Histone H2B Monoubiquitination in the Chromatin of FLOWERING LOCUS C Regulates Flowering Time in Arabidopsis

Abstract

Ubiquitination is one of many known histone modifications that regulate gene expression. Here, we examine the Arabidopsis thaliana homologs of the yeast E2 and E3 enzymes responsible for H2B monoubiquitination (H2Bub1). Arabidopsis has two E3 homologs (HISTONE MONOUBIQUITINATION1 [HUB1] and HUB2) and three E2 homologs (UBIQUITIN CARRIER PROTEIN [UBC1] to UBC3). hub1 and hub2 mutants show the loss of H2Bub1 and early flowering. By contrast, single ubc1, ubc2, or ubc3 mutants show no flowering defect; only ubc1 ubc2 double mutants, and not double mutants with ubc3, show early flowering and H2Bub1 defects. This suggests that ubc1 and ubc2 are redundant, but ubc3 is not involved in flowering time regulation. Protein interaction analysis showed that HUB1 and HUB2 interact with each other and with UBC1 and UBC2, as well as self-associating. The expression of FLOWERING LOCUS C (FLC) and its homologs was repressed in hub1, hub2, and ubc1 ubc2 mutant plants. Association of H2Bub1 with the chromatin of FLC clade genes depended on UBC1,2 and HUB1,2, as did the dynamics of methylated histones H3K4me3 and H3K36me2. The monoubiquitination of H2B via UBC1,2 and HUB1,2 represents a novel form of histone modification that is involved in flowering time regulation.

INTRODUCTION

The transition from vegetative to reproductive development is a critical event in the plant life cycle that is coordinated by both endogenous signals and environmental cues (Searle and Coupland, 2004; Baurle and Dean, 2006; Imaizumi and Kay, 2006). Genetic analysis has identified several pivotal genes involved in the control of flowering time in Arabidopsis thaliana via multiple pathways (Simpson and Dean, 2002; Searle and Coupland, 2004; He and Amasino, 2005; Imaizumi and Kay, 2006), including FLOWERING LOCUS C (FLC), which is the central repressor of flowering in Arabidopsis. FLC inhibits the expression of flowering-time integrators, including FLOWERING LOCUS T (FT) and SUPPRESSOR OF OVEREXPRESSION OF CONSTANS1, and it indirectly represses such floral meristem identity genes as LEAFY and APETALA1 (He and Amasino, 2005; Baurle and Dean, 2006; Imaizumi and Kay, 2006; Searle et al., 2006). Inactivation of FLC leads to flowering; thus, its level of expression is tightly regulated (Michaels and Amasino, 1999; Sheldon et al., 1999). There are five FLC homologs, MADS AFFECTING FLOWERING (MAFs), in the Arabidopsis genome, each of which is also involved in the control of flowering time (Ratcliffe et al., 2001, 2003; Scortecci et al., 2001).

The covalent modification of histones is important for chromatin remodeling and proper transcriptional regulation (Jenuwein and Allis, 2001). These modifications include methylation, acetylation, phosphorylation, ADP-ribosylation, and ubiquitination (Turner, 2002; Kouzarides, 2007). The ubiquitination of proteins is performed in a specific manner by E1, E2, and E3 enzymes (Hellmann and Estelle, 2002; Sullivan and Deng, 2003; Pickart, 2004; Smalle and Vierstra, 2004). Two types of protein ubiquitination, multiubiquitination and monoubiquitination, have been described to date. Multiubiquitination targets proteins for degradation by the 26S proteasome, whereas monoubiquitination mediates the localization and activation of the ubiquitinated proteins (Hicke, 2001; Zhang, 2003).

Histone monoubiquitination involves mainly histones H2A and H2B (Zhang, 2003). H2A monoubiquitination, which is mediated by polycomb repression complex 1 in animals, is necessary for the silencing of homeobox genes (Wang et al., 2004). In comparison, monoubiquitinated H2B (H2Bub1) controls the binding of histone H3 lysine methylases, such as COMPASS, to the chromatin in yeast and thus is required for the methylation of H3K4 and H3K36 (Dover et al., 2002; Sun and Allis, 2002; Henry et al., 2003; Wood et al., 2003a; Lee et al., 2007). H2Bub1 worked together with COMPASS, RNA polymerase II associated factor 1 (PAF1) complex, and RNA polymerase II to stimulate transcription elongation (Pavri et al., 2006; Lee et al., 2007; Weake and Workman 2008). In yeast, the monoubiquitination of H2B to H2Bub1 is mainly achieved by two enzymes, a ubiquitin E3 ligase encoded by Brefeldin A-sensitivity protein1 (BRE1) and a ubiquitin E2 conjugase encoded by Radiation sensitivity protein6 (RAD6) (Robzyk et al., 2000; Hwang et al., 2003; Kao et al., 2004). H2Bub1 functions in the control of cell size in yeast (Hwang et al., 2003). The homologs of BRE1 have been found to perform the same biochemical function in animals, Drosophila BRE1 is required for wing development (Bray et al., 2005), and human BRE1 is involved in HOX gene regulation (Zhu et al., 2005).

In the Arabidopsis genome, there are two BRE1 homologs (HISTONE MONOUBIQUITINATION1 [HUB1] and HUB2) and three RAD6 homologs (UBIQUITIN CARRIER PROTEIN1 [UBC1], UBC2, and UBC3). Liu et al. (2007) reported that mutations in HUB1 or HUB2 led to defects in seed dormancy and bushy plants with pale green leaves, while Fleury et al. (2007) found that hub1 or hub2 plants were abnormally small with narrow leaf laminae and stunted primary roots. Based on these results, it was suggested that HUB1,2 controls seed dormancy and leaf and root growth in Arabidopsis. However, the function of UBC1-3 has not been documented.

Recent studies have suggested the involvement of methylation and acetylation of histone H3 in FLC expression and flowering time control (He and Amasino, 2005). For example, it was found that FLOWERING LOCUS D (FLD), a homolog of animal lysine-specific demethylase 1, represses FLC expression by decreasing the histone acetylation level of FLC chromatin (He et al., 2003). In addition, the polycomb group protein and the plant homeodomain protein VERNALIZATION INSENSITIVE3 (VIN3) establishes, while VERNALIZATION2 (VRN2) maintains, the repression of FLC expression by modifying the H3K27 and H3K9 methylation state of FLC chromatin in the vernalization pathway (Bastow et al., 2004; Sung and Amasino, 2004). Finally, the conserved PAF1 complex is necessary for enhancement of the H3K4me3 level of FLC chromatin and FLC expression in Arabidopsis (He et al., 2004).

In this article, we show that Arabidopsis E2 conjugases UBC1 and UBC2 work together with the E3 ligases HUB1 and HUB2 to monoubiquitinate histone H2B. We also show that the ubiquitinated H2B preferentially associates with the gene body of FLC clade genes and that this H2Bub1 is required for the enhancement of H3K4me3 and H3k36me2 and transcriptional activation in FLC/MAFs chromatin. Thus, this study provides evidence that H2Bub1 is linked to histone H3 methylation in an H2Bub1-dependent dynamic manner and is specifically involved in regulating the expression of key flowering time genes in Arabidopsis.

RESULTS

HUB1 and HUB2 Are Necessary for Flowering Time Control in Arabidopsis

There are two BRE1 homologs, HUB1 (At2g44950) and HUB2 (At1g55250 and At1g55255), in the Arabidopsis genome. To analyze the function of these homologs in Arabidopsis, two alleles of hub1 (hub1-4 and hub1-5) and one allele of hub2 (hub2-2) from the SALK insertion mutant collection were characterized. No full-length transcripts were detected for any of the three alleles; however, partial transcripts were identified in all three cases (Figure 1A).

Figure 1.

Mutation of HUB1 or HUB2 Leads to an Early Flowering Phenotype in Arabidopsis.

(A) Gene structures and characterization of the steady state mRNA levels of HUB1 and HUB2 by RT-PCR in the hub1 and hub2 mutants. Exons are represented by filled black boxes, introns by lines, untranslated regions by filled gray boxes, and T-DNA insertions by triangles. The primers used to detect the transcripts are indicated as P1 to P8. Pairs of P3/P4 and P7/P8 were used to detect the full-length transcripts of HUB1 and HUB2, respectively; Pairs of P1/P2 and P5/P6 were used to detect the partial transcripts of HUB1 and HUB2, respectively.

(B) and (C)Phenotypes of hub1 and hub2 single mutants and the hub1 hub2 double mutant at 26 (B) and 42 d old (C).

(D) and (E) Complementation of hub1 and hub2 by wild-type (D) (HUB1/hub1 and HUB2/hub2) or mutant (E)(HUB1M/hub1 and HUB2M/hub2) HUB1 or HUB2, respectively. Thirteen out of 15 and 18 out of 20 single copy insertion lines were rescued by the transformation of HUB1 into hub1-5 and hub1-4, respectively, whereas 16 out of 19 single copy insertion lines were rescued by the transformation of HUB2 into hub2-2. No line out of 54 lines for each genotype was rescued by the transformation of HUB1M into hub1-5 and hub1-4, and no line out of 64 lines was rescued by the transformation of HUB2M into hub2-2. Bars = 1 cm.

Mutations in HUB1 or HUB2 led to an early-flowering phenotype under both long-day (Figure 1B, Table 1) and short-day (Table 1) conditions. In addition, the small plant size and pale green leaves observed by Fleury et al. (2007) and Liu et al. (2007) using the alleles we used and two additional alleles were evident in the mutants (Figures 1B and 1C). The three alleles of hub1 and hub2 produced quite similar phenotypes (Figure 1B, Table 1). Furthermore, the phenotype of the double mutant was very similar to that of each single mutant (Figure 1B, Table 1), suggesting that both HUB1 and HUB2 are required for the control of flowering time in Arabidopsis.

Table 1.

Summary of Flowering of the Wild Type and Mutants

	LD		SD
Genotype	Rosette Leaf No.	n	Rosette Leaf No.	n
Wild type (Col-0)	12.9 ± 0.8	20	65.1 ± 3.8	18
hub1-5	10.5 ± 0.6	21	49.8 ± 5.2	12
hub1-4	8.4 ± 0.7	22	45.5 ± 4.2	17
hub2-2	10.3 ± 0.8	24	46.7 ± 4.3	15
hub1-5 hub2-2	8.7 ± 0.9	23	45.8 ± 4.8	17
hub1-4 hub2-2	9.0 ± 0.6	23	47.4 ± 3.7	14
ubc1-1	12.5 ± 0.8	22	67.9 ± 3.5	16
ubc2-1	12.7 ± 0.9	21	65.4 ± 5.2	16
ubc3-1	12.8 ± 0.7	21	63.3 ± 4.2	18
ubc1-1 ubc2-1	9.0 ± 0.7	23	44.3 ± 7.0	14
ubc1-1 ubc3-1	12.8 ± 0.7	23	66.8 ± 4.1	15
ubc2-1 ubc3-1	13.0 ± 0.8	20	62.6 ± 4.0	15

LD, long-day conditions (16 h light/8 h dark); SD, short-day conditions (8 h light/16 h dark).

Next, a complementation assay was conducted by transforming HUB1 and HUB2 into hub1 and hub2 mutant plants, respectively. Native promoter-driven HUB1 and HUB2 coding sequence (CDS) completely rescued the hub1 and hub2 phenotypes (Figure 1D, Table 2). By contrast, mutant versions of HUB1 (C⁸⁴¹A and H⁸⁴³A) and HUB2 (C⁸⁴⁴A and H⁸⁴⁶A) (which alter the two conserved and necessary amino acid residues for E3 activity in yeast; see Supplemental Figure 2B online) were unable to rescue the phenotypes of the hub1 and hub2 mutants (Figure 1E), suggesting that the E3 activity of HUB1 and HUB2 is necessary for its function in Arabidopsis flowering time control.

Table 2.

Summary of Flowering of the Wild Type, Mutants, and Complementation Lines

Genotype	Rosette Leaf No.	n
Wild type (Col-0)	13.9 ± 1.1	25
hub1-5	10.5 ± 0.8	29
hub1-4	9.2 ± 0.6	29
hub2-2	10.4 ± 0.9	28
HUB1/hub1-5	13.8 ± 0.9	24
HUB1/hub1-4	14.3 ± 1.0	24
HUB2/hub2-2	14.2 ± 1.0	24
ubc1-1 ubc2-1	8.7 ± 0.7	20
UBC1/ubc1-1 ubc2-1a	13.4 ± 1.5	12
UBC2/ubc1-1 ubc2-1b	14.0 ± 1.4	12

^aTransformant containing UBC1 genomic DNA in the ubc1 ubc2 mutant background.

^bTransformant containing UBC2 genomic DNA in the ubc1 ubc2 mutant background.

Neither HUB1 nor HUB2 was functional if yellow fluorescent protein (YFP) was fused to its C terminus; however, both proteins functioned normally with YFP fused to their N termini. The RING finger domain in HUB1 and HUB2 is located at the very end of the C terminus, which suggests that a free RING finger domain is necessary for proper functionality in HUB1 and HUB2.

HUB1 and HUB2 Form Homodimers and Heterodimers

Genetic analysis revealed that both HUB1 and HUB2 are necessary for flowering time control in Arabidopsis (Figure 1B); this observation is also true for HUB1,2-mediated leaf development and seed dormancy (Fleury et al., 2007; Liu et al., 2007). One possible explanation for this result is that the two proteins work together in the same protein complex. Thus, we tested for a physical interaction between HUB1 and HUB2 by yeast two-hybrid assay. HUB1 and HUB2 were able to self-associate and to interact in a pairwise fashion (Figure 2A), suggesting that they form both homodimers and heterodimers.

Figure 2.

Self-Association and Pairwise Interaction of HUB1 and HUB2.

(A) Yeast two-hybrid assay. Positive control: pGADT7-T + pGBKT7-53 (encoding the fusions between GAL4 DNA-BD and AD and murine p53 and SV40 large T-antigen, respectively). p53 and large T-antigen interact in a yeast two-hybrid assay. Negative control: pGADT7-T + pGBKT7-Lam (encoding a fusion of the DNA-BD with human lamin C and provides a control for a fortuitous interaction between an unrelated protein and either the pGADT7-T control or AD/library plasmid). Lamin C neither forms complexes nor interacts with most other proteins. The indicated combinations of plasmids were cotransformed into the yeast reporter strain, and the interaction of HUB1 with HUB2 was assessed by growth on plates lacking Leu, Trp, His, and adenine or by analysis of the activation of a second reporter gene (β-galactosidase). Bar = 1 cm.

(B) BiFC assay to detect the interaction of HUB1 with HUB2 in Arabidopsis protoplasts. One (YFP^N-HUB1 + YFP^C) out of four negative controls (YFP^N-HUB1 + YFP^C; YFP^C-HUB1 + YFP^N; YFP^N-HUB2 + YFP^C; and YFP^C-HUB2 + YFP^N) is shown. Bar = 20 μm.

This result was confirmed in Arabidopsis protoplasts using a bimolecular fluorescence complementation (BiFC) assay. HUB1 and HUB2 interacted with both HUB1 and HUB2, whereas neither HUB1 (YFP^N-HUB1 or YFP^C-HUB1) nor HUB2 (YFP^N-HUB2 or YFP^C-HUB2) was able to interact with YFP^C or YFP^N (Figure 2B), suggesting that the observed interactions were specific. The interaction between HUB1 and HUB2 took place in the nucleus (Figure 2B), which is consistent with the function (histone modification) of the proteins.

UBC1 and UBC2 Are Functionally Redundant in the Control of Flowering Time in Arabidopsis

RAD6 is conserved in a number of species, and there are three RAD6 homologs, UBC1, UBC2, and UBC3, in the Arabidopsis genome (see Supplemental Figure 1 online). To determine the function of each of these E2s and whether they are involved in the HUB1,2 pathway, we first characterized T-DNA insertion mutants for the three genes. One T-DNA insertion mutant was available for each of the three genes in the SALK insertion mutant collection (Alonso et al., 2003). Transcript analysis indicated a lack of full-length transcripts in all three mutants. ubc1 and ubc2 had partial transcripts, whereas ubc3 did not, suggesting that it was a null mutant (Figure 3A).

Figure 3.

Characterization of ubc1, ubc2, and ubc3 Mutants.

(A) Gene structures and characterization of the steady state mRNA levels of UBC1, UBC2, or UBC3 by RT-PCR. Exons are represented by filled black boxes, introns by lines, untranslated regions by filled gray boxes, and T-DNA insertions by triangles. The primers used to detect the transcripts are indicated as P1 to P8. Pairs of P1/P2, P4/P5, and P7/P8 were used to detect the full-length transcripts of UBC1, UBC2, and UBC3, respectively; pairs of P1/P3 and P4/P6 were used to detect the partial transcripts of UBC1 and UBC2, respectively.

(B) The phenotype of ubc1, ubc2, ubc3, and each double mutant at 26 d old.

(C) Complementation of ubc1 ubc2 by UBC1 (UBC1/ubc1 ubc2) or UBC2 (UBC2/ubc1 ubc2) genomic DNA (135 of 144 lines were rescued by the transformation of UBC1, while 153 of 161 lines were rescued by the transformation of UBC2).

(D)Phenotype of the ubc1 ubc2 ubc3 triple mutant. Bars = 1 cm.

None of the single mutants exhibited obvious defects in flowering time (Figure 3B, Table 1). The ubc1 ubc2 double mutant flowered early and exhibited a phenotype similar to that of hub1 or hub2 (Figures 1B and 3B, Table 1), whereas the ubc1 ubc3 and ubc2 ubc3 double mutants did not show visible defects in flowering time under both long- and short-day conditions (Figure 3B, Table 1). The early-flowering phenotype observed in ubc1 ubc2 double mutant was completely rescued by transformation of the double mutant with UBC1 or UBC2 genomic DNA (Figure 3C, Table 2), indicating that UBC1 and UBC2 are functionally redundant and that UBC3 is not involved in the control of flowering time in Arabidopsis.

Although ubc1 ubc3 and ubc2 ubc3 double mutants did not exhibit an obvious flowering phenotype, the ubc1 ubc2 ubc3 triple mutant exhibited pleiotropic defects, including dwarfism, an increased number of branches, and sterility (Figure 3D). These results suggest that the three E2s are involved in multiple developmental processes and that they are functionally redundant in processes other than flowering time.

UBC1 and UBC2 Exhibit a Similar but Different Expression Pattern from UBC3 in the Shoot Apical Meristem

UBC1, UBC2, and UBC3 are homologous proteins with very high identities (see Supplemental Figure 1 online). Expression pattern analysis using promoter-β-glucuronidase (GUS) transgenic plants revealed that the three genes are expressed ubiquitously in seedlings, leaves, roots, and flowers (Figure 4A), which is consistent with the results of a previous report (Thoma et al., 1996), and that they are functionally redundant in the control of multiple developmental processes.

Figure 4.

Expression Pattern of UBC1, UBC2, UBC3, HUB1, and HUB2 by Promoter-GUS Fusion.

(A) Histochemical analysis of UBC1, UBC2, UBC3, HUB1, and HUB2 expression in leaf, root, and flower.

(B) Expression pattern of HUB1, HUB2, and UBCs in the SAMs of 8-d-old seedlings. Nine out of 10 (UBC1), 10 out of 10 (UBC2), 10 out of 11 (UBC3), eight out of 10 (HUB1), and nine out of 10 (HUB2) independent lines exhibited the expression pattern shown. Bar = 10 μm.

However, the results of the above functional analysis suggested that UBC3 is different from UBC1 and UBC2 in terms of its effect on flowering time in Arabidopsis (Figure 3B). As floral organ identity is determined by the genes expressed in the shoot apical meristem (SAM) (Baurle and Dean, 2006), we analyzed the expression pattern of UBC1, UBC2, and UBC3 in the SAM. Our efforts to determine the expression level of these genes in the SAM by in situ hybridization were hampered by their apparently weak expression. We therefore examined the expression pattern of the E2s in the SAM using promoter-GUS fusion transgenic lines. As shown in Figure 4B, UBC1:GUS and UBC2:GUS were ubiquitously expressed in the SAM, whereas no UBC3:GUS expression was detected in the central region of the SAM. This suggests that the expression pattern of UBC3 is different from that of UBC1 or UBC2 in the SAM, and this difference in expression may be responsible for the distinct role that each gene plays in the control of flowering time.

Physical and Genetic Interactions between HUB1 or HUB2 and UBC1 or UBC2

The hub1 or hub2 single mutant phenotype is similar to the ubc1 ubc2 double mutant phenotype (Figures 1B and 3B), indicating that these two groups of proteins may interact and that they may be involved in the same pathway. We found that the expression patterns for HUB1, HUB2, UBC1, and UBC2 were very similar in seedlings, leaves, roots, flowers, and the SAM (Figure 4). We examined the physical interaction between UBC1/2 and HUB1/2 in yeast. It was found that HUB1 and HUB2 interacted with UBC1 and UBC2 in yeast, as indicated by the growth of yeast strains cotransformed with HUB1 or HUB2 and UBC1 or UBC2 on high-stringency selection medium (Figure 5A). The RING finger domain of HUB1 or HUB2 was shown to be necessary for its interactions with UBC1 and UBC2, due to the fact that each interaction was abolished if the RING finger domain was deleted (Figure 5B). However, the observed interactions were weak, as indicated quantitatively by a β-galactosidase activity assay (Figure 5A). No interaction was detected for UBC1-UBC1, UBC2-UBC2, or UBC1-UBC2, suggesting that these E2s cannot form homodimers or heterodimers.

Figure 5.

Physical and Genetic Interactions between HUB1 or HUB2 and UBC1 or UBC2.

(A) HUB1 or HUB2 interacts with UBC1 or UBC2.

(B) Deletion of the RING finger domain in HUB1 and HUB2 (constructs pGBKT7- HUB1N_1-824 and pGADT7-HUB2N_1-848) abolishes the interaction between HUB1 or HUB2 and UBC1 or UBC2. Positive control: pGADT7-T + pGBKT7-53; negative control: pGADT7-T + pGBKT7-Lam. The indicated combinations of plasmids were cotransformed into the yeast reporter strain, and the interaction between HUB1,2 and UBCs was assessed by growth on plates lacking Leu, Trp, His, and adenine or by analysis of the activation of β-galactosidase. The full-length and RING finger domain-deleted HUB proteins expressed in the yeast are shown in the bottom.

(C) The early-flowering phenotype of hub1 or hub2 and ubc1,2 double and triple mutants. Bar = 1 cm.

(D) The rosette leaf number at flowering of the hub1 or hub2 and ubc1,2 double and triple mutants. The data for wild type and each mutant are from at least 20 plants. Error bars indicate sd.

To analyze the genetic interaction between HUB1 or HUB2 and UBC1 or UBC2, we generated double and triple mutants for the corresponding genes. The phenotypes of the double and triple mutants were similar to those of the hub1 or hub2 single and ubc1 ubc2 double mutants (Figures 5C and 5D), thus providing genetic evidence of the involvement of the UBC1,2/HUB1,2 pathway in flowering in Arabidopsis.

HUB1/2 and UBC1/UBC2 Are the E3s and E2s Responsible for the Production of H2Bub1 in Arabidopsis

To test whether HUB1/2 and UBC1/UBC2 are the E3s and E2s responsible for the production of H2Bub1 in flowering time control in Arabidopsis under our conditions, we examined the in vivo H2Bub1 pattern in wild-type and mutant plants. As the anti-H2B antibodies for animal H2B did not work in Arabidopsis and anti-H2Bub1 antibody was not available when we conducted the experiment, FLAG-tagged H2B (At3g45980) was transformed into wild-type and hub1/2 or ubc1/2/3 plants. Anti-FLAG antibodies detected two bands (upper and lower) in wild-type, ubc1, ubc2, ubc3 single mutant, ubc1 ubc3, and ubc2 ubc3 double mutant plants (Figure 6A). However, no upper band was detected in hub1 and hub2 single, double, or ubc1 ubc2 double mutant plants (Figure 6A), indicating that the upper band corresponds to monoubiquitinated H2B. This result further suggests UBC1,2 and HUB1,2 are required for H2Bub1 formation and that the UBC1,2/HUB1,2-mediated production of H2Bub1 is genome-wide and not gene specific.

Figure 6.

Analysis of the H2Bub1 Level in 7-d-Old Seedlings.

(A) Detection of H2Bub1 in wild-type and mutant plants. The upper band was monoubiquitinted Flag-H2B, and the lower band was Flag-H2B. The bands were detected by anti-Flag antibody. H2Bub1, monoubiquitinated H2B; MW, molecular weight.

(B) No H2Bub1 was detected in wild-type plants transformed with a mutant form of H2B (K¹⁴⁶A). The U in the circle is ubiquitin, and peptide in the bottom of the panel is the C-terminal sequence of H2B.

(C) The H2B-specific upper band was recognized by antiubiquitin antibodies after the immunoprecipitation by FLAG antibodies.

(D) and (E) Overexpression of the K¹⁴⁶A mutant form of histone H2B in wild-type plants (ProH2B:H2BK146A/WT) leads to early flowering. The levels of the mutated form of H2B protein in different lines are shown at the top of the graph in (E). Ten out of 34 independent lines exhibited an early-flowering phenotype (four are shown). Bar in (D) = 1 cm.

To obtain additional evidence in support of this notion, we transformed a mutant version of H2B into wild-type plants. Based on the site of ubiquitination in human H2B (Zhu et al., 2005), K¹⁴⁶ was identified as the likely site of monoubiquitination in Arabidopsis H2B. The upper band was absent from transgenic plants expressing mutant (K¹⁴⁶A) H2B, whereas it was present in transgenic plants carrying wild-type H2B (Figures 6B and 6C). The upper band was recognized by antiubiquitin antibodies as well (Figure 6C). These results confirm that the upper band corresponds to monoubiquitinated H2B and that the site of monoubiquitination in Arabidopsis is K¹⁴⁶.

Moreover, the H2Bub1 patterns in the wild-type and mutant plants clearly indicate that the early-flowering phenotype of the mutants is caused by the loss of H2Bub1 (Figure 6A). In addition, we found that the phenotype of 10 out of 34 independent transgenic lines expressing the mutated version of H2B (H2BK¹⁴⁶A) was indistinguishable from that of hub1 and hub2 single mutants and the ubc1 ubc2 double mutant (Figures 6D and 6E), and the early flowering phenotype was correlated with the expression level of H2BK¹⁴⁶A protein (Figure 6E, lines 2 and 30 exhibited phenotype, while lines 8 and 22 did not), providing further evidence that H2Bub1 is required for flowering time control in Arabidopsis.

To further confirm that HUB1 and HUB2 are functional homologs of yeast BRE1, HUB1 or HUB2 was transformed into yeast bre1. Both HUB1 and HUB2 rescued the bre1 phenotype when expressed under the control of the yeast BRE1 promoter (see Supplemental Figure 2A online). However, the bre1 phenotype was not rescued if HUB1 or HUB2 was driven by the yeast alcohol dehydrogenase1 (highly and constitutively expressed) promoter (see Supplemental Figure 2C online), suggesting that the precise level of HUB1,2 is important for its function.

To further provide evidence that the E3 activity of HUB1,2 is necessary for the rescue of the yeast bre1 phenotype, a mutated version of HUB1 (C⁸⁴¹A or H⁸⁴³A) or HUB2 (C⁸⁴⁴A or H⁸⁴⁶A) (see Supplemental Figure 2B online) was transformed into yeast bre1 cells. The bre1 phenotype was not rescued by the mutated version of HUB1 or HUB2, even if they were driven by the BRE1 promoter (see Supplemental Figure 2D online). Thus, HUB1 and HUB2 are functional homologs of yeast BRE1, and the E3 activity of the proteins is necessary for their functionality in yeast.

Defects in H2Bub1 Inhibit the Expression of FLC Clade Genes in Arabidopsis

As FLC and its homologs are central players in the control of flowering time (Simpson and Dean, 2002; Searle and Coupland, 2004; He and Amasino, 2005), we examined whether the expression of FLC and its homologs was altered in hub1 and hub2 single mutants and ubc1 ubc2 double mutants by quantitative real-time PCR. FLC expression was repressed in hub1-4 and ubc1 ubc2, but surprisingly not in hub2-2 or hub1-4 hub2-2 (Figure 7A). In addition, the expression of some other members of the FLC clade was repressed in the mutants, albeit to different extents (Figure 7A). Among them, the expression of MAF4 and MAF5 in the mutants was significantly repressed compared with that in wild-type plants (Figure 7A). As the expression of FLC was not repressed in hub2 or hub1 hub2, we examined the expression of flower-time integrator gene FT and found that FT expression was significantly increased in all four mutants (Figure 7B), which is consistent with the early flowering phenotype for those mutants. This result suggests that repression of other FLC clade genes, such as MAF1, MAF4, and MAF5, are responsible for the early flowering phenotype in hub2-2 and hub1-4 hub2-2. The reason for this inconsistent repression of FLC expression in the different mutants is unclear. There is no obvious change for the expressions of ACTIN, TUBULIN, elongation factor 1 (eEF-1), or glyceraldehyde- 3-phosphate dehydrogenase (GAPDH) between wild-type and the four mutant plants (see Supplemental Figure 3 online). This result indicates that defects in H2Bub1 specifically lead to the repression of FLC and some other FLC clade genes, thus promoting flowering.

Figure 7.

H2Bub1 Is Required for the Transcriptional Activation of FLC/MAFs.

(A) Analysis of the relative expression level of FLC/MAFs by quantitative real-time PCR using RNA from the SAM of 7-d-old seedlings. The levels were normalized to the wild type.

(B) Analysis of the relative expression level of FT by quantitative real-time PCR using 10-d-old seedlings. The levels were normalized to the wild type. Three biological replications were performed for each experiment. Actin was used as an internal control. Error bars indicate sd.

Monoubiquitination of FLC/MAF Chromatin by the UBC1,2/HUB1,2 Pathway

We next examined whether FLC/MAF chromatin can be monoubiquitinated and whether the UBC1,2/HUB1,2 pathway is responsible for the modification. As our double chromatin immunoprecipitation (ChIP) assay using anti-FLAG(H2B) and -ubiquitin antibodies was unsuccessful, we used a monoclonal anti-H2Bub1 antibody (Minsky et al., 2008) for this purpose. We first examined whether the antibody could be used in Arabidopsis and found that the antibody clearly recognized a protein band with a molecular mass of ∼28 kD. The band was not detected in hub1, or ubc1 ubc2 mutant plants (Figure 8B), suggesting that the antibody works in Arabidopsis and indicating again that the UBC1,2/HUB1,2 pathway is required for H2Bub1.

Figure 8.

Monoubiquitination of FLC/MAF Chromatin.

(A) The structure of FLC. P1 to P8 represent the primers used to assess the level of H2Bub1, H3K4me3, or H3K36me2 by ChIP. The filled boxes represent exons, while the open boxes represent introns. +1, the translation initiation point.

(B) Detection of H2Bub1 in Arabidopsis using a monoclonal antibody against H2Bub1. H3 was used as a loading control.

(C) ChIP analysis of the H2Bub1 level in FLC/MAFs chromatin followed by PCR. Antibodies used for IP are indicated at top, and plant genotypes are indicated above each lane.

(D) ChIP was used to analyze the level of H2Bub1 in chromatin across the FLC locus (P1 to P8 as in [A]), including the promoter and gene body region, from wild-type, hub1, and ubc1 ubc2 seedlings (light gray, dark gray, and white bars, respectively). The amount of immunoprecipitated chromatin as normalized to the total amount of chromatin used in FLCP1 region in wild-type plants was given as 1.

(E) Differences in the level of H2Bub1 in the gene body region of FLC/MAF chromatin between wild-type and mutant plants as detected by ChIP analysis.

(F) ChIP analysis of the H2Bub1 level in FLC/MAF chromatin from wild-type seedlings. The data represent the amount of immunoprecipitated chromatin versus the total amount of chromatin used.

(G) Detection of H2Bub1, H3K4me3, and H3K36me2 in ACTIN chromatin by ChIP analysis. At least three independent immunoprecipitations were performed for each experiment. Unless otherwise indicated, the immunoprecipitated DNA was quantified by quantitative real-time PCR. The data in (E) and (G) were normalized to the input chromatin.

H2B was monoubiquitinated in the gene body region of FLC chromatin; however, it was either not ubiquitinated or monoubiquitinated at a very low level in the promoter region (Figures 8C and 8D). There was a significant reduction in the level of H2Bub1 in the gene body region of FLC chromatin in hub1 and ubc1 ubc2 plants (Figures 8C and 8D). H2B was also monoubiquitinated in the chromatin of other FLC clade genes (Figures 8C, 8E, and 8F), though the level of monoubiquitination in MAF2 and MAF3 chromatin was much higher than that in FLC, MAF4, or MAF5 (Figure 8F). The level of H2Bub1 in the chromatin of all FLC clade genes was significantly reduced in the mutants compared with wild-type plants (Figures 8C and 8E). Thus, FLC clade gene-specific H2Bub1 is dependent on the UBC1,2/HUB1,2 pathway. We also detected H2Bub1 in the chromatin of the ACTIN gene, and it was determined that the monoubiquitination was dependent on the activity of UBC1,2/HUB1,2 (Figures 8C and 8G).

H2Bub1 Is Required for the H3K4 and H3K36 Hypermethylation of FLC/MAF Chromatin

The activation and repression of a gene are controlled by the modification state of its chromatin. In general, histones H3K4me3 and H3K36me2 are necessary for gene activation, while H3K9me2 is associated with gene silencing (Li et al., 2007). Thus, we examined the interaction between H2Bub1 and H3 methylation in Arabidopsis. Although no H2Bub1 was detected in hub1, hub2, or ubc1 ubc2 mutant plants, there was no obvious difference in the genome-wide level of H3K4me3, H3K4me2, H3K4me1, H3K36me2, or H3K9me2 modification between wild-type and hub1, hub2, or ubc1 ubc2 mutant plants (Figure 9A).

Figure 9.

H2Bub1 Is Required for H3K4 and H3K36 Hypermethylation in FLC/MAF Chromatin.

(A)Analysis of H3K4, H3K9, and H3K36 methylation on a genome-wide scale by protein gel blotting.

(B) Detection of H3K4me3, H3K36me2, and H3K9me2 in the gene body region of FLC/MAF chromatin from wild-type and mutant plants by ChIP analysis followed by PCR. ACTIN was used as an internal control for H3K4me3 and H3K36me2, while TA3 was used as an internal control for H3K9me2.

(C) Detection of H3K4me3 in the promoter and gene body regions of FLC chromatin by ChIP analysis.

(D) ChIP analysis of the H3K4me3 level in FLC/MAF chromatin.

(E) ChIP analysis of the H3K4me3 level in FLC/MAF chromatin from wild-type seedlings. The data represent the amount of immunoprecipitated chromatin versus the total amount of chromatin used.

(F) Detection of H3K36me2 in the promoter and gene body regions of FLC chromatin by ChIP analysis.

(G)ChIP analysis of the H3K36me2 level in FLC/MAF chromatin.

(H)ChIP analysis of the H3K9me2 level in FLC/MAF chromatin. At least three independent immunoprecipitations were performed for each experiment. Unless otherwise indicated, the immunoprecipitated DNA was quantified by quantitative real-time PCR.

The data in (C), (D), (F), and (G) were normalized to ACTIN, while the data in (H) were normalized to TA3.

Thus, we examined whether H2Bub1 is necessary for gene-specific histone H3 methylation by comparing the levels of H3K9me2, H3K36me2, and H3K4me3 in FLC clade gene chromatin between wild-type and hub1-4 (a representative hub mutant) or ubc1 ubc2 plants. There was no obvious difference between the wild-type and hub1 or ubc1 ubc2 plants in terms of their H3K9me2 levels in the chromatin of FLC clade genes (Figures 9B and 9H). However, the H3K4me3 level was reduced in the gene body region of FLC chromatin, whereas no obvious reduction was detected in the promoter region in the hub1 single or ubc1 ubc2 double mutants (Figures 9B and 9C). The H3K4me3 level was reduced in the chromatin of MAF1, MAF4, and MAF5 in the hub1 single and ubc1 ubc2 double mutants as well (Figures 9B and 9D). The level of H3K4me3 in the chromatin of MAF2 and MAF3 was much higher than that in the chromatin of FLC, MAF4, or MAF5 (Figure 9E); moreover, there was no obvious change in the H3K4me3 level in the chromatin of MAF2 and MAF3 between wild-type and hub1 or ubc1 ubc2 plants (Figures 9B and 9D). Similarly, no obvious change in the level of H3K4me3 in the chromatin of ACTIN was found between wild-type and hub1 or ubc1 ubc2 plants (Figures 8G and 9B).

In terms of H3K36me2, the modification pattern in FLC chromatin was similar to that for H3K4me3 in wild-type plants (Figure 9F), and a similar change in the H3K4me3 modification pattern was observed in the chromatin of FLC clade genes between wild-type and hub1 or ubc1 ubc2 plants (Figures 9B and 9G). There was no obvious change in the level of H3K36me2 in the chromatin of ACTIN between wild-type and hub1 or ubc1 ubc2 plants (Figures 8G and 9B). These results indicate that the UBC1,2/HUB1,2 pathway is involved in the regulation of H3K4me3 and H3K36me2 levels and that H2Bub1 is required for the enhancement of H3K4 and H3K36 hypermethylation in the chromatin of FLC and some other FLC clade genes.

Genetic Interaction between HUB1 or HUB2 and the PAF1 Complex Component Gene ELF7

The above results suggested that the UBC1,2/HUB1,2 pathway regulates H2Bub1 to mediate flowering time in Arabidopsis. It was previously shown that E2 and E3 involved in H2Bub1 work together with the PAF1 complex in the same pathway in yeast and animals (Ng et al., 2003; Wood et al., 2003b; Zhu et al., 2005). It was also reported that defects in Arabidopsis homologs of the components of the PAF1 complex, such as EARLY FLOWERING7 (ELF7), ELF8, and VERNALIZATION INDEPENDENCE4, led to early flowering (Zhang and van Nocker, 2002; He et al., 2004). We therefore examined whether UBC1,2/HUB1,2 and the PAF1 complex function in the same pathway in Arabidopsis by characterizing the phenotype of hub and elf7 single and double mutants. The hub1 or hub2 and elf7 single mutants exhibited an early flowering phenotype, while their double mutants exhibited a similar but slightly stronger early flowering phenotype (Figure 10). Moreover, the elf7 mutants exhibited a stronger phenotype during leaf and flower organ development than the hub1 or hub2 mutants (Figure 10; He et al., 2004), suggesting that the function of UBC1,2/HUB1,2 and the PAF1 complex is synergistic, and indicating the action of no identical mechanisms during Arabidopsis development.

Figure 10.

Genetic Interaction between HUB1, HUB2, and ELF7.

The plants were grown under long-day conditions for 26 d. The rosette leaf number at flowering of the hub1 or hub2 and elf7 double and triple mutants is shown at the bottom. The data for the wild type and each mutant are at least from 20 plants. Error bars indicate sd. Bar = 1 cm.

DISCUSSION

H2Bub1 and Arabidopsis Development

The covalent modification of histone H2B to H2Bub1 is conserved from yeast to human (Hwang et al., 2003; Wood et al., 2003a; Bray et al., 2005; Zhu et al., 2005). In Arabidopsis, mutations in HUB1,2 have been shown to cause defects in cellular proliferation during early leaf and root development, resulting in plants of reduced size with narrow leaf laminae (Berna et al., 1999; Fleury et al., 2007). In addition, defects in HUB1 or HUB2 have been shown to reduce seed dormancy and to result in plants with pale green leaves and a bushy appearance (Liu et al., 2007). Furthermore, a recent report suggested that H2B deubiquitination is critical for non-CpG DNA methylation and transcriptional gene silencing (Sridhar et al., 2007).

In this article, we provide several lines of evidence showing that mutations in UBC1,2/HUB1,2, a set of E2/E3 genes, lead to early flowering in addition to the defects in plant size and leaf development in Arabidopsis (Figures 1, 3, and 5). Furthermore, we found that the phenotype was highly correlated with the ubiquitination state of H2B (Figures 1, 3, and 6) and that a loss of H2Bub1 leads to early flowering (Figures 1, 3, and 6). Together, these results suggest that a novel form of histone modification (H2Bub1) is involved in the control of flowering time in Arabidopsis.

We also found that the monoubiquitination site in H2B is in the C terminus, at K¹⁴⁶ (Figure 6); however, Sridhar et al. (2007) identified K¹⁴³ as the ubiquitination site of H2B. It should be noted that K¹⁴⁶ in the C terminus (AVTKFTSS) of At3g45980 (the gene used in our experiments) corresponds to K¹⁴³ (AVTKFTSS) in another isoform of H2B (At1g07790), suggesting that the modification of H2B is not isoform specific.

A Conserved E2/E3 Pathway Regulates the Production of H2Bub1 from Yeast to Human and Plant

Protein ubiquitination is achieved by a cascade involving an E1, an E2, and an E3. The E2 and E3 responsible for the production of H2Bub1 were first identified in yeast (RAD6 and BRE1, respectively) (Robzyk et al., 2000; Hwang et al., 2003; Kao et al., 2004). Since then, homologs of BRE1 with the same biochemical function have been identified in animals (Bray et al., 2005; Zhu et al., 2005) and plants (Figure 6; Fleury et al., 2007; Liu et al., 2007). In this study, we provided genetic and biochemical evidence that Arabidopsis UBC1 and UBC2 act as E2s together with HUB1,2 to produce H2Bub1 (Figures 1, 3, 5, and 6). Interestingly, several features of the ubiquitination of H2B are conserved in multiple species: (1) homologs of the E2 and E3 responsible for the production of H2Bub1 in yeast (Robzyk et al., 2000; Hwang et al., 2003; Kao et al., 2004) have been identified in plant (this article); (2) the site of ubiquitination in H2B and the primary sequence surrounding the ubiquitination site (AVTKFTSS) are highly conserved from yeast to human and plant; (3) histone H2Bub1 is genome-wide, not gene-specific, from yeast to plant; (4) H2Bub1 is associated with transcriptional activation in diverse eukaryotic organisms. Thus, chromatin remodeling via the RAD6/BRE1-mediated production of H2Bub1 has been conserved during evolution.

H2Bub1, H3K4, and H3K36 Methylation in Arabidopsis

We found that a loss of H2Bub1 had no obvious effect on the global level of H3K4me3 or H3K36me2 in Arabidopsis (Figure 9A). By contrast, a defect in H2Bub1 led to the loss of genome-wide H3K4me3 in yeast (Hwang et al., 2003), and the knock down of H2Bub1 in human cells significantly decreased the genome-wide H3K4me3 level (Zhu et al., 2005). These findings suggest that H3K4me3 is entirely dependent on H2Bub1 in yeast, but not in plants.

In addition, H2Bub1 has more functions than H3K4me3 in fission yeast, as the loss of H2Bub1 produced stronger defects than the loss of H3K4me3 (Tanny et al., 2007). Furthermore, a defect in the putative histone H3 methyltransferase EARLY FLOWERING IN SHORT DAYS, which decreased the level of H3K4me3 in FLC chromatin, produced pleiotropic and stronger phenotypes in Arabidopsis than in the hub1 or hub2 mutant, which lacks H2Bub1 (Kim et al., 2005; this article). These results indicate that H3K4me3 plays a greater number of roles in Arabidopsis than H2Bub1. Therefore, the importance of H2Bub1 appears to have been progressively weakened, while that of H3K4me3 has become strengthened, during evolution.

H2Bub1 is associated with active transcription; thus, it was proposed to be coupled to H3K4me3 and H3K36me2 in yeast and humans (Henry et al., 2003; Zhu et al., 2005; Lee et al., 2007). We found that the H2B in chromatin of FLC-clade flowering repressor genes was monoubiquitinated (Figure 8); in particular, the H2Bub1 level in the gene body was higher than that in the promoter region (Figure 8), which is consistent with the role of H2Bub1 in transcriptional elongation (Pavri et al., 2006; Lee et al., 2007). The levels of H2Bub1 and H3K4me3 in the chromatin of several FLC clade genes were variable, and the level of H2Bub1 was correlated with the level of H3K4me3 for individual genes (Figures 8F and 9E). A defect in H2Bub1 significantly decreased the level of H3K4me3 and H3K36me2 in the chromatin of FLC, MAF1, MAF4, and MAF5 (Figure 9) and repressed the expression of those genes (Figure 7), suggesting that H2Bub1 is required for the enhancement of H3K4me3 and H3K36me2 and the increased expression of those genes.

However, there was no obvious change in the steady state level of H3K4me3 or H3K36me2 in the chromatin of MAF2 and MAF3 or in the expression of either gene after the removal of H2Bub1 from the chromatin of MAF2 and MAF3 in the hub1, hub2, or ubc1 ubc2 mutant (Figures 8 and 9). The same results were obtained for the constitutively expressed gene ACTIN (Figures 8G and 9B). All three genes showed a higher level of H3K4me3 in their chromatin (Figure 9E), indicating that H3K4me3 and the expression of those genes with a high level of H3K4me3 in their chromatin may not be dependent on H2Bub1. This result is consistent with the lack of an obvious change in the global H3K4me3 level between wild-type and hub1 or hub2 mutant plants (Figure 9A).

The reason for the uncoupling between H2Bub1 and H3 methylation in the chromatin of those genes with a high level of H2Bub1 and H3K4me3 is unclear. FLC family proteins show a high level of sequence identity (53 to 87%), and MAF2 to MAF5 are arranged in a tight cluster in the Arabidopsis genome (Ratcliffe et al., 2003). Significantly greater H2Bub1 was detected in the chromatin of MAF2/MAF3 than in that of other FLC clade genes (Figure 8F), suggesting that H2Bub1 in the chromatin of MAF2 and MAF3 is more stable, while that in the chromatin of such genes as FLC, MAF4, and MAF5 is more dynamic. It was reported in yeast and humans that H2Bub1 was highly transient and that the dynamics of H2Bub1 were important for the regulation of transcription (Henry et al., 2003; Minsky et al., 2008). Both the monoubiquitination and deubiquitination of H2B are involved in transcriptional activation. The disruption of either process affects the transient dynamics of H2B ubiquitination, leading to alterations in the levels of H3K4me3 and H3K36me2 (Henry et al., 2003). Thus, the dynamic and transient features of H2Bub1 may depend on a low level of monoubiquitination, and only those genes with a low level of H2Bub1 in their chromatin may couple H2Bub1 and H3 methylation with transcriptional activation. This is consistent with the viable and subtle phenotypes of the hub1,2 and ubc1 ubc2 mutants (Figures 1 and 3). The function of H2Bub1 for those genes with a stable level of H2Bub1 in their chromatin remains to be determined.

UBC1, UBC2, and UBC3 Have Overlapping but Not Identical Roles in Arabidopsis Development

UBC1, UBC2, and UBC3 are three highly homologous proteins in Arabidopsis (see Supplemental Figure 1 online). The ubc1 ubc2 ubc3 triple mutant is defective in several developmental processes (Figure 3D), suggesting that these three E2s are involved in both vegetative and reproductive development in Arabidopsis. The phenotype of the triple mutant was much stronger than that of the single or double mutants (Figures 3B and 3D), suggesting that UBC1, UBC2, and UBC3 are functionally redundant for the above-mentioned developmental processes.

In terms of floral development, the ubc1 ubc2 double mutant exhibited an early-flowering phenotype; however, there was no obvious difference between ubc1 ubc3 or ubc2 ubc3 double mutants and the wild type in terms of flowering time (Figure 3B), and the early-flowering phenotype of the triple mutant was no stronger than that of the ubc1 ubc2 double mutant (Figure 3).

This indicates that UBC1 and UBC2 are functionally redundant and that UBC3 is not involved in flowering time control. UBC1 and UBC2 were both expressed ubiquitously in the SAM (Figure 4B), whereas UBC3 was not expressed in the central region of the SAM (Figure 4B), which greatly overlaps with the expression region of FLC in the SAM (Searle et al., 2006). The results of our mutant phenotype analysis (Figures 1, 3, and 5), in vivo histone ubiquitination assay (Figures 6A to 6C), and interaction analysis in yeast (Figure 5A) suggest that UBC1 and UBC2 are the E2s responsible for the production of H2Bub1. Thus, UBC1 and UBC2, but not UBC3, work together with the E3s HUB1 and HUB2 to regulate H2Bub1 and flowering time in Arabidopsis.

H2Bub1 Is a Novel Form of Histone Modification Essential for the Enhancement of H3K4 and H3K36 Hypermethylation and FLC/MAF Transcriptional Activation in Arabidopsis

The MADS box transcription factor FLC is a central repressor of the floral transition in Arabidopsis, other members of FLC family (MAFs) have similar roles in floral transition, and increases in FLC/MAF expression delay flowering (Henderson and Dean, 2004; He and Amasino, 2005). It has been shown that the acetylation and methylation of histone H3 is critical for the expression of FLC and flowering time control in Arabidopsis (He and Amasino, 2005; Zhao et al., 2005). FLD mediates the acetylation of FLC chromatin (He et al., 2003). VRN1, VRN2, and VIN3 regulate the level of H3K9me2 (Bastow et al., 2004; Sung and Amasino, 2004), while the H3K9 hyperdimethylation of FLC chromatin and the repression of FLC expression are maintained by LIKE HETEROCHROMATIN PROTEIN1 in Arabidopsis (Mylne et al., 2006; Sung et al., 2006). In this work, we found that H2Bub1 is essential for the enhancement of H3K4 and H3K36 hypermethylation but that it has no effect on the H3K9me2 level in FLC/MAFs chromatin (Figure 9). Thus, H2Bub1 is required for FLC/MAF transcriptional activation.

We found that both HUB1 and HUB2 were required, while UBC1 and UBC2 were functionally redundant, for the production of H2Bub1 and the control of flowering time (Figures 1, 3, 5, and 6). Moreover, HUB1 and HUB2 were capable of self-association and pairwise interaction (Figure 2). Thus, HUB1 and HUB2 may form a heterotetramer with two HUB1 and two HUB2 molecules. This E3 complex may then recruit UBC1 or UBC2 to FLC/MAF chromatin, followed by the transfer of ubiquitin from UBC1/UBC2 to H2B. The UBC1,2/HUB1,2 pathway worked together with H2Bub1 deubiquitylating enzyme (Ubp) to control the dynamics of H2Bub1, and the dynamics of ubiquitinated H2B may then serve as a platform for the recruitment of other components, such as histone H3 methyltransferase, which are responsible for the modification of H3 and the enhancement of H3K4 and H3K36 hypermethylation. Hypermethylation of H3 in FLC/MAF chromatin leads to increased FLC/MAF expression and the delay of flowering in Arabidopsis; thus, defects in this pathway repress the expression of FLC/MAFs and promote early flowering (Figure 11). This study shows that a novel form of histone modification (H2Bub1) is required for a key developmental process in Arabidopsis (i.e., floral transition), thus providing new insight into the mechanism of floral transition control in Arabidopsis.

Figure 11.

Proposed Model for the Modulation of FLC/MAF Chromatin and Transcription by the UBC1,2/HUB1,2 Pathway in the Control of Flowering Time in Arabidopsis.

In this model, HUB1 and HUB2 form a heterotetramer with two HUB1 and two HUB2 molecules to recruit UBC1 or UBC2 to FLC/MAF chromatin, followed by the transfer of ubiquitin from UBC1 or UBC2 to H2B. The ubiquitinated H2B may then serve as a platform for the recruitment of histone H3 methyltransferase for the enhancement of H3K4 and H3K36 hypermethylation, thus leading to increased FLC/MAF expression and the delay of flowering in Arabidopsis.

METHODS

Plant Materials and Growth Conditions

All of the Arabidopsis thaliana alleles used in this study were in a Columbia background. Seeds of hub1-5, hub1-4, and hub2-2, corresponding to SALK_044415, SALK_122512, and SALK_071289, seeds of ubc1-1, ubc2-1, and ubc3-1, corresponding to SALK_109196, SALK_060994, and SALK_045233, and seeds of elf7-3, corresponding to SALK_019433, were obtained from the ABRC. All T-DNA insertions were confirmed by PCR and sequencing.

For the mutant complementation assay, native promoter-driven HUB1 (2000 bp upstream of ATG; see Supplemental Table 3 online), HUB2 (2000 bp upstream of ATG; see Supplemental Table 3 online) CDS, UBC1 (2458 bp upstream ATG + gene body + 1059 bp downstream TAG; see Supplemental Table 3 online), or UBC2 (2553 bp upstream ATG + gene body + 1439 bp downstream TAG; see Supplemental Table 3 online) genomic DNA was transformed into hub1 or hub2 single or ubc1 ubc2 double mutant plants by the floral dip method using Agrobacterium tumefaciens (Clough and Bent, 1998). The seeds of the transformants were sterilized with 2.25% bleach, washed with sterilized water, kept at 4°C for 3 d, and then germinated on Murashige and Skoog medium containing 1% sucrose. After 8 d of growth under continuous white light (160 μmol m⁻² s⁻¹), the plants were transplanted to soil and grown under long-day (16 h of light [22°C]/8 h of dark [18°C]) or short-day (8 h of light [22°C]/16 h of dark [18°C]) conditions at 65% relative humidity.

RNA Extraction and Real-Time PCR

Total RNA was isolated from the SAMs of 7-d-old wild-type or mutant seedlings using an RNeasy plant mini kit (Qiagen) according to the manufacturer's protocol and then treated with RNase-free DNase (Promega) to degrade any remaining DNA. A total of 3 μg of RNA was used for cDNA synthesis using an oligo(dT) primer and SuperScript II reverse transcriptase (Invitrogen). Real-time PCR was subsequently performed to quantify the cDNA using TaKaRa SYBR Premix Ex Taq in a 7500 real-time PCR instrument (Applied Biosystems). ACTIN was included as an internal control, and the level of other housekeeping genes, such as TUBULIN, eEF-1, and GAPDH, among the wild type and mutants were also detected. The primers used for real-time PCR are provided in Supplemental Table 1 online.

ChIP

ChIP was performed as described by Gendrel et al. (2005) using 7-d-old seedlings, which were grown on Murashige and Skoog medium under continuous white light. Anti-H3K9me2, anti-H3K4me3, and anti-H3k36me2 antibodies were purchased from Upstate Biotechnology, and anti-H2Bub1 monoclonal antibody was purchased from Médimabs. As the H3K4me3 and H3K36me2 levels in ACTIN chromatin and the H3K9me2 level in TA3 chromatin were similar among the hub1, hub2, ubc1 ubc2, and wild-type plants, we used ACTIN as internal controls for the H3K4me3 and H3K36me2 levels and TA3 for the H3K9me2 level, respectively. The primers used for PCR and quantitative real-time PCR are provided in Supplemental Table 2 online.

GUS Staining

The ∼2-kb promoter of HUB1 (2000 bp upstream of ATG; see Supplemental Table 3 online), HUB2 (2000 bp upstream of ATG; see Supplemental Table 3 online), or UBCs (2007, 1997, and 2069 bp upstream of ATG for UBC1, 2, and 3, respectively; see Supplemental Table 3 online) was fused with GUS and transformed into wild-type plants, and T2 lines were subjected to histochemical analysis. GUS staining was performed as described by Jefferson et al. (1987) and Hemerly et al. (1993); the samples were cleared as described by Malamy and Benfey (1997). To analyze the expression pattern of HUB1, HUB2, UBC1, UBC2, and UBC3 in the SAM, 10 plants each from 10 independent T2 lines were selected for each gene. The level of GUS activity was then assessed by staining as described above, and the seedlings with GUS staining were embedded in Paraplast Plus. Each seedling was then sectioned every 8 μm and observed under a Zeiss light microscope as described by Sessions et al. (1999).

Protoplast Transient Expression and BiFC Assays

Protoplasts were isolated from 3- or 4-week-old Arabidopsis leaves grown under 12 h of light/12 h of dark as described by Asai et al. (2002). To generate BiFC constructs, full-length HUB1 and HUB2 CDS were obtained by RT-PCR and cloned into pSY735 and pSY736, respectively. The primers used for amplification of full-length HUB1 and HUB2 CDS were as follows: HUB1 forward 5′-ACGCGTCGACAATGGCGAGCACAGGCGAGCCTGAC-3′ and reverse 5′-CGCGGATCCTCATATGTAGATAGGTTTAATAT-3′; HUB2 forward 5′-ACGCGTCGACAATGGAGAATCAGGAATCGGACGAG-3′ and reverse 5′-CGCGGATCCTTACATTTTGACAAGCCGGACGTC-3′. All BiFC plasmids were purified by density gradient centrifugation. The test constructs were then cotransformed into the protoplasts using the polyethylene glycol transformation method (Bracha-Drori et al., 2004; Walter et al., 2004). YFP fluorescence in the transformed protoplasts was detected by confocal laser scanning microscopy (LSM 510 Meta; Zeiss) after 20 to 24 h of incubation in the dark. Roughly 26 to 28% of the protoplasts cotransformed with pSY735-HUB1 or HUB2 and pSY736-HUB1 or HUB2 exhibited YFP fluorescence, whereas YFP fluorescence was observed in <2 to 3% of the negative controls.

In Vivo H2B Ubiquitination and H3 Methylation Assay

FLAG-H2B (At3g45980) driven by the H2B native promoter was transformed into wild-type and hub1, hub2, or ubc plants as indicated in Figure 6. T2 seedlings were then used for detection of the H2Bub1 level, and total protein was extracted using plant extraction buffer (50 mM Tris-Cl, pH 7.5, 150 mM NaCl, 10 mM MgCl₂, and 0.1% Nonidet P-40, plus 1 mM PMSF and protease inhibitor cocktail). The proteins were then separated by 15% SDS-PAGE and transferred to PVDF membranes for blotting with an anti-FLAG monoclonal antibody (Sigma-Aldrich); tubulin, detected by antitubulin antibodies (Sigma-Aldrich), was used as a loading control. The bound antibodies were visualized using the ECL system (Amersham).

For the confirmation of H2Bub1 using antiubiquitin antibodies, 10-d-old T2 transgenic seedlings, produced by the transformation of FLAG-H2B or FLAG-H2BK¹⁴⁶A into wild-type plants, were used. Ten grams of plant material were ground in liquid nitrogen to a fine powder then mixed with 8 mL of IP buffer (50 mM Tris-Cl, pH 7.5, 150 mM NaCl, 10 mM MgCl₂, 1 mM EDTA, 10% glycerol, and 0.1% Nonidet P-40, plus 1 mM PMSF and protease inhibitor cocktail). After centrifugation at 14,000 rpm for 20 min, the supernatant was transferred to a fresh cold tube, passed through a 0.45-μm filter, and mixed overnight with 100 μL of anti-FLAG M2 Affinity Gel (Sigma-Aldrich) on a rotator at 4°C. The beads were then collected by centrifugation at 800g for 30 s, washed three times with cooled IP buffer, collected by centrifugation at 4500 rpm for 2 min at 4°C, and eluted with 5× SDS-PAGE sample buffer (without β-ME or DTT). The protein samples were then probed with anti-FLAG and antiubiquitin antibodies (Abcam), respectively.

For the detection of global-scale H3 modification and H2Bub1 level among wild-type and mutant plants, protein extracted from 7-d-old seedlings following the chromatin extraction and sonication of the ChIP protocol, chromation solution were diluted with 2× loading buffer and denatured at 100°C for 10 min and then the samples were probed with anti-H3K4me1, anti-H3K4me2, anti-H3K4me3, anti-H3K36me2, anti-H3K9me2 antibodies (all from Upstate Biotechnology), and anti-H2Bub1 antibody (Médimabs), respectively.

Yeast Complementation Assay

The yeast strains used in this study were kindly provided by Hiter D. Madhani (University of California, San Francisco). YM1740 corresponds to bre1Δ in a YM1736 (S288C) background. Yeast complementation plasmids (BRE1-HUB1 and BRE1-HUB2) were generated from pADH 1-BRE1 (Hwang et al., 2003). The pADH 1 promoter, which is the promoter of alcohol dehydrogenase 1, was replaced by the native promoter of yeast BRE1, while BRE1 was replaced by its Arabidopsis homolog, HUB1 or HUB2. Cell size was determined as described previously (Hwang et al., 2003).

Yeast Two-Hybrid Interaction Assay

All of the genes tested were cloned into two vectors (pGADT7 and pGBKT7) and cotransformed into yeast strain AH109. Transformation, yeast growth, and β-galactosidase quantitative assays were performed as described in the Clontech Yeast Protocols Handbook.

Sequence Alignment and Phylogenetic Analysis

The amino acid sequences of the open reading frames were initially aligned using the program ClustalX 2.0 with the following parameter value: gap opening = 10, gap extension = 0.2, delay divergent sequences% = 30, protein weight matrix: gonnet series. The phylogenetic tree was generated using the neighbor-joining method in Phylip3.6 and opened with TreeView 1.6. Numbers at nodes indicate the bootstrap value obtained for 1500 replicates.

Accession Numbers

Sequence data from this article can be found in the Arabidopsis Genome Initiative or GenBank/EMBL databases under the following accession numbers: HUB1 (At2g44950), HUB2 (At1g55250 and At1g55255), UBC1 (At1g14400), UBC2 (At2g02760), UBC3 At5g62540), H2B (At3g45980, At1g07790), ACTIN (At5g09810), TUBLIN (At1g64740), TA3 (At1g37110), eEF-1 (At5g60390), GAPDH (At1g42970), FLC (At5g10140), MAF1 (At1g77080), MAF2 (At5g65050), MAF3 (At5g65060), MAF4 (At5g65070), and MAF5 (At5g65080).

Supplemental Data

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

• Supplemental Figure 1. Sequence Alignment and Phylogenetic Tree of the RAD6 Homologs in Different Organisms.

• Supplemental Figure 2. HUB1 and HUB2 Rescue the Yeast bre1 Phenotype.

• Supplemental Figure 3. Relative Expression Level of ACTIN, TUBLIN, eEF-1, and GAPDH in Wild-Type and Mutant Plants.

• Supplemental Table 1. The Primers Used for Real-Time PCR in Gene Expression Assay.

• Supplemental Table 2. The Primers Used for Real-Time PCR in ChIP Assay.

• Supplemental Table 3. The Primers Used for Specific DNA Fragment Amplification.

• Supplemental Data Set 1. Sequence Alignment of RAD6 Homologs Corresponding to Supplemental Figure 1.