Mutation of histone H3 serine 28 to alanine influences H3K27me3-mediated gene silencing in Arabidopsis thaliana

Abstract

Histone modifications are essential for chromatin activity and play an important role in many biological processes. Trimethylation of histone H3K27 (H3K27me3) is a repressive modification established by Polycomb Repressive Complex 2 (PRC2). Although the presence of the histone H3 serine 28 phosphorylation (H3S28ph) modification at adjacent amino acid residues has both positive and negative effects on Polycomb silencing in mammals, little is known about the effect of H3S28ph on H3K27me3-mediated gene silencing in plants. In this study, we show that mutating H3S28A in Arabidopsis (Arabidopsis thaliana) causes a dominant-negative effect that leads to an early-flowering phenotype by promoting the expression of flowering-promoting genes independently of abnormal cell division. While H3S28ph levels decreased due to the H3S28A mutation, H3K27me3 levels at the same loci did not increase. Moreover, we observed decreased H3K27me3 levels at some known PRC2 target genes in H3.3S28A transgenic lines, rather than the expected enhanced H3K27me3-mediated silencing. In line with the reduced H3K27me3 levels, the expression of the PRC2 catalytic subunits CURLY LEAF and SWINGER decreased. Taken together, these data demonstrate that H3.3S28 is required for PRC2-dependent H3K27me3-mediated silencing in Arabidopsis, suggesting that H3S28 has a noncanonical function in H3K27me3-mediated gene silencing.

Introduction

In eukaryotes, chromatin is a highly condensed structure composed of DNA and histones (Luger et al., 1997). Chromatin structure and function are regulated by chromatin remodeling, DNA methylation, histone modifications, and other factors. Histone modifications such as methylation, acetylation, ubiquitination, and phosphorylation are covalent additions at histone N-terminal tails (Jenuwein and Allis, 2001). These modifications are involved in many nuclear activities, such as replication, chromatin assembly, and transcriptional regulation (Marmorstein, 2001).

Histone H3 lysine 27 trimethylation (H3K27me3) is a repressive histone mark essential for chromatin structure and gene silencing (Mozgova et al., 2015; Xiao and Wagner, 2015; Wiles and Selker, 2017). The deposition of H3K27me3 is controlled by the evolutionarily conserved Polycomb Repressive Complex 2 (PRC2), an H3K27 methyltransferase complex (Pien and Grossniklaus, 2007; Lu et al., 2008; Margueron and Reinberg, 2011; Xiao et al., 2016). In Arabidopsis,CURLY LEAF (CLF), SWINGER (SWN), and MEDEA (MEA) are SET domain-containing protein subunits of PRC2 that catalyze H3K27me3 deposition (Goodrich et al., 1997; Grossniklaus et al., 1998; Liu et al., 2010). The local chromatin environment is also crucial for H3K27me3 deposition. In Arabidopsis and Marchantia polymorpha, H2A monoubiquitination (H2Aub) is required for H3K27me3 deposition (Kralemann et al., 2020; Liu et al., 2021; Yin et al., 2021). Active histone marks, such as H3K4me3 and H3K36me2/3, inhibit the function of PRC2 in both mammals and plants (Schmitges et al., 2011). Moreover, the level of an adjacent histone mark, histone H3 serine 28 phosphorylation (H3S28ph), can also influence PRC2 function.

In mammals, H3S28 is phosphorylated mainly by Mitogen- and Stress-activated Kinase 1 (MSK1) and MSK2, and by mitotic Aurora B kinase specifically during mitosis (Giet and Glover, 2001; Goto et al., 2002; Dunn and Davie, 2005; Dyson et al., 2005). H3S28ph has both positive and negative effects on Polycomb silencing. On the one hand, H3S28ph counteracts mammalian Polycomb-mediated silencing in response to biological stimuli (Gehani et al., 2010; Lau and Cheung, 2011). On the other hand, H3S28A, a point mutation at H3S28, reduces H3K27me3 levels and compromises Polycomb silencing in Drosophila melanogaster (Yung et al., 2015). In Arabidopsis, systematic profiling of histone readers shows that H3K27me3 increases the interaction between H3S28ph and the phosphorylation reader GRF2 (Zhao et al., 2018), indicating a potential functional cooperation between H3K27me3 and H3S28ph. In contrast, H3S28ph blocks the binding of PWWP domain-containing protein PWO1, a recently discovered H3K27me3 reader, to histones in vitro (Hohenstatt et al., 2018), suggesting a negative role of H3S28ph in the downstream part of H3K27me3-mediated silencing. However, the effect of H3S28ph on H3K27me3-mediated silencing in Arabidopsis has been unclear. Therefore, further genetic and genome-wide analysis of the relationship between H3S28ph and H3K27me3 are required to understand the role of H3S28ph in H3K27me3-mediated silencing.

Here, we show that the S28A mutation of histone H3 results in an early-flowering phenotype without interfering with cell division in Arabidopsis. We revealed that decreased H3S28ph derived from the H3S28A mutation is not sufficient to induce extra H3K27me3 at either transposable elements (TEs) or genes, although H3S28ph is considered antagonistic to H3K27me3. Surprisingly, we found that the H3.3S28A mutation could decrease H3K27me3 levels, potentially by regulating the expression of the PRC2 components CLF and SWN. Although histone phosphorylation is well known to inhibit the methylation of adjacent histones (Gehani et al., 2010; Lau and Cheung, 2011), H3S28 is required for H3K27me3-mediated silencing at some loci in Drosophila (Yung et al., 2015). Thus, our results suggest an unexpected but probably conserved function of H3S28 in H3K27me3-mediated silencing in Arabidopsis.

Results

The H3S28A mutation causes an early-flowering phenotype

To explore the effect of H3S28ph on adjacent H3K27me3 deposition and transcriptional regulation in Arabidopsis, we generated Arabidopsis plants expressing variants of histone H3.1 and H3.3. Since HTR13 (H3.1) and HTR5 (H3.3) have been used in previous publications (Sanders et al., 2017; Lu et al., 2018), we decided to use them as well in our experiments. We generated Arabidopsis plants expressing HTR13 or HTR5 with serine-to-alanine substitutions at S28 of each protein (hereafter referred to as H3.1S28A and H3.3S28A, respectively). H3.1 and H3.3 fragments were driven by their native promoters, which are cell cycle dependently or ubiquitously expressed during development. As controls, we also generated Arabidopsis expressing intact HTR13 or HTR5 under the control of their native promoters (hereafter referred to as H3.1 and H3.3, respectively) (Figure 1A). In around 40 T₁ transgenic lines, H3.1 displayed no phenotypic differences from wild-type (WT) Columbia-0 (Col-0). In comparison, 36 out of 45 H3.3 T₁ lines were late flowering compared to the WT, in line with the early-flowering phenotype of h3.3kd mutants (Zhao et al., 2021). In contrast to H3.1 and H3.3, around 50% of H3.1S28A and H3.3S28A lines showed an early-flowering phenotype (Supplemental Figure S1A).

Figure 1.

The phenotype of H3S28A point mutation plants. A, Schematic diagram of S28A under the native promoter of H3.1 (HTR13) gene, left and H3.3 (HTR5) gene, right. B, Four-week-old Col-0 and transgenic plants expressing H3.1 and H3.1S28A under LD condition, two independent lines are shown. C, Five-week-old Col-0 and transgenic plants expressing H3.3 and H3.3S28A under LD condition, two independent lines are shown. D, Flowering time of Col-0, H3.1, and H3.1S28A plants expressed as the rosette leaves number under LD conditions. n = 20. E, Flowering time of Col-0, H3.3, and H3.3S28A plants expressed as the rosette leaves number under LD conditions. n = 20. Symbols represent median values. Different letters indicate significant differences among means according to one-way ANOVA and post-hoc Tukey’s test at P < 0.05.

The early-flowering phenotype was further confirmed in T₃ generation plants. Overall, two independent homozygous lines from the T₃ generations of H3.1, H3.1S28A, and H3.3 showed phenotypes similar to those of the T₁ plants (Figure 1, B and C). We failed to obtain homozygous H3.3S28A lines due to infertility (Supplemental Figure S1, B and C). Both of male and female fertility were reduced in the reciprocal cross (Supplemental Figure S2). Since heterozygous H3.3S28A also showed an early-flowering phenotype compared to H3.3 (Figure 1C), we used heterozygous H3.3S28A for further study. Although both H3.1S28A and H3.3S28A were early flowering (Figure 1, D and E), we observed dwarfing and smaller leaves in H3.3S28A but not in H3.1S28A, suggesting that the S28A mutation in histone H3.3 has more severe dominant-negative effects on plant development. The aforementioned T₃ plants were utilized for further analysis. Thus, the H3S28A mutation displays an early flowering phenotype.

The H3S28A mutation influences plant development independently of abnormal cell division

To explore if decreased H3S28ph confers the aforementioned developmental defects, we examined H3S28ph levels in H3.1S28A and H3.3S28A. Given that H3S28ph strongly accumulates at centromeric regions during metaphase and anaphase of cell division (Gernand et al., 2003), we investigated H3S28ph levels at metaphase via immunostaining, by normalizing to a cell cycle-dependent phosphorylation histone H3 threonine 3 (H3T3ph) (Caperta et al., 2008). H3.1 and H3.3 presented similar H3S28ph signal intensity as the WT (Figure 2, A, B, D, and F). In contrast, H3S28ph signal intensity was significantly lower in H3.3S28A compared to H3.3 (P < 0.05, one-way ANOVA [AQ]; Figure 2, D–F). Unlike H3.3S28A, H3.1S28A exhibited a subtle decrease in H3S28ph levels compared to H3.1 (Figure 2, B, C, and F). Therefore, the dominant effect of H3.1S28A is weaker than that of H3.3S28A, in line with the weaker developmental defects observed in H3.1S28A (Figure 1, B and D).

Figure 2.

Fluorescence of H3S28ph on metaphase chromosomes. A–E, Confocal images of H3S28ph and H3T3ph on metaphase chromosomes in Col-0, H3.1, H3.1S28A, H3.3, and H3.3S28A, respectively. F, Quantifications of the ratio of H3S28ph/H3T3ph level on metaphase chromosomes. The relative intensity of n = 15 metaphase chromosomes was analyzed using ImageJ. The data are presented as the means ± SE, and individual data points as overlays. G, Ploidy analysis of H3.3 and H3.3S28A. Pictures represent five independent plants in each line. 2n and 4n Col-0 were used as the ploidy control. The numbers below the peaks indicate ploidy levels of the nuclei. Marked area represents diploidy peak. Different letters indicate significant differences among means according to one-way ANOVA and post-hoc Tukey’s test at P < 0.05. Scale bars, 5 μm.

We next examined the effect of decreased H3S28ph levels on cell division. Given that abnormal mitosis is usually associated with aneuploidy in somatic cells, we examined the ploidy levels of H3S28A and H3 lines by flow cytometry and found that all lines were diploid (Figure 2G; Supplemental Figure S3A). In addition to mitosis, we also investigated the tetrad products of male meiosis. The abnormal chromosome segregation usually results in dyads, triads, or polyads. However, we did not observe any difference in H3S28A plants compared to Col-0 (Supplemental Figure S3B). Moreover, abnormal meiosis may also result in aneuploidy pollen or aborted pollen. In line with the normal tetrad formation in H3S28A, pollen grains in H3S28A were as expected, like in Col-0 (Supplemental Figure S3C). Taken together, these results indicate that the H3S28A mutation affects plant development but cell division is normal.

H3S28A alters the expression of H3K27me3-targeted flowering-promoting genes

To determine if the phenotypic defects observed in H3.1S28A and H3.3S28A were due to transcriptional alterations, we generated transcriptome profiles of 2-week-old H3 and H3S28A seedlings. We identified only 301 differentially expressed genes (DEGs, log2FC > 1 or log2FC < −1, FDR < 0.05) in H3.1S28A compared to H3.1. In contrast, we identified 4,116 DEGs in H3.3S28A compared to H3.3 (Figure 3A;Supplemental Table S1), consistent with the stronger dominant phenotypic effects of H3.3S28A. Although more genes were downregulated in H3.3S28A, we found that upregulated genes (UGs) had greater fold changes than downregulated genes (DGs) in this line. Considering the role of H3S28ph in H3K27me3-mediated gene silencing, we further analyzed the expression of PRC2 targets (defined H3K27me3 targeted genes in leaves with methylation score > 0) (Lafos et al., 2011) in the H3.1S28A and H3.3S28A RNA-seq data sets. The overlap between DEGs in H3.1S28A and PRC2 targets was not very significant (Figure 3B). In contrast, UGs in H3.3S28A significantly overlapped with PRC2 targets: 750 out of 1,734 UGs are known PRC2 targets (Figure 3B) (Lafos et al., 2011). These data suggest that H3.3S28A plays a role in the transcriptional repression of a group of PRC2 targets.

Figure 3.

RNA-seq of H3.1S28A and H3.3S28A. A, Volcano plots of DEGs in H3.1S28A (left) and H3.3S28A (right). The X-axis and Y-axis represent log2FC and the statistical significance as the negative log10 (P-value). The red dots represent genes that were significantly UGs, while the blue dots represent the significantly DGs. Genes that were not significantly changed were shown with gray dots. Two biological replicates were performed. B, Venn diagram showing overlap of UGs and DGs with PRC2 targets in H3.1S28A (left) and H3.3S28A (right). Significance was tested using a hypergeometric test (P  < 0.001). C, Enriched biological processes of significantly UGs and DGs in H3.1S28A compared to H3.1. D, Enriched biological processes of significantly UGs and DGs in H3.3S28A compared to H3.3. The X-axis represents negative log10 (P-value). E, Heat map showing the relative mRNA expression levels of indicated genes in H3.1S28A and H3.3S28A, log2Fold Change was used. F, Relative expression level of the selected genes (arrows in E) by RT–qPCR from 2-week-old H3.1, H3.1S28A, H3.3, and H3.3S28A seedlings. Two biological replicates were performed. Data are means ± sd.

To examine the molecular pathways affected by the H3.1S28A and H3.3S28A mutations, we performed a gene ontology enrichment analysis on all DEGs. DEGs in H3.1S28A were enriched in stress- and hormone-related pathways (Figure 3C). In H3.3S28A, consistent with the early flowering phenotype, DEGs were enriched in flower development and transcriptional regulation (Figure 3D). Among the flower development-related genes, we found that PRC2 targeted flowering-promoting genes such as AGAMOUS (AG), APETALA1 (AP1), FLOWERING LOCUS T (FT), SEPALLATA3 (SEP3), and AG-LIKE17 (AGL17) were upregulated in H3S28A compared to H3 (Figure 3, E and F). Although the flowering-repressive genes FLOWERING LOCUS C and MADS AFFECTING FLOWERING5 also showed increased expression in H3S28A, the difference was subtle (Figure 3E). Thus, the early-flowering phenotype observed in H3S28A was likely caused by the upregulation of flowering-promoting genes.

H3S28A mutations are not sufficient to induce H3K27me3 deposition

To investigate whether the transcriptional alterations in H3S28A lines were caused by reduced H3S28ph and differential H3K27me3, we generated genome profiles of H3S28ph and H3K27me3 for H3S28A, along with WT Col-0 and H3 as controls. We first examined the distribution of H3S28ph and H3K27me3 on chromosomes. In contrast to strong signals of H3K27me3 at chromosome arms, H3S28ph is enriched at centromeres (Supplemental Figure S4). We applied the Limma package (Ritchie et al., 2015) to directly identify TEs and genes showing reduced H3S28ph levels in H3.1S28A and H3.3S28A compared to H3.1 and H3.3. We identified 562 and 644 TEs with reduced H3S28ph in H3.1S28A and H3.3S28A compared to H3.1 and H3.3, respectively (Figure 4, A and B;Supplemental Table S2). The TEs showing decreased H3S28ph were largely different in H3.1S28A and H3.3S28A, with only 23 TEs shared in common (Figure 4C). The TEs showing reduced H3S28ph levels in H3.1S28A compared to H3.1 presented substantially increased levels of H3S28ph in H3.1 compared to the WT (Figure 4A). Similarly, TEs showing reduced H3S28ph levels in H3.3S28A also exhibited increased H3S28ph levels in H3.3 compared to the WT (Figure 4B). Therefore, TEs with decreased H3S28ph levels in H3.1S28A or H3.3S28A were sensitive to increased dosage from transgenic histones H3.1 or H3.3. Such TEs were usually short, often <1,000 bp (Figure 4, D and E). These TEs were randomly distributed in pericentromeric and euchromatic regions along chromosomes (Figure 4F).

Figure 4.

Characterization of decreased H3S28ph at TEs. A, Box plot showing median values of H3S28ph in Col-0, H3.1 and H3.1S28A, H3.3, and H3.3S28A over TEs with reduced H3S28ph level in H3.1S28A mutant (n = 562). B, Box plot showing median values of H3S28ph in Col-0, H3.1 and H3.1S28A, H3.3, and H3.3S28A over TEs with reduced H3S28ph level in H3.3S28A mutant (n = 644). Two biological replicates were performed. C, Venn diagram showing overlap of TEs with reduced H3S28ph between H3.1S28A and H3.3S28A. D, Genome browser views of two selected TEs showing decreased H3S28ph level in both H3.1S28A and H3.3S28A mutants. E, TE length distribution of all TEs and TEs with reduced H3S28ph in H3.1S28A (n = 562) and H3.3S28A (n = 644). F, Percentage of pericentromeric and euchromatic TEs in all TEs and TEs with reduced H3S28ph in H3.1S28A and H3.3S28A. G, Box plot showing median values of H3K27me3 at TEs lost H3S28ph (with reduced H3S28ph, n = 562) and all TEs in H3.1 and H3.1S28A. H, Box plot showing median values of H3K27me3 at TEs lost H3S28ph (with reduced H3S28ph, n = 644) and all TEs in H3.3 and H3.3S28A. NS: P > 0.05, *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001. (Kolmogorov–Smirnov test).

In addition to reduced H3S28ph levels at TEs, we also examined whether reduction of H3S28ph is sufficient to induce ectopic H3K27me3 at TEs. In H3.1S28A, the group of TEs with reduced H3S28ph did not acquire ectopic H3K27me3. In contrast, these TEs showed significantly reduced levels of H3K27me3 (Figure 4G;Supplemental Table S2). However, given that H3K27me3 levels at TEs are very low, the reductions observed at such loci were not informative. Nonetheless, H3K27me3 levels did not increase when TEs lost H3S28ph. Similarly, TEs with reduced H3S28ph levels in H3.3S28A did not acquire additional H3K27me3 (Figure 4H;Supplemental Table S2). Hence, reduced H3S28ph in H3.1S28A or H3.3S28A is not sufficient to induce H3K27me3 deposition at TEs.

We further explored whether reduced H3S28ph in H3.1S28A and H3.3S28A influences H3K27me3 levels at genes. To address this question, we identified 757 and 478 genes showing substantially decreased H3S28ph levels in H3.1S28A and H3.3S28A, respectively (Figure 5, A and B;Supplemental Table S3). Similar to the TEs with reduced H3S28ph, genes with reduced H3S28ph in H3.1S28A and H3.3S28A also presented enhanced H3S28ph levels in H3.1 and H3.3 compared to the WT (Figure 5, A and B). Hence, genes showing decreased H3S28ph are also sensitive to H3.1 or H3.3 dosage. Moreover, the genes showing decreased H3S28ph in H3.1S28A and H3.3S28A were quite different (Figure 5C) and were typically small genes shorter than 1,000 bp (Figure 5, D and E); these observations are similar to those for TEs. This indicates that small genes and TEs are more sensitive to the dominant effect of H3S28A. It seems plausible that such loci are easily fully replaced by the mutated H3.1 or H3.3. We then analyzed H3K27me3 levels at genes showing decreased H3S28ph in H3.1S28A and H3.3S28A, finding that a decrease in H3S28ph in H3.1S28A and H3.3S28A did not induce H3K27me3 deposition at genes, either (Figure 5, F and G; Supplemental Table S3). Moreover, we also identified 431 and 325 upstream regions (−2,000 bp to start codon) losing H3S28ph in H3.1S28A and H3.3S28A, and 532 and 388 downstream regions (+2,000 bp to stop codon). Similarly, we did not observe increased H3K27me3 levels at upstream and downstream regions losing H3S28ph (Supplemental Figure S5, A–D; Supplemental Table S4). Therefore, even though the H3S28ph modification is associated with transcriptional activation and is considered antagonistic to H3K27me3 (Sawicka et al., 2014), reducing H3S28ph in H3.1S28A and H3.3S28A is not sufficient to induce enhanced H3K27me3 deposition at the subgroup of short TEs or genes identified in this study in Arabidopsis.

Figure 5.

Characterization of decreased H3S28ph at genes. A, Box plot (left) and Metagene plot (right) showing median values of H3S28ph in Col-0, H3.1 and H3.1S28A, H3.3, and H3.3S28A over genes with reduced H3S28ph level in H3.1S28A mutant (n = 757). B, Box plot (left) and Metagene plot (right) showing median values of H3S28ph in Col-0, H3.1 and H3.1S28A, H3.3, and H3.3S28A over genes with reduced H3S28ph level in H3.3S28A mutant (n = 478). Two biological replicates were performed. Transcription start sites (TSSs), transcription end sites (TES). C, Venn diagram showing overlap of genes with reduced H3S28ph between H3.1S28A and H3.3S28A. D, Genome browser views of two selected genes showing decreased H3S28ph level in both H3.1S28A and H3.3S28A mutants. E, Gene length distribution of all genes and genes with reduced H3S28ph in H3.1S28A (n = 757) and H3.3S28A (n = 478). F, Box plot showing median values of H3S28ph (left) and H3K27me3 (right) at genes lost H3S28ph (with reduced H3S28ph, n = 757) and PRC2 targets (n = 6909) in H3.1 and H3.1S28A. G, Box plot showing median values of H3S28ph (left) and H3K27me3 (right) at genes lost H3S28ph (with reduced H3S28ph, n = 478) and PRC2 targets (n = 6909) in H3.3 and H3.3S28A. N.S. P > 0.05, *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001. (Kolmogorov–Smirnov test).

H3.3S28 is required for H3K27me3 deposition

Although the reduction of H3S28ph in H3.1S28A and H3.2S28A did not influence H3K27me3 levels as expected, H3K27me3 levels at known PRC2 targets decreased in H3.3S28A (Figure 5G). Moreover, UGs in H3.3S28A showed significant overlap with known PRC2 targets (Figure 3B). Both these points indicated that the H3.3S28A mutation reduced the H3K27me3 levels of a group of genes, resulting in their transcriptional activation. To validate this finding, we analyzed the H3K27me3 levels at UGs in H3.3S28A. Indeed, H3K27me3 levels at these UGs were lower in H3.3S28A than in H3.3 (Figure 6A;Supplemental Figure S6; Supplemental Table S5), in line with their greater expression. Unlike the H3K27me3 levels, H3S28ph levels at these genes were similar between H3.3S28A and H3.3 (Figure 6B;Supplemental Figure S6; Supplemental Table S5). Therefore, H3S28A indirectly influenced H3K27me3 deposition at these genes. PRC2 is known to be the core complex depositing H3K27me3 (Pien and Grossniklaus, 2007; Margueron and Reinberg, 2011). Therefore, we asked whether the H3.3S28A mutation might influence the expression of PRC2 components, leading to reduced H3K27me3 levels. Indeed, we found that CLF and SWN expression was substantially lower in H3.3S28A compared to that in H3.3 (Figure 6, C and D). In contrast, the expression levels of CLF and SWN, as well as their H3K27me3 levels, were similar between H3.1S28A and H3.1 (Figures 5F and 6, C and D; Supplemental Figures S6 and S7; Supplemental Table S5). Hence, H3.3S28A influences H3K27me3 deposition and transcriptional repression by regulating the expression of the PRC2 components CLF and SWN.

Figure 6.

H3.3S28 contributes to H3K27me3 deposition. A, Metagene plot showing H3K27me3 level of UGs (n = 1,734) in H3.3 and H3.3S28A mutants. B, Metagene plot showing H3S28ph level of UGs (n = 1,734) in H3.3 and H3.3S28A mutants. C, Heat map showing the relative mRNA expression levels of PRC2 subunits in H3.1S28A and H3.3S28A, log2Fold Change was used. D, Relative expression level of the CLF and SWN genes by RT–qPCR from 2-week-old H3.1, H3.1S28A, H3.3, and H3.3S28A seedlings. Two biological replicates were performed. Data are means ± sd.

Discussion

Histone modifications are influenced by the local environment, including other nearby histone modifications. This study employed a dominant-negative mutation approach to investigate the phenotypic, transcriptomic, and epigenetic consequences of disrupting one such histone modification, H3S28ph. Our work revealed that a transgene expressing H3S28A acts in a dominant-negative manner to influence plant development by transcriptional regulation. The H3S28ph histone mark is considered to be negatively associated with H3K27me3, and enhanced H3S28ph may disrupt PRC2 recruitment and H3K27me3 (Lau and Cheung, 2011). We sought to determine whether a reduction in H3S28ph would lead to increased H3K27me3. Contrary to our expectation, H3K27me3 levels did not increase at either TE or gene loci, despite decreased H3S28ph in H3.1S28A and H3.3S28A. Hence, reduced H3S28ph is not sufficient to induce H3K27me3, whereas ectopic H3S28ph decreases H3K27me3 levels. It is known that multiple factors are involved in regulating H3K27me3 levels. H2A ubiquitination catalyzed by the PRC1 complex is essential for PRC2 function (Liu et al., 2021). Active histone modifications, including H3K4me3 and H3K36me2/3, play a role in repressing H3K27me3 (Schmitges et al., 2011). Another possibility for the lack of increased H3K27me3 levels is the limited reduction of H3S28ph in H3.1S28A and H3.3S28A transgenic lines. Both H3.1/HTR13 and H3.3/HTR5 are ubiquitously expressed during development. A very strong dominant-negative effect is observed upon mutation of H3K36M, which shows a notable reduction of H3K36me3 levels in Arabidopsis (Sanders et al., 2017). However, only a few hundred genes and TEs showed substantially reduced H3S28ph levels in the H3.1S28A and H3.3S28A transgenic lines, and the global H3S28ph levels were unchanged. Thus, plants probably have more chance to survive with H3S28A than H3K36M. Alternatively, it might indicate that H3S28ph is vital to survival, resulting in transgenic plants with severely decreased H3S28ph levels dying at early stages. Thus, the lines we analyzed are all weak alleles with limited reduction of H3S28ph.

We found that many known PRC2 targets with high H3K27me3 deposition lost H3K27me3 in the H3.3S28A mutant. Moreover, loss of H3K27me3 at these known PRC2 targets was independent of the variation in H3S28ph. In line with the partial reduction of H3K27me3, the activity of the PRC2 complex was potentially repressed in H3.3S28A due to decreased expression of two catalytic PRC2 subunits, CLF and SWN. Alternatively, given that lacking transcription may trigger H3K27me3 accumulation in human cells (Hosogane et al., 2016), it is also possible that H3.3S28A induces gene transcription, and in turn, leads to reduced H3K27me3 in Arabidopsis. At last, it is also possible that H3S28A influences the recruitment of PRC2 components or other steps of H3K27me3 deposition independently of phosphorylation levels at those loci. Similar to this hypothesis, the H3S28A mutation destabilizes nucleosome structure and impairs PRC2-dependent H3K27 methylation in Drosophila (Yung et al., 2015). Hence, here we showed that H3S28 in Arabidopsis functions in maintaining H3K27me3 levels. While the mechanism is not fully clear or completely the same, the contribution of H3S28 to H3K27me3 deposition is potentially conserved between Drosophila and Arabidopsis.

In addition, being involved in H3K27me3 deposition, H3S28ph is also well known to strongly accumulate at centromeric regions during metaphase and anaphase in cell division (Gernand et al., 2003). The evolutionarily conserved protein kinase family Aurora is involved in monitoring chromosome segregation via phosphorylation of different substrates, such as H3S28ph (Andrews et al., 2003; Kawabe et al., 2005). Although our transgenic H3.3S28A lines presented substantially decreased H3S28ph levels at the centromeric regions of mitotic chromosomes (Figure 2), both mitotic and meiotic cell division in such plants were normal (Supplemental Figure S3). Given that residual H3S28ph could be clearly seen in H3.3S28A plants, it is possible that this remaining amount of H3S28ph is sufficient to maintain chromosome segregation during both mitotic and meiotic segregation. Another possibility is that H3S28ph is not required for chromosome segregation, even though it accumulates at centromeric regions during metaphase (Gernand et al., 2003). The expected role of H3S28ph is based on the deposition of H3S28ph and the phenotype of aurora mutants (Petrovska et al., 2012; Demidov et al., 2014; Magnaghi-Jaulin et al., 2019). Given that Aurora kinases have multiple phosphorylation targets (Willems et al., 2018), substrates other than H3S28ph are likely responsible for maintaining chromosome segregation.

Although H3.1S28A and H3.3S28A did not present defects in cell division, both showed an early-flowering phenotype similar to that of mutants with disrupted H3K27me3 deposition (Goodrich et al., 1997; Chanvivattana et al., 2004; Jiang et al., 2008). Accordingly, flowering-promoting genes, such as AG, AP1, FT, SEP3, and AGL17, were upregulated in both H3.1S28A and H3.3S28A, and the increase was stronger in H3.3S28A than in H3.1S28A. In mammals, H3S28ph is involved in the transcriptional activation of immediate early genes (Mahadevan et al., 1991). Moreover, a genome-wide study in human cells revealed a novel function of H3S28ph in the transcriptional activation of stress-inducible genes (Sawicka et al., 2014). Our data suggest that H3S28ph contributes to both H3K27me3-dependent and H3K27me3-independent gene repression in Arabidopsis. On one hand, we detected a loss of H3K27me3 at many known PRC2 targets in H3.3S28A. On the other hand, the increased expression of PRC2 targets, such as AG, AP1, FT, SEP3, and AGL17 are independent of altered H3K27me3 levels (Supplemental Figure S8), suggesting an indirect effect of H3.3S28A. Therefore, H3S28 has dual roles in H3K27me3-mediated silencing. While we had the control of H3.1 and H3.3 to evaluate the dosage effect of exogenous H3.1 and H3.3, potential side effect of exogenous H3.1S28A or H3.3S28A could not be completely excluded. In the future, it will be worth simultaneously mutating multiple H3S28 in one plant using a base-editing or prime-editing approach. Such materials will be ideal for dissecting the function of H3S28A in Arabidopsis.

In conclusion, our study reveals that H3S28 is essential for plant development and gene expression, as it influences H3K27me3 deposition and downstream transcriptional regulation in Arabidopsis.

Materials and methods

Plant materials and growth conditions

The Arabidopsis Col-0 ecotype was used as the WT. Seeds were stratified at 4°C for 3 days before being sown on soil. Alternatively, seeds were surface sterilized in a solution containing 75% (v/v) ethanol with 0.1% (v/v) Triton X-100 for 10 min and rinsed with 95% ethanol 3 times before being sown on half-strength Murashige and Skoog (1/2 MS) medium plates (1% (w/v) sucrose, 0.8% (w/v) plant agar). The plates were kept at 4°C for 3 days to synchronize germination before being moved into the light. All plants were grown in the greenhouse under a long-day (LD; 16-h-light/8-h-dark photoperiod) condition at 22°C with 70% humidity.

Plasmid construction and generation of transgenic plants

Full-length genomic DNA fragments of histone H3.3 (AT4G40040) and H3.1 (AT5G10390) and their promoters (−1 kb) were cloned into pENTR/D-TOPO (Invitrogen, Waltham, MA, USA). The S28A mutation was created via site-directed mutagenesis using the Fast Mutagenesis Kit V2 (Vazyme, Nanjing, China; C214-02). Constructs were recombined into a pEarleyGate302 binary vector with a FLAG tag (Earley et al., 2006) using the LR reaction kit (Invitrogen; 11791020). The pEarleyGate destination vectors were transformed into Agrobacterium tumefaciens strain GV3101 and then transformed into Arabidopsis Col-0 plants using the floral dip method (Clough and Bent, 1998). Transgenic plants were selected on 1/2 MS plates containing 30 μg⋅L⁻¹ Basta. All primers used in this study are listed in Supplemental Table S6.

Immunostaining

Four-day-old seedlings grown on 1/2 MS plates were synchronized overnight at 4°C and then fixed with 4% paraformaldehyde in Tris buffer (10-mM Tris, 10-mM EDTA, 100-mM NaCl, 0.1% Triton X-100, pH 7.5) under vacuum for 5 min. Fixation continued on ice for 30 min. Fixed tissue was then chopped using a fresh sharp razor blade in LB01 buffer (2-mM Na₂EDTA, 20-mM NaCl, 2-mM EDTA, 80-mM KCl, 0.5-mM spermine, 15-mM β-mercaptoethanol, 0.1% Triton X-100, pH 7.5) and filtered through a 30-μm Cell Trics filter (Sysmex, Germany). The cell suspension was centrifuged onto slides using a Cytospin3 (Shandon) at 700 rpm for 5 min. The slides were then blocked in 5% BSA for 1 h at room temperature. Nuclei were incubated at 4°C overnight in Monoclonal Anti-H3S28ph (1:500; Sigma, St. Louis, MO, USA; H9908) and Polyclonal Anti-H3T3ph (1:500; Millipore, Burlington, MA, USA; 07-424). Slides were washed in phosphate-buffered saline (PBS) and incubated with Goat Anti-Rat Alexa Fluor 488 (1:250; Jackson ImmunoResearch, West Grove, PA, USA; 112-545-167) and Goat Anti-Rabbit Alexa Fluor 568 (1:250; Invitrogen, A-11029) for 1 h at 37°C. Following PBS washes, nuclei were stained with 4′,6-diamidino-2-phenylindole (DAPI) and analyzed using a Zeiss LSM780 confocal microscope (laser 488 nm, 600 gains for H3S28ph, laser 561 nm, 400 gains for H3T3ph and laser 405 nm, 500 gains for DAPI). Relative fluorescence intensity was analyzed in ImageJ. Statistical analyses were performed by one-way ANOVA followed by post-hoc Tukey’s test at P < 0.05.

Flow cytometry

Two-week-old leaves were chopped up, and the leaf material was mixed with 1 mL of Galbraith’s buffer (45-mM MgCl₂, 20-mM 3-[N-morpholino] propanesulfonic acid, 30-mM sodium citrate, 0.1% [v/v] Triton X-100). The released cell nucleus suspension was filtered using a 30-μm CellTrics filter (Sysmex, Germany, 04-0042-2316). The cytometer was subsequently calibrated by setting the DNA content peak of radish (Raphanus sativus) to 300. Peaks at 100 and 200 indicated 2× (diploid), whereas lack of these peaks in the presence of other higher-value peaks indicated 4×.

RNA extraction and RT–qPCR analysis

Total RNA from 2-week-old seedlings was extracted using the MagMAX Plant RNA Isolation kit (Thermo Fisher, Waltham, MA, USA; #A33784) according to the manufacturer’s instructions. Second-strand cDNA was synthesized using the cDNA Synthesis kit (Thermo Fisher, K1612). Reverse transcription–quantitative PCR (RT–qPCR) reactions contained Solis BioDyne-5x Hot FIREPol EvaGreen qPCR Supermix (ROX, Solis BioDyne, 08-36-00008), and the runs were performed in a QuantStudio 6 Flex Real-Time PCR System (Applied Biosystems, Waltham, MA, USA). Two biological replicates were performed, using GADPH as the reference gene. All primers used in this study are listed in Supplemental Table S6.

RNA-seq and data analysis

Total RNA from 2-week-old seedlings was extracted using the RNeasy Plant Mini Kit (Qiagen; 74904). For RNA-seq, libraries were constructed with the VAHTS mRNA-seq V3 Library Prep Kit (Vazyme; NR611-01) according to the manufacturer’s instructions. The libraries were sequenced at Novogene (UK) via a Novaseq instrument in 150-bp paired-end mode. Two biological replicates were performed. Reads were mapped to the TAIR10 WT Arabidopsis genome with HISAT2 (Kim et al., 2019) in paired-end mode. DEGs were analyzed via the Subread (Liao et al., 2019) and DESeq2 (Love et al., 2014) R packages with a 0.05 false discovery rate (FDR). The heatmap was generated using TBtools (Chen et al., 2020).

ChIP-seq and data analysis

Approximately 1 g of 2-week-old seedlings grown under LD conditions was collected.

The ChIP experiments were performed as described by Moreno-Romero et al. (2016), with minor changes. In brief, after the isolation of nuclei, the chromatin was sonicated in Nuclei lysis buffer (50-mM Tris–HCl, pH 8.0, 10-mM EDTA, 1% (w/v) SDS, 0.1-mM PMSF, 1-μM pepstatin A, Protease Inhibitor Cocktail (Roche, Basel, Switzerland; 11836145001)) using a Bioruptor Plus sonication device (Diagenode) to obtain the desired DNA fragment size (enriched at 300 bp). The following procedures were described by Moreno-Romero et al. (2016). After the de-crosslinking, DNA was purified according to the IPure kit v2 Kit manual (Diagenode, C03010015). ChIP-seq libraries were prepared using NEBNext Ultra II DNA Library Prep Kit (NEB, E7645). The following antibodies were used: anti-H3 (Sigma; H9289), anti-H3S28ph (Millipore, Burlington, MA, USA; 07-145), and anti-H3K27me3 (Millipore; 07-449).

The ChIP-seq libraries were sequenced at Novogene (UK) via a Novaseq instrument in 150-bp paired-end mode, and 15 million reads were generated for each library. Two biological replicates were performed. Bioinformatic analysis was performed following previously described procedures (Jiang et al., 2017). Quality control and adapter trimming of ChIP-seq reads were performed with fastp (Chen et al., 2018). Afterwards, reads were mapped to the Arabidopsis genome using Bowtie2 (Langmead and Salzberg, 2012) in pair-end mode. Mapped reads were deduplicated using MarkDuplicates (https://broadinstitute.github.io/picard). Coverage was estimated and normalized to 10 million reads. H3S28ph and H3K27me3 ChIP signals were normalized by subtracting their coverage with H3 ChIP data at every position in the genome. These data were standardized and normalized for comparative purposes across samples with a z-score (Cheadle et al., 2003) and represented in bedGraph files of 50-bp bins. Genes with differential levels of modifications between the WT and mutants were identified using linear models as implemented in the limma R package (Ritchie et al., 2015), using the information in the two replicates for both conditions. Genes with a log2 fold change of P <0.05 were selected as differentially modified loci. Kolmogorov–Smirnov tests were performed at https://scistatcalc.blogspot.com.

Data access

All sequencing data have been submitted to the NCBI Gene Expression Omnibus (GEO; https://www.ncbi.nlm.nih.gov/geo/) under accession number GSE197502.

Accession numbers

Sequence data from this article can be found in the GenBank/EMBL data libraries under accession number GSE197502.

Supplemental data

The following materials are available in the online version of this article.

Supplemental Figure S1. The phenotype of H3S28A point mutation plants.

Supplemental Figure S2 . Fertility analysis of H3S28A homozygous lines.

Supplemental Figure S3. Morphological phenotype of H3S28A grown under LD conditions.

Supplemental Figure S4. Screen shot of genome-wide H3S28ph and H3K27me3 levels for all five chromosomes in Col-0, H3.1, H3.1S28A, H3.3, and H3.3S28A.

Supplemental Figure S5. H3S28ph and H3K27me3 levels in regions 2,000-bp upstream of start codon and 2,000-bp downstream of stop codon.

Supplemental Figure S6. Genome browser views of PRC2-targeted genes with increased expression in H3.3S28A showing H3S28ph and H3K27me3 levels in H3.1, H3.1S28A, H3.3, and H3.3S28A.

Supplemental Figure S7. H3.1S28 does not contribute to H3K27me3 deposition.

Supplemental Figure S8. Genome browser views of H3K27me3-targeted flowering genes (blue box) showing H3S28ph and H3K27me3 levels in H3.1, H3.1S28A, H3.3, and H3.3S28A.

Supplemental Table S1. DEGs in H3.1S28A and H3.3S28A.

Supplemental Table S2. Levels of H3S28ph and H3K27me3 in Col, H3.1, H3.1S28A, H3.3, and H3.3S28A over TEs with decreased H3S28ph level in H3.1S28A and H3.3S28A.

Supplemental Table S3. Levels of H3S28ph and H3K27me3 in Col, H3.1, H3.1S28A, H3.3, and H3.3S28A over genes with decreased H3S28ph level and PRC2 targets in H3.1S28A and H3.3S28A.

Supplemental Table S4. Levels of H3S28ph and H3K27me3 in Col, H3.1, H3.1S28A, H3.3, and H3.3S28A over regions 2,000-bp upstream of start codon or 2,000-bp downstream of stop codon with decreased H3S28ph level in H3.1S28A and H3.3S28A.

Supplemental Table S5. Levels of H3S28ph and H3K27me3 over UGs in H3.1S28A and H3.3S28A.

Supplemental Table S6. List of primers used in this study.

Supplemental Table S7. Quality of ChIP-sequencing samples.

 

L.X. and J.C. executed the experimental procedures. L.X. and H.J. analyzed the NGS data. L.X. and H.J. performed the experimental design. L.X. and H.J. wrote the manuscript. All authors discussed the results and commented on the manuscript.

The author responsible for distribution of materials integral to the findings presented in this article in accordance with the policy described in the Instructions for Authors (https://academic.oup.com/plphys/pages/general-instructions) is Hua Jiang (jiangh@ipk-gaterslebemn.de).