ATX1–ATX5) have been identified and belong to the Trx subfamily of SET‐domain proteins based on their conserved SET domain and other characteristic domains; in addition, seven proteins were classified as ARABIDOPSIS TRITHORAX‐RELATED (ATXR1–ATXR7; Avramova, 2009; Pontvianne et al, 2010).

Arabidopsis FLOWERING LOCUS C (FLC) functions as a central floral repressor (Michaels & Amasino, 1999), and the epigenetic state of FLC is critical for regulation of flowering in response to vernalization (Bastow et al, 2004; Whittaker & Dean, 2017). FLC expression affects the floral transition and is regulated by its active or repressive histone modification state (He, 2009). The atx1 mutant exhibits an early flowering phenotype correlated with reduced FLC expression; ATX1 functions as an activator of FLC by targeting to FLC chromatin to establish an H3K4me3 modification mark (Pien et al, 2008). ATXR7 regulates flowering time by transcriptionally activating FLC expression, and atx1‐2 atxr7‐2 double mutant shows additive effects on the regulation of flowering time and H3K4 methylation at the FLC locus (Tamada et al, 2009). Histone methylation plays multiple significant roles during plant development (Liu et al, 2010; Pontvianne et al, 2010; Thorstensen et al, 2011); however, the mechanisms regulating the activity of histone methyltransferases remain poorly understood.

Previous work in winter wheat (Triticum aestivum) indicated that O‐GlcNAcylation‐dependent interaction between the lectin protein vernalization‐related 2 (VER2) and the RNA‐binding protein TaGRP2 regulates vernalization response and flowering by releasing TaGRP2‐mediated repression of TaVRN1 (Xiao et al, 2014). In this study, we investigated the effect of loss of function of O‐GlcNAc transferase SEC on flowering‐time regulation using a null mutant, sec‐5, in Arabidopsis. The sec‐5 mutant exhibits an early flowering phenotype with down‐regulated expression of FLC. Further biochemical and genetic studies indicated that the histone methyltransferase ATX1 is activated in an O‐GlcNAcylation‐dependent manner. Together, our results not only identify the role of the O‐GlcNAc transferase SEC in the epigenetic mediation of floral transition but also reveal a novel mechanism regulating HKMT activity in plants.

Results

Loss of SEC function results in early flowering and reduced FLC expression

To explore whether the O‐GlcNAc transferase SEC is involved in flowering‐time regulation in Arabidopsis, we identified two homozygous mutant lines in the Col‐0 background, sec‐4 and sec‐5, which contain T‐DNA insertions in the promoter region and in exon 2 of SEC, respectively (Fig 1A and B). Quantitative real‐time PCR (qRT–PCR) analysis to detect a transcript between exon 18 and 19 of SEC showed that SEC mRNA expression was not altered in the sec‐4 mutant but was dramatically lower in the sec‐5 mutant as compared to the wild type (Fig 1C). Further phenotypic observation showed that the sec‐4 mutant had a similar flowering phenotype to Col‐0, while the sec‐5 mutant exhibited a reproducible weak early flowering phenotype with reduced rosette leaf number and flowering time compared with wild type under long‐day (LD) condition (Fig 1D, Table 1).

Figure 1. Identification and characterization of sec mutants.

1. Schematic genomic structure of the SEC locus. Exons are shown as black boxes, promoter and introns are indicated as lines, and 5′ and 3′ untranslated regions are shown as gray boxes. T‐DNA insertion positions are indicated by triangles. P represents the SEC transcript detected by qRT–PCR.
2. Identification of sec homozygotes. LB represents the left border primer of the T‐DNA insertion. LP and RP represent the left and right genomic primers, respectively.
3. qRT–PCR analysis of SEC mRNA levels in sec‐4 and sec‐5 mutants. The expression level was normalized to that of TUBULIN, a reference gene for qRT–PCR. Independent biological experiments were repeated three times, and one representative result is shown here. Data shown are means ± standard deviation (s.d.), n = 3.
4. The early flowering phenotype of sec‐5 mutant under long‐day (LD) conditions. Scale bar: 1 cm.
5. qRT–PCR analysis of the expression levels of FLC, SOC1, and MAF1‐3 in 12‐d‐old vegetative Col‐0, sec‐4, and sec‐5 plants. Independent biological experiments were repeated three times, and one representative result is shown here. The expression level was normalized to that of TUBULIN. Data are means ± s.d., n = 3. Statistical significance (two‐tailed t‐test) with *P < 0.05.
6. The sec‐5 mutant exhibits an early flowering phenotype under short‐day (SD) conditions. Scale bar: 1 cm.
7. qRT–PCR analysis of FLC and SOC1 expression in Col‐0 and sec‐5 plants under SD conditions. The expression level was normalized to that of TUBULIN. Experiments were repeated three times, and one representative result is shown here. Data are mean ± s.d., n = 3. Statistical significance (two‐tailed t‐test) with *P < 0.05.

Source data are available online for this figure.

Table 1.

The sec‐5 mutant exhibits an early flowering phenotype under LD conditions

Genotype	Visible buds (%)	Bolting rates (%)	Days to first flower opening	Rosette leaf no.	n
Col‐0	22.2 ± 1.7	8.3 ± 0.4	32.7 ± 1.2	11.5 ± 1.0	72
sec‐4	23.6 ± 1.2	9.6 ± 1.7	32.3 ± 0.6	11.3 ± 1.1	72
sec‐5	63.9 ± 1.4**	47.2 ± 1.2**	29.3 ± 0.6*	9.4 ± 0.9**	72

LD, long‐day conditions.

Percentages of plants with visible buds were scored at 21 days after germination, and bolting rates were calculated at 24 days after germination. Experiments were repeated three times, and the values are the means ± standard deviation (s.d.).

Significant differences at **P < 0.01 between Col‐0 and mutants, and at *P < 0.05 by two‐tailed t‐test. n, total numbers of plants used in statistical analysis.

The qRT–PCR analysis showed that the key gene of the autonomous and vernalization pathways (Kim et al, 2009), FLC, was markedly down‐regulated in the sec‐5 mutant. In addition, mRNA expression of the FLC‐clade genes MADS AFFECTING FLOWERING 1 (MAF1), MAF2 and MAF3 were slightly down‐regulated, and SUPPRESSOR OF OVEREXPRESSION OF CONSTANS 1 (SOC1) was up‐regulated, which is consistent with the flowering phenotype of sec‐5. Nevertheless, the expression level of these genes in sec‐4 was similar to that in Col‐0 plants (Fig 1E). Further mRNA expression analysis showed that four genes flanking SEC, At3g04220, At3g04230, At3g04250, and At3g04260, were unaltered in sec‐5 (Fig EV1A). Thus, the sec‐5 mutant was used for further functional analysis.

Figure EV1. Characterization and flowering phenotype of sec‐5 under SD conditions with vernalization, and flowering phenotype rescue of sec‐5 .

1. qRT–PCR analysis of the expression of genes upstream and downstream of SEC in the sec‐5 mutant. The expression level was normalized to that of TUBULIN. Experiments were repeated three times, and one representative result is shown here. Data are mean ± s.d., n = 3.
2. Flowering phenotype of Col‐0 and sec‐5 plants with vernalization treatment under SD conditions. Plants were vernalized for 20 days before being moved to SD conditions. Scale bar: 1 cm.
3. Rosette leaf number analysis of Col‐0 and sec‐5 plants under SD with or without vernalization treatment. A total number of 62 plants were scored for rosette leaf number analysis of each line. Data are mean ± s.d., statistical significance (two‐tailed t‐test) with **P < 0.01.
4. qRT–PCR analysis of CO, GI, FPA, FLD, FY, and FLK mRNA expression levels in Col‐0 and sec‐5 plants under LD conditions. The expression level was normalized to that of TUBULIN. Data are mean ± s.d., n = 3. Experiments were repeated three times, and one representative result is shown here.
5. The early flowering phenotype of the sec‐5 mutant was rescued by pSEC::SEC under LD conditions. Three independent transgenic lines are shown. Scale bar: 1 cm.
6. qRT–PCR analysis of SEC and FLC transcription levels in Col‐0, sec‐5 and pSEC::SEC transgenic lines in a sec‐5 background. The expression level was normalized to that of TUBULIN. Data are mean ± s.d. of three independent biological replicates, n = 3. Statistical significance (two‐tailed t‐test) with *P < 0.05.
7. Flowering‐time analysis of plant lines in (E). Error bars indicate s.d. of three biological replicates. Statistical significance (two‐tailed t‐test) with *P < 0.05.
8. Rosette leaf number analysis of plant lines in (E). Three biological replicates were conducted, and a total number of 52 plants were scored for rosette leaf number analysis. Error bars indicate s.d. Statistical significance (two‐tailed t‐test) with **P < 0.01.

When grown under short‐day (SD) conditions, sec‐5 plants showed a more obvious early flowering phenotype. The sec‐5 mutant also exhibited down‐regulation of FLC and up‐regulation of SOC1 transcript levels compared with wild‐type plants (Fig 1F and G, Appendix Table S1). In addition, vernalization treatment under SD conditions accelerated flowering of both sec‐5 and wild‐type Col‐0 plants, and reduced the difference in rosette leaf numbers between sec‐5 and Col‐0 plants (Fig EV1B and C, Appendix Table S1). Further qRT–PCR analysis indicated that the transcript levels of key genes in the photoperiod and autonomous pathways, CONSTANS (CO), GIGANTEA (GI), FPA, FLOWERING LOCUS D (FLD), FY, and FLOWERING LOCUS K (FLK), were unaltered in sec‐5 under LD conditions (Fig EV1D).

We generated transgenic plants expressing full‐length SEC cDNA driven by the endogenous SEC promoter in the sec‐5 background. These pSEC::SEC transgenic lines showed similar flowering‐time phenotypes and FLC transcript levels to Col‐0 under LD conditions (Fig EV1E and F), indicating that the early flowering phenotype of sec‐5 was genetically complemented by pSEC::SEC transformation (Fig EV1G and H). The sec‐1 (in Wassilewskija background) and sec‐2 (in Col‐0 background) mutants have previously been reported to have T‐DNA insertions in exon 12 and intron 22, respectively (Hartweck et al, 2002, 2006). The number of days to flowering was unchanged in sec‐1 and sec‐2 as compared with wild type, but both mutants show reduced total leaf numbers at flowering, which has been suggested to result from a reduced leaf production rate of sec‐1 and sec‐2 mutants. The sec‐1 did not show a flowering phenotype neither in long days nor short days (Hartweck et al, 2002, 2006), and this may be an ecotype difference. Another allele, sec‐3 (in Landsberg erecta background), has been shown to have reduced gibberellin (GA) response and a dwarf phenotype, but its flowering phenotype was not reported (Zentella et al, 2016). Our findings support the possibility that sec‐5 accelerates flowering by regulating FLC transcription.

To determine the genetic relationship between SEC and FLC, we generated 35S::SEC‐overexpressing transgenic plants in C24 and flc‐20 (C24 background), respectively. 35S::SEC overexpression in C24 resulted in a late‐flowering phenotype; however, flc‐20 35S::SEC plants showed almost the same flowering phenotype as flc‐20, indicating that SEC might be functionally dependent on FLC in the same genetic pathway for flowering (Fig EV2A–C).

Figure EV2. Genetic interaction analysis between SEC and FLC, and flowering phenotype comparison between Col‐0, sec‐5, atx1‐2, sec‐5 atx1‐2, and flc‐3 plants.

1. Flowering phenotype of C24, C24 35S::SEC, flc‐20, and flc‐20 35S::SEC plants.
2. Relative SEC mRNA expression levels of lines in (A). The expression level was normalized to that of TUBULIN. Data are mean ± s.d., n = 3. Experiments were repeated three times, and one representative result is shown here.
3. Rosette leaf number analysis of plants in (A). The values are mean ± s.d.; a total of 40 plants were scored for each genotype. Statistical significance (two‐tailed t‐test) with **P < 0.01.
4. Flowering phenotype of Col‐0, sec‐5, atx1‐2, sec‐5, sec‐5 atx1‐2, and flc‐3 plants. Scale bar: 1 cm.
5. Rosette leaf numbers of plant lines in (D). The values are mean ± s.d.; a total number of 62 plants were scored for rosette leaf number analysis. Statistical significance (two‐tailed t‐test) with **P < 0.01.
6. Expression level of ATX1 mRNA in Col‐0 and sec‐5 plants. Twelve‐day‐old plants were used for total RNA extraction. Three independent biological replicates were done, and one representative result is shown here. Data are mean ± s.d., n = 3. The data were normalized to that of UBIQUITIN. ATX1 mRNA levels were analyzed for significant differences by two‐tailed t‐test. n.s., not significant.

Taken together, our data demonstrate that SEC regulates flowering time in an FLC‐dependent manner, and endogenous SEC plays a negative role in regulating flowering time by activating FLC transcription.

Loss of SEC function alters the histone modification of FLC chromatin

To investigate whether histone modifications are involved in the down‐regulation of FLC in the sec‐5 mutant, we analyzed the H3K4me2, H3K4me3, H3K36me3, and H3K27me3 modification states, which are associated with either active or repressive states of FLC chromatin, by chromatin immunoprecipitation (ChIP). Because the regions near the transcription start site play important roles in FLC transcription activity (Bastow et al, 2004), we assayed three regions corresponding to the FLC promoter (region P1), transcription start site (region P2), and intron 1 (region P3; Fig 2A, Appendix Fig S1A) for enrichment of various histone modifications. In sec‐5 plants, as compared with Col‐0, the H3K4me2 levels were slightly elevated in P1, P2, and P3 (Fig 2B and Appendix Fig S1B), whereas H3K4me3 levels (associated with active transcription of FLC) were dramatically decreased, particularly in P2 (Fig 2C and Appendix Fig S1C). H3K36me3, another chromatin modification associated with active transcription of FLC (Yang et al, 2014), showed nearly identical levels in sec‐5 and wild‐type plants in all three regions (Fig 2D and Appendix Fig S1D). The repressive mark H3K27me3 was increased in all three regions (Fig 2E and Appendix Fig S1E). These data indicate that loss of SEC significantly decreases the deposition of the H3K4me3 mark and also increased H3K27me3 levels in FLC chromatin. H3K4me3 is involved in FLC activation, and thus, the reduced expression of FLC in sec‐5 correlates with down‐regulation of H3K4me3 levels at the FLC locus. The data support the proposal that SEC negatively regulates flowering time by maintaining an active FLC chromatin state.

Figure 2. Loss of SEC function alters histone modification states on FLC chromatin.

1. Schematic of the FLC locus. Exons are indicated with black boxes; promoter and introns are shown as lines. P1, P2, and P3 refer to genomic regions examined by ChIP.
2. ChIP‐qPCR assay of H3K4me2 levels of indicated regions at FLC chromatin.
3. ChIP‐qPCR assay of H3K4me3 levels of indicated regions at FLC chromatin.
4. ChIP‐qPCR assay of H3K36me3 levels of indicated regions at FLC chromatin.
5. ChIP‐qPCR assay of H3K27me3 levels of indicated regions at FLC chromatin.

Data information: For ChIP analysis, 12‐day‐old plants were collected and three independent experiments were conducted. Each bar represents the mean ± s.d. of three independent experiments, n = 3. The relative abundance was normalized to the input.

ATX1 interacts with and functionally depends on SEC

ATX1 directly mediates H3K4me3 modification at the FLC locus (Pien et al, 2008). OGT O‐GlcNAcylates MLL5β, an isoform of the mammalian TrxG protein MLL5, and acts as a component of the transcription activation complex (Nin et al, 2015). This prompted us to explore the possibility that SEC might regulate H3K4me3 modification of the FLC locus by catalyzing O‐GlcNAc modification of the histone lysine methyltransferase ATX1. A yeast two‐hybrid assay showed that ATX1 directly interacts with either SEC or SEC‐N, a truncated SEC including only the TPR domains at the N‐terminus, indicating that these N‐terminal TPR domains mediate the interaction between SEC and ATX1 (Fig 3A). To define whether ATX1 is a potential substrate of SEC with O‐GlcNAc modification, we expressed HA‐tagged ATX1 alone or together with FLAG‐tagged SEC in tobacco mesophyll cells. ATX1‐HA was immunoprecipitated with anti‐HA antibody and then probed with the antibody CTD110.6 which recognizes O‐GlcNAc sites on serine and threonine residues to detect O‐GlcNAc modification on ATX1. ATX1‐HA co‐expressed with FLAG‐SEC showed positive O‐GlcNAc modification, while no signal for the ATX1‐HA expressed alone (Fig 3B).

Figure 3. ATX1 interacts with and is functionally dependent on SEC .

• A
  Yeast two‐hybrid assay of the interaction between ATX1 and SEC. The full‐length coding sequence (CDS) of ATX1 was cloned into the bait vector pGBKT7. A full‐length SEC CDS and a truncated fragment were cloned into the prey vector pGADT7 to express complete SEC or truncated SEC‐N (including only TPR domains at the N‐terminus), respectively. AD vectors expressing SEC or SEC‐N were each cotransformed with BD‐ATX1. SD/2, SD/‐Leu/‐Trp medium; SD/4, SD/‐Ade/‐His/‐Leu/‐Trp medium. Cotransformed pGBKT7‐53/pGADT7‐T was used as a positive control.
• B
  Immunoblot analysis for O‐GlcNAcylation of ATX1 by SEC using anti‐CTD110.6 antibody. HA‐tagged ATX1 was expressed alone or together with Flag‐tagged SEC in tobacco leaves. Nuclear proteins were extracted from tobacco mesophyll cells, and expression of FLAG‐SEC and ATX1‐HA was confirmed by Western blotting.
• C
  Flowering phenotype of 35S::ATX1‐HA overexpression transgenic lines in Col‐0 and sec‐5 backgrounds, respectively. Scale bar = 1 cm.
• D
  Immunoblot analysis using anti‐HA antibody to confirm ATX1‐HA expression in Col‐0 and 35S::ATX1‐HA‐overexpressing lines shown in (C). ACTIN was used as a loading control.
• E–G
  Flowering‐time phenotype analysis for Col‐0, sec‐5, and 35S::ATX1‐HA transgenic lines in (C). The percentages of plants with visible buds and the bolting rates were calculated on day 21 and day 24 after plant germination, respectively. Experiments were repeated three times, and for each line, a total of 62 plants were counted for statistical analysis of rosette leaf numbers, visible buds percent, and bolting rates. Data are mean ± s.d., statistical significance (two‐tailed t‐test) with **P < 0.01. n.s., not significant.

Source data are available online for this figure.

To further understand whether ATX1 is functionally dependent on SEC, we generated 35S::ATX1‐HA overexpression transgenic lines in both the Col‐0 and sec‐5 backgrounds. Compared Col‐0 plants, homozygous lines overexpressing ATX1‐HA in Col‐0 displayed a late‐flowering phenotype with significantly increased rosette leaf numbers, and dramatically reduced visible buds at 21 days and bolting rate at 24 days after germination. In contrast, sec‐5 35S::ATX1‐HA lines exhibited similar rosette leaf numbers and flowering‐time phenotypes as that of sec‐5 (Fig 3C‐G). Loss of SEC function significantly suppressed the late‐flowering phenotype of 35S::ATX1‐HA plants, suggesting that SEC is genetically necessary for the function of ATX1 in floral transition.

We also generated a sec‐5 atx1‐2 double mutant, and the flowering phenotype analysis showed that sec‐5 atx1‐2 exhibited an early flowering phenotype similar to those of sec‐5, atx1‐2, and flc‐3 mutants (Fig EV2D–F, Appendix Table S2), indicating that SEC and ATX1 regulate flowering time through the same pathway.

SEC activates ATX1 by O‐GlcNAc modification

To explore the regulation of ATX1 activity by SEC, we used the catalytic domains of each protein, which are easier to recombinantly purify than the full‐length proteins, for further biochemical analysis. The recombinant ATX1ΔN includes the SET domain (amino acids 899–1017) at the C‐terminus, along with the regions immediately upstream (amino acids 801–898) and downstream (amino acids 1,018–1,062) of SET (Appendix Fig S2A; Liu et al, 2015). The recombinant SECΔN (amino acids 592–952) contains the complete OGT catalytic domain at the C‐terminus of SEC (Appendix Fig S2B). The truncated versions of these two proteins, His‐tagged ATX1ΔN and SECΔN, were expressed in Escherichia coli and affinity purified (GenScript, Appendix Fig S2C). Biochemical analysis showed that recombinant SECΔN exhibited OGT activity in vitro and that O‐GlcNAcylation of ATX1ΔN could be recognized by the monoclonal antibody CTD110.6. The subsequent β‐elimination assay confirmed O‐GlcNAc modification of ATX1ΔN (Fig 4A). Combined with the down‐regulation of H3K4me3 levels at the FLC locus in the sec‐5 mutant (Fig 2C), these results suggest that SEC may directly regulate HKMT activity of ATX1 by O‐GlcNAcylation. ATX1ΔN activity was analyzed using unlabeled S‐adenosyl‐L‐methionine (SAM) as the methyl donor. Immunoblot analysis with an antibody against H3K4me3 showed that ATX1ΔN was capable of catalyzing H3K4me3 of recombinant histone H3. When SECΔN and UDP‐GlcNAc were added to the reaction system to allow the O‐GlcNAcylation reaction in advance, the activity of ATX1ΔN was notably enhanced compared with the background activity of ATX1ΔN. In addition, a recombinant mutated H3 (H3K4A) with Lys4 replaced by alanine was not methylated by ATX1ΔN (Fig 4B), indicating that SEC activates ATX1‐regulated H3K4 trimethylation.

Figure 4. SEC O‐GlcNAcylates and activates ATX1 in vitro and in vivo .

1. Detection and conformation of O‐GlcNAc modification of His‐ATX1ΔN catalyzed by His‐SECΔN in vitro. His‐ATX1ΔN (expressing residues 592–952 of the C‐terminus) and His‐SECΔN (expressing residues 801–1,062 of the C‐terminus) were recombinantly expressed and affinity purified separately. O‐GlcNAcylation of His‐ATX1ΔN was detected by anti‐CTD110.6 antibody and further conformed by β‐elimination analysis.
2. His‐SECΔN activates His‐ATX1ΔN in vitro. Histone methyltransferase activity of ATX1ΔN was detected with or without recombinant SECΔN. H3K4A: mutated H3 in which the fourth amino acid, a lysine (K), was replaced with alanine (A).
3. Analysis of ATX1 O‐GlcNAc modification in wild‐type and sec‐5 plants. Total soluble protein extracts from 12‐day‐old seedlings were subjected to SDS–PAGE followed by immunoblotting using the indicated antibodies. CBB: Coomassie brilliant blue staining, showing relative protein loading amount.
4. Loss of SEC function reduced ATX1 activity in Arabidopsis. Nuclear proteins were extracted, and ATX1 was immunoprecipitated with anti‐ATX1 antibody from wild‐type and sec‐5 mutant plants, respectively, and then used for histone H3K4 methyltransferase activity analysis with recombinant H3 as catalyzing substrate. Band intensities were quantified with ImageJ. The H3 signal was first normalized by input signal and then was used for H3K4me3 signal normalization.
5. Comparison of ATX1 protein levels in Col‐0 and sec‐5 plants at stages before and after flowering. Seedlings were collected for protein extraction at 21 days (before flowering) and 35 days (after flowering) after seeds were planted on plates. Anti‐ATX1 antibody was used for immunoblot assay.

Source data are available online for this figure.

To investigate further whether SEC O‐GlcNAcylates ATX1 in vivo, we analyzed the effect of loss of SEC function on the O‐GlcNAcylation level of ATX1 in Arabidopsis. The immunoblot assay showed that O‐GlcNAc modification of ATX1 was detected in Col‐0 but not sec‐5 plants, indicating that SEC is necessary for ATX1 O‐GlcNAcylation in vivo. Meanwhile, sec‐5 plants showed reduced ATX1 protein level (Fig 4C). But qRT–PCR analysis showed that loss of SEC has no effect on the transcript level of ATX1 (Fig EV2F). Subsequently, ATX1 was immunoprecipitated from 12‐d‐old Col‐0 wild‐type and sec‐5 plants using anti‐ATX1 polyclonal antibody (GenScript) and then used for an HKMT activity assay. As expected, SEC mutation resulted in reduction in the endogenous HKMT activity of ATX1 (Fig 4D). To further investigate the effect of SEC mutation on ATX1 protein level, we immunodetected endogenous ATX1 protein in plants at development stages before and after flowering. The data revealed that loss of SEC function decreased ATX1 protein level (Fig 4E). Together, these results showed that SEC plays a role in activating the methyltransferase ATX1 through O‐GlcNAc modification and also in maintaining its stability in Arabidopsis.

S947 is a key site for O‐GlcNAcylation‐dependent activation of ATX1 by SEC

The role of O‐GlcNAcylation in regulating ATX1 activity prompted us to look for functional O‐GlcNAc sites in ATX1. The recombinant His‐ATX1ΔN was catalyzed by His‐SECΔN for O‐GlcNAcylation modification in vitro, and then for higher‐energy collisional dissociation (HCD)–mass spectrometry (MS) analysis to identify potential O‐GlcNAc modification sites. MS analysis identified a peptide in the ATX1 SET domain that has potential sites of O‐GlcNAc modifications at S947 and/or T953 (Fig EV3). We further investigated the possibility of O‐GlcNAcylation on S947. To explore the potential role of S947 O‐GlcNAcylation in regulating ATX1 HKMT activity, we mutated S947 to A for recombinant expression of the mutant protein His‐ATX1ΔN‐m in E. coli. Simultaneously, we replaced all 12 serine and threonine residues in the ATX1 SET domain with alanine for expression of another mutant protein, His‐ATX1ΔN‐12m. Compared with His‐ATX1ΔN, His‐ATX1ΔN‐m and His‐ATX1ΔN‐12m showed similarly reduced O‐GlcNAcylation levels (Fig 5A), indicating that S947 is a key O‐GlcNAc site in SET domain of ATX1.

Figure EV3. Identification of potential O‐GlcNAcylation site(s) in ATX1 SET domain.

1. The mass spectrometry (MS) analysis showed that the serine and/or threonine residue(s) in the EQLIYNSMVGAGTYMFR peptide are (is) modified with O‐GlcNAc.
2. Partial enlarged detail of (A) from 100 to 400 of m/z.

Figure 5. S947 of ATX1 is necessary for its O‐GlcNAc modification and activation by SEC .

1. Mutation of either S947 alone or all 12 serine and threonine residues in the SET domain reduced O‐GlcNAc modification level of ATX1ΔN.
2. Site mutation of S947 or all 12 serine and threonine residues in the ATX1 SET domain inhibited the activation of ATX1 by SEC; moreover, mutation of five conserved amino acids in the SEC functional domain inhibited SEC activity.
3. Quantitative analysis of HKMT activity of ATX1ΔN, ATX1ΔN‐12 m, and ATX1ΔN‐m and the effect of SECΔN and ΔSEC5m in activating those proteins. Experiments were repeated three times. Error bars are s.d. His‐ΔSEC5m and His‐SECΔN were used as the control.

Source data are available online for this figure.

Based on the previously reported binary and ternary crystal structure of human OGT (Lazarus et al, 2011), we created a model of the overall structure of SEC protein using the Phyre2 web portal (http://www.sbg.bio.ic.ac.uk/phyre2; Kelley et al, 2015; Fig EV4A). We found a strong resemblance between SEC and OGT. Compared with human OGT, a small motif corresponding to amino acids 727–929 of human OGT is deleted in the O‐GlcNAc transferase domain of SEC; however, this motif does not participate in the binding of substrates and peptides to OGT (Lazarus et al, 2011). According to our prediction, there are five extremely conserved amino acids, Phe540, His541, His604, Gln776, and Lys779, located within the catalytic region, which are critical for SEC binding with substrates (Fig EV4B). Among these, His541 and His604 are reported to be conserved and required for SEC activity (Zentella et al, 2016). A truncated version of SEC with these five residues replaced with alanine (amino acids 539–952, containing the OGT domain of SEC and 53 amino acids immediately upstream, designated ΔSEC5m) was expressed with His tag in E. coli and affinity purified. The in vitro biochemical activity assay showed that, compared with His‐SECΔN, His‐ΔSEC5m activated His‐ATX1ΔC relatively weakly. Meanwhile, the single‐site mutation (His‐ATX1ΔN‐m, S947A) and the simultaneous mutation of 12 sites (His‐ATX1ΔN‐12m) in the SET domain of ATX1 reduced its activation by SEC in vitro (Fig 5B).

Figure EV4. SEC structure and functionally conserved amino acids.

1. SEC structure model. SEC N‐terminal TPR domain (partial) is colored in gold (residues 50–538), residues 510–754 in green, and the C‐terminal domain (residues from 755 to end) in blue. The magenta sphere represents the peptide binding to human OGT.
2. Analysis of conserved amino acids for substrate binding of HsOGT and SEC. Conserved residues are shown in red.

To further verify the role of His‐SECΔN in activating His‐ATX1ΔN, we conducted an in vitro quantitative HKMT activity assay (EpiGentek). The results showed that His‐ATX1ΔN was strongly activated by His‐SECΔN and less effectively activated by His‐ΔSEC5m. In addition, as compared with His‐ATX1ΔN, His‐ATX1ΔN‐12m and His‐ATX1ΔN‐m were activated only weakly by His‐SECΔN (Fig 5C). These results indicate that the five conserved amino acids are necessary for SEC activity, and the activation of ATX1 by SEC is dependent on O‐GlcNAcylation of the ATX1 SET domain.

S947 on ATX1 genetically mediates the regulation of flowering in Arabidopsis

To evaluate the effect of S947 O‐GlcNAcylation on the biological function of ATX1 in plants, we generated 35S::ATX1‐FLAG and 35S::ATX1m‐FLAG (in which S947 was replaced with A) transgenic plants in the atx1‐2 mutant background (Fig 6A and B) and assayed their flowering phenotype. The S947A mutation repressed ATX1‐activated FLC transcription (Fig 6B and Appendix Fig S3). The early flowering phenotype of atx1‐2 was rescued by ATX1 but not by ATX1m, and ATX1m transgenic lines flowered earlier than ATX1 overexpression and Col‐0 plants (Fig 6C–E), suggesting that O‐GlcNAcylation of ATX1 at S947 is required for its function in activating FLC to regulate floral transition.

Figure 6. Mutation of S947A inhibits ATX1 function in Arabidopsis .

• A
  Flowering phenotypes of Col‐0, atx1‐2, atx1‐2 35S::ATX1‐FLAG, and atx1‐2 35S::ATX1m FLAG (with Ser947 replaced with alanine) plants under LD conditions. Scale bar: 1 cm. Two independent lines are shown for each transformation.
• B
  qRT‐PCR analysis of ATX1 and FLC transcript levels in Col‐0, atx1‐2, atx1‐2 35S::ATX1‐FLAG, and atx1‐2 35S::ATX1m FLAG plants. The transcript levels were normalized to that of TUBULIN. Data are mean ± s.d. of three independent biological replicates, n = 3.
• C–E
  Flowering‐time phenotype analysis of lines in (A). The percentages of plants with visible buds and the bolting rates were calculated on day 21 and day 24 after plant germination, respectively. Three independent biological repeats were conducted, and data from a total of 71 plants were used for statistical analysis. Data are mean ± s.d., statistical significance (two‐tailed t‐test) with **P < 0.01.
• F
  Working model of SEC activation of ATX1 to regulate FLC transcription by O‐GlcNAc modification.

Together, these results strongly suggest that activation of ATX1 is dependent on its O‐GlcNAcylation in the SET domain, and in particular on S947. In addition, SEC has conserved residues for substrate binding, which is similar to what is seen in human OGT and important for activating ATX1. These findings support that, in the Col‐0 background, SEC acts as a negative regulator to inhibit flowering transition by activating ATX1 and maintaining its stability through O‐GlcNAcylation, leading to increased expression of FLC.

Discussion

SEC mediates epigenetic regulation of flowering by catalyzing O‐GlcNAc modification of HKMT ATX1

The O‐GlcNAc modification of RNA‐binding protein TaGRP2 is known to mediate the vernalization response for flowering in wheat (Xiao et al, 2014). In Arabidopsis Landsberg erecta (Ler) plants, SEC functions as a positive regulator of GA signaling by modifying the DELLA protein RGA with O‐GlcNAc to coordinate development signaling pathways, which may be dependent on the genetic background (Hartweck et al, 2006; Zentella et al, 2016). Here, the sec‐5 mutant showed a wild‐type GA response (Fig EV5). Recently, more than 200 proteins with O‐GlcNAc modification were identified by a large‐scale proteomic analysis, suggesting potential roles of O‐GlcNAc in mediating epigenetic regulation and multiple other cellular processes in Arabidopsis (Xu et al, 2017). In this study, we provide biochemical and genetic evidence that the O‐GlcNAc transferase SEC regulates flowering time via an HKMT ATX1‐associated epigenetic process in Arabidopsis. Loss of function of SEC decreased the level of H3K4me3 in the FLC chromatin, leading to repressed FLC transcription and an early flowering phenotype. OGT associates mostly with the promoter, transcription start site, and gene body regions to regulate H3K4me3 accumulation (Deplus et al, 2013; Vella et al, 2013), indicating that O‐GlcNAc has a role in regulating histone‐related gene expression.

Figure EV5. Response to paclobutrazol and GA3 treatment of Col‐0 and sec‐5 plants.

• A
  The sec‐5 mutant shows a wild‐type response to paclobutrazol (PAC).
• B
  Germination rate analysis of Col‐0, spy‐3, and sec‐5 seeds under PAC treatment. Experiments were repeated three times. Errors bars are s.d. Two asterisks (**) indicate significant differences between Col‐0 and spy‐3 plants (P < 0.01); P‐value was determined by two‐tailed t‐test.
• C, D
  Comparison of plant height between Col‐0 and sec‐5 plants. Experiments were repeated three times. Error bars are s.d. More than 20 plants were analyzed. n.s., not significant by two‐tailed t‐test. Scale bar: 1 cm.
• E
  Comparison of hypocotyl length of Col‐0 and sec‐5 plants. Plants were grown under SD conditions. Scale bar: 0.5 cm.
• F
  Statistical analysis of hypocotyl length of plants in (E). Hypocotyl length was measured after plants were grown under SD conditions for 5 days, and more than 20 plants were calculated for each line. Experiments were repeated three times. Error bars are s.d., n.s., not significant by two‐tailed t‐test.
• G
  Comparison of flowering phenotype of Col‐0 and sec‐5 plants with or without GA3 treatment.
• H
  Calculation of bolting rate of Col‐0 and sec‐5 plants at 20 days after plant germination. Experiments were repeated three times, and more than 20 plants were calculated for each line. Error bars are s.d. Two asterisks (**) indicate a significant difference between Col‐0 and sec‐5 plants (P < 0.01); P‐value was determined by two‐tailed t‐test.

In addition to down‐regulation of H3K4me3, loss of function of SEC also resulted in an increase in H3K27me3 levels in FLC chromatin (Fig 2D). The PRC2 complex maintains a state of transcriptional repression through deposition of H3K27me3 at specific chromatin region, which is conserved in animals and plants (Cao et al, 2002; Schubert et al, 2005; Whitcomb et al, 2007; Xiao et al, 2016). In Drosophila, the location of O‐GlcNAc on chromosomes is coincident with Polycomb group (PcG) response elements (PREs; Gambetta et al, 2009; Sinclair et al, 2009), and PREs and TREs (trithorax response elements) may represent the same regions for regulation of target gene expression (Ringrose & Paro, 2004). The active chromatin mark H3K4me3 specifically inhibits PRC2‐mediated H3K27me3 in mammals and flies, and TrxG and PcG proteins antagonistically regulate target gene expression (Papp & Muller, 2006; Schmitges et al, 2011). In Arabidopsis, PRE‐like functional components have been found to regulate H3K27me3 (Berger et al, 2011; Xiao et al, 2017). During vernalization, the PRC2 complex plays a role in triggering FLC silencing in Arabidopsis (De Lucia et al, 2008). In addition, ATX1 has been suggested to counteract PcG‐mediated gene silencing in Arabidopsis (Pien & Grossniklaus, 2007). Our data suggest that the increased H3K27me3 observed at FLC chromatin in sec‐5 plants might be a result of the down‐regulation of H3K4me3. These findings may shed light on O‐GlcNAc‐mediated epigenetic regulation during plant development.

O‐GlcNAc modification mediates ATX1 activity and stability regulation in plants

There are five ATX genes in the Arabidopsis genome, of which ATX1 and ATX2 form one subgroup and ATX3–ATX5 another (Ng et al, 2007). As the plant orthologs of MLL proteins, ATX1 and ATX2 have highly similar structures, but they show different temporal and special expression patterns. Despite the high degree of similarity, ATX1 catalyzes H3K4me3 of FLC chromatin specifically (Pien et al, 2008), but ATX2 mediates H4K4me2 of target genes (Saleh et al, 2008). The triple mutant atx3‐1 atx4‐1 atx5‐1 exhibits dwarf and small rosette leaf phenotypes, but no alteration of flowering time (Chen et al, 2017). In contrast to its significant reduction in H3K4me3 in FLC chromatin, loss of SEC resulted in slight up‐regulation of H3K4me2 in region P3 of FLC (Fig 2B), indicating that ATX1 is not required for the deposition of H3K4me2. Non‐reduced H3K4me2 has been reported in silenced FLC during the vernalization response in Arabidopsis (Bastow et al, 2004). The slight up‐regulation of H3K4me2 levels in sec‐5 mutants suggested that not ATX2 but rather ATX1 activity is influenced by sec mutation and that SEC regulates flowering transition by affecting O‐GlcNAc‐dependent ATX1 activity.

O‐GlcNAc modification plays a role in regulating protein stability and activity (Yi et al, 2012; Ruan et al, 2013). Tyr1015 is reported to be critical for ATX1 activity, but the regulatory mechanism underlying this effect is unknown (Ding et al, 2012). In this study, we presented evidence that SEC O‐GlcNAcylates ATX1 to activate its histone methyltransferase activity and S947 is critical for O‐GlcNAc modification‐mediated activity regulation of ATX1. According to our results, the S947A mutation did not completely inhibit ATX1 activity (Fig 6B), indicating that there may be other mechanisms for regulation of ATX1 based on HKMT activity. Our results here suggest that O‐GlcNAcylation acts as a pattern of HKMT activity and stability regulation in plants. It will be of great interest to define whether other TrxG or PcG proteins can also be regulated by O‐GlcNAc modification in further studies.

O‐GlcNAcylation and COMPASS‐like complexes

H3K4 methylation is catalyzed by COMPASS (Complex Proteins Associated with Set1) complexes to activate target genes transcription in yeast, human, and plants. The core components of COMPASS complexes are conserved and include Ash2, RbBP5, and WDR5 (Miller et al, 2001; Shilatifard, 2008; Ding et al, 2012; Xiao et al, 2016). In Arabidopsis, an Ash2R‐containing COMPASS‐like complex mediates H3K4me3 in FLC to regulate flowering time (Jiang et al, 2011). ATX1 acts as one of the components of a COMPASS‐like complex to participate in FLC activation (Kim & Sung, 2012; Fromm & Avramova, 2014). ATX1 is required for recruitment of AtCOMPASS‐like to promoters, but the assembly of the transcription machinery at target promoters is not influenced by the activity of ATX1, indicating that an ATX1/AtCOMPASS‐like generated H3K4me3 mark is required for transcription elongation but not initiation (Ding et al, 2012).

ATX1‐catalyzed H3K4me3 of FLC chromatin is required for FLC transcription activation. Our data indicate that SEC functions as a negative regulator of flowering to activate ATX1 by O‐GlcNAc modification, suggesting that O‐GlcNAc modification‐mediated ATX1 activity plays a role in regulating ATX1/AtCOMPASS‐like complex function (Fig 6F). In summary, our findings not only uncover a previously unknown mechanism regulating the activity of the TrxG HKMT ATX1, but also establish a distinct epigenetic role of O‐GlcNAc signaling in plants. OGT is necessary for the formation of the SET1/COMPASS complex. One of the components of the SET1/COMPASS complex, host cell factor 1 (HOS1), is O‐GlcNAcylated by OGT to regulate SET1/COMPASS integrity, and OGT activity helps promote SET1 binding to chromatin (Deplus et al, 2013). The present study provides novel insight into the regulatory mechanism of the ATX1/AtCOMPASS‐like complex in plants. It will be intriguing to explore the role of O‐GlcNAc modification in regulating HKMT substrate recognition and complex assembly in plants.

Materials and Methods

Plant materials and growth conditions

Arabidopsis thaliana ecotypes Columbia‐0 (Col‐0) and C24 were used as wild type and as background for the transgenic lines, respectively. T‐DNA insertion mutants of sec‐4 (SALK_106339) and sec‐5 (SALK_034290) were obtained from the SALK T‐DNA collection (Alonso et al, 2003).

Seeds were sterilized and grown on half‐strength Murashige and Skoog medium (1/2 MS) plates for 3 days at 4°C and then transferred to the greenhouse for germination at 22°C with the photoperiods of 16‐h light/8‐h dark (long day, LD) or 8‐h light/16‐h dark (short day, SD). Transgenic seeds were grown on 1/2 MS containing 20 mg/l hygromycin as screening pressure to obtain transgenic lines. Flowering time was calculated as rosette leaf numbers at bolting.

In vitro O‐GlcNAcylation and methyltransferase assay

1 μg of recombinant expressed His‐SECΔN (aa 592–952, containing the OGT domain of SEC) was incubated with 1 μg His‐ATX1ΔN (aa 801–1,062, containing the SET domain) and 50 μM UDP‐N‐acetylglucosamine in 25 μl of reaction system for 1 h at 37°C. Immunoblotting analysis was conducted to detect O‐GlcNAc modification of proteins with antibody CTD110.6.

For ATX1 histone methyltransferase enzyme activity assay, purified His‐ATX1ΔN was incubated with recombinant expressed H3 (Millipore) and unlabeled S‐adenosyl‐L‐methionine (SAM, Sigma) in methyltransferase buffer. After the HMTase assay, the products were analyzed by immunoblot analysis with the indicated antibodies.

Plasmid construction and plant transformation

To generate a complementation construct for the sec‐5 mutant, a 1,798‐bp genomic sequence upstream of the SEC ATG and the 2,934‐bp cDNA fragment of SEC were amplified by PCR and cloned into the vector pFGC5941. To generate the p35S::SEC construct, cDNA fragments of SEC were amplified by RT–PCR and cloned into the pSN1301 vector. The p1307‐SEC‐FLAG construct was generated by fusion of amplified SEC with FLAG in the p1307‐cFLAG plasmid to express the SEC‐FLAG fusion protein in plants. To express an ATX1‐HA fusion protein in plants, the cDNA fragment of ATX1 was amplified and inserted into the SacI‐SalI site of the pCsVMV‐HA3‐N‐1300 plasmid. The single‐site mutation of the ATX1 cDNA was generated by overlap PCR technique. The mutated products were confirmed by sequencing and cloned to p1307‐cFLAG plasmid. The primers used for plasmid construction are listed in Appendix Table S3.

The Arabidopsis plants were transformed by the floral‐dip method (Clough & Bent, 1998) using the GV3101 Agrobacterium strain, and positive plants were screened on hygromycin B medium. More than ten independent transformants with a single T‐DNA insertion were obtained for each construct. The T3 homozygotes were used for subsequent analysis.

Reverse transcription PCR and quantitative real‐time reverse transcription PCR

Total RNAs were extracted with TRIzol reagent (Invitrogen), and 1 μg of total RNA was used for reverse transcription reaction with SuperScript III reverse transcriptase (Invitrogen). Real‐time quantitative reverse transcription PCR (qRT–PCR) analyses were performed using SYBR green mixture (TOYOBO) as described previously (Xiao et al, 2014), and UBIQUITIN was used as internal control. Three independent biological replicates were performed for each qRT–PCR analysis. The primers used for qRT–PCR are listed in Appendix Table S3.

ChIP

Chromatin immunoprecipitation (ChIP) experiments were performed according to methods described previously (Bowler et al, 2004). Antibodies specific to H3K4me2, H3K4me3, H3K36me3, and H3K27me3 (Millipore) were used to immunoprecipitate FLC chromatin. The relative abundances of histone modifications were normalized to input DNA. Primers are listed in Appendix Table S3.

Yeast two‐hybrid assay

For yeast two‐hybrid assay, the open reading frame of ATX1 was amplified and inserted into NcoI and SmaI sites of the bait vector pGBKT7 (BD). The open reading frame of SEC was cloned into ClaI and SacI sites of the prey vector pGADT7, and the cDNA fragment of SEC‐N (cDNA only encoding TPR domains at N‐terminus) was cloned into NdeI and EcoRI sites of pGADT7, respectively. Primers used for plasmid construction are shown in Appendix Table S3. To confirm protein interactions in yeast cells, BD‐ATX1 was cotransformed with AD‐SEC and AD‐SEC‐N into Saccharomyces cerevisiae strain AH109 (Clontech) according to the manufacturer's instructions. Transformants first were grown on SD/‐Leu/‐Trp (SD/‐2) medium, and then, positive clones were screened on SD/‐Leu/‐Trp/‐His/‐Ade (SD/4) medium. pGBKT7‐53 and pGADT7‐T vectors were cotransformed as positive controls, and pGBKT7‐Lam and pGADT7‐T were cotransformed as a negative control.

Protein expression, purification, and immunoblot analyses

The cDNA fragments of SEC and ATX1 were cloned into the pET28a vector following codon optimization (GenScript) and used to express peptides corresponding to 361 amino acids of the SEC C‐terminus and 262 amino acids of the ATX1 C‐terminus, respectively. The same method was used to generate constructs expressing mutated peptides of ATX1‐12m (a total of 12 serine and threonine residues were replaced with alanine), ATX1‐m (S947 was replaced with alanine), and SEC‐5m (Phe540, His541, His604, Gln776, and Lys779 were all replaced with alanine). The recombined plasmids were transformed into the E. coli BL21 (DE3) line to generate fusion proteins with His tags. The His‐SECΔN, His‐ATX1ΔN, His‐ATX1‐12m, His‐ATX1‐m, and His‐SEC5m fusion proteins were induced and purified by Protein Service (GenScript).

For immunoblot assays, proteins were separated on SDS–PAGE and then electrotransferred to PVDF membranes (Millipore). The membranes were blocked at room temperature for 2 h with TBST buffer (25 mM Tris, 140 mM NaCl, 3 mM KCl, and 0.1% Tween‐20, pH 7.4). Membranes were incubated overnight at 4°C with appropriate antibodies and washed with TBST for 3 × 10 min. Then, membranes were incubated with secondary antibody conjugated to horseradish peroxidase (Sigma). Immunoreactive bands were detected by using the enhanced chemiluminescence (ECL) or super ExPlus system.

In vitro O‐GlcNAcylation assay

1 μg of recombinant expressed His‐SECΔN was incubated with 1 μg His‐ATX1ΔN and 50 μM UDP‐N‐acetylglucosamine in 25 μl of reaction system for 1 h at 37°C. The reaction buffer contained 50 mM Tris–HCl, pH 7.8, 12.5 mM MgCl₂, and 1 mM DTT, pH 7.5. Samples were denatured at 95°C for 10 min in 5× loading buffer (100 mM Tris–HCl, pH 6.8, 200 mM DTT, 4% SDS, 20% glycerol, and 0.2% bromophenol blue) and electrophoresed by SDS–PAGE. Immunoblot analysis was conducted to detect O‐GlcNAc modification of proteins with antibody CTD110.6 specific to O‐GlcNAc sites.

β‐Elimination assay

The O‐GlcNAc linkage is easily hydrolyzed in alkaline solution with mild conditions. A β‐elimination assay was used to confirm the presence of the O‐GlcNAc modification on proteins as previously described (Duk et al, 1997). The O‐GlcNAcylated proteins catalyzed by SECΔN were electrotransferred to PVDF membrane (Millipore), and then, membranes were treated with 55 mM NaOH for 16 h at 40°C. The target proteins were redetected with antibody CTD110.6 by immunoblot assay.

Histone methyltransferase assay

For ATX1 histone methyltransferase enzyme activity assay, 1 μg purified His‐ATX1ΔN protein was incubated with 2 μg of recombinant expressed H3 (Millipore) and 1 mM S‐(5′‐adenosyl)‐L‐methionine iodide (Sigma) in methyltransferase buffer (50 mM Tris, pH 9.0, 1 mM PMSF and 0.5 mM DTT) for 1 h at 30°C. The products were analyzed by immunoblot with anti‐H3K4me3 antibody (Millipore). The quantitative analysis of ATX1 histone methyltransferase enzyme activity was based on the amount of methylated H3K4 converted by ATX1 with EpiQuik Histone Methyltransferase Activity/Inhibition Assay Kit (EpiGentek).

Identification of O‐GlcNAc sites by higher‐energy collisional dissociation (HCD)–mass spectrometry (MS) and electron transfer dissociation (ETD)‐MS

Recombinant His‐SECΔN and His‐ATX1ΔN proteins expressed in E. coli strain BL21 (DE3) were purified (GenScript) and incubated for SEC O‐GlcNAcylation enzyme activity reaction. Samples were subsequently separated by SDS–PAGE and digested in‐gel with trypsin. Peptides were dried to completion and enriched with O‐GlcNAc antibody (PTM‐Biolabs Co., Ltd, PTM20160616C0) and analyzed by tandem mass spectrometry (MS/MS) in Q Exactive (Thermo).

Author contributions

LX designed experiments and performed immunoblotting, phenotyping, and data analysis. YL performed ChIP, immunoblotting, qRT–PCR, and plant material preparation. SX performed qRT–PCR and transgenic analyses. JX provided useful critiques of the manuscript. BW performed plasmid construction. HD prepared plant materials. ZL analyzed mass‐spectrum data. YX analyzed data. KC designed experiments, KC and LX wrote the manuscript. All authors commented on the manuscript.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Appendix

Expanded View Figures PDF

Source Data for Expanded View

Review Process File

Source Data for Figure 1

Source Data for Figure 3

Source Data for Figure 4

Source Data for Figure 5

The EMBO Journal (2018) 37: e98115