Plants are sessile organisms, but they may undergo changes in physical, chemical, and biological environments, such as temperature, water, nutrition, and pathogens, throughout their life. Abnormal environmental conditions can exert deleterious effects on plant growth and development. In order to tolerate or resist the adverse environment, plants reprogram genome-wide expression profiles in which a large number of stress-responsive genes may play protective roles (Fujita et al., 2006). Therefore, transcriptional regulation plays a key role in every aspect of stress regulation and has been intensively studied (Yamaguchi-Shinozaki and Shinozaki, 2006); however, the histone modification and chromatin remodeling that are probably involved in transcription have not been well addressed in plants.
Here, we show that H3K27me3, the conserved histone marker for Polycomb group proteins in both plants and animals (Kohler and Villar, 2008), is removed in two cold-responsive genes, COR15A and ATGOLS3, during cold-induced transcription. In the transcribed regions of COR15A and ATGOLS3, H3K27me3 gradually decreases during cold-induced transcription without notable depletion of the H3 histone. In contrast, in the promoter region of ATGOLS3, a decrease of H3K27me3 is coupled with a loss of histone H3, particularly at earlier time points during cold-induced transcription. At the same time, removal of H3K27me3 independent of removal of H3 is also observed in the promoter region of ATGOLS3. Based on our results, we propose three, possibly non-exclusive, mechanisms that contribute to the removal of H3K27me3 in the promoter and transcribed regions of actively transcribed genes. First, chromatin remodeling resulting in reduced histone occupancy can be used to eliminate H3K27me3 in transcriptionally activated promoter regions (Figure S2). In general, a higher occupancy of histones in promoter regions negatively influences transcription (Workman and Kingston, 1998). Consistent with this, lower nucleosome density has often been found in the promoter regions of constitutively expressed eukaryotic genes (Barrera and Ren, 2006). In contrast, certain inducible genes contain nucleosomes in the promoter region, which should be remodeled or removed for transcriptional induction (Williams and Tyler, 2007). One of the best examples is the effect of phosphate deficiency on the PHO5 promoter in yeast, where four positioned nucleosomes are disassembled by the histone H3/H4 chaperone Asf1 (Adkins et al., 2004). Evolutionarily conserved ATP-dependent chromatin remodelers such as SWI/SNF can also change histone occupancy through histone eviction, nucleosome sliding, and other nucleosome restructuring in eukaryotic promoter regions (Kwon and Wagner, 2007; Clapier and Cairns, 2009). In pea plants, illumination results in increased accessibility of nuclease to the promoter and enhancer regions of the plastocyanin gene, suggesting that nucleosome disassembly may occur in these regions (Chua et al., 2001). Under drought conditions, histone density decreases in the promoter regions of RD20 and RAP2.4 but not in those of RD29A and RD29B in Arabidopsis (Kim et al., 2008). Here, we investigated H3 occupancy to show that chromatin remodeling resulting in reduced histone occupancy may occur in the promoter regions of ATGOLS3 and COR15A during cold-induced transcription. Second, H3K27me3 can be eliminated by a jumonji-class enzyme(s) specific for H3K27 that can be targeted to the promoter and transcribed regions (Figure S2). Two jumonji-class enzymes, UTX and JMJD3, have recently been identified as histone demethylases for both dimethylated and trimethylated H3K27 (Agger et al., 2007; De Santa et al., 2007; Hong et al., 2007; Lan et al., 2007). In Drosophila, these enzymes can be targeted to promoters and coding regions by specific transcriptional activators and the general transcription machinery (Smith et al., 2008). In humans, UTX can be found in association with Set1-like H3K4 methyltransferase complexes that are targeted to the 5′ ends of certain actively transcribed genes (Cho et al., 2007; Issaeva et al., 2007; Lee et al., 2007). Among the 21 predicted jmjC domain-containing proteins in Arabidopsis, one or more might be an enzyme(s) responsible for the demethylation of H3K27me3 (Lu et al., 2008). Third, H3K27me3 can be removed via histone replacement in the promoter and transcribed regions (Figure S2). It is known that the histone variant H3.3 replaces the canonical histone H3.1 in actively transcribed genes (Mito et al., 2005). In Arabidopsis, the histone variant H3.2 (H3.3 in animals) does not appear to contain a detectable level of H3K27me3 (Johnson et al., 2004). Therefore, we expect that the histone variant H3.2 replaces trimethylated H3.1 at Lys27 during active transcription in plants (Ahmad and Henikoff, 2002; Schwartz and Ahmad, 2005).
Our study also reveals that the decreased level of H3K27me3 following cold exposure can be maintained for up to 3 days after transcriptional attenuation. This suggests that the quantitative level of H3K27me3 can be inherited to some degree through cell division. However, the decrease in H3K27me3 alone does not appear to be sufficient for transcriptional reactivation of COR15A and ATGOLS3. One possibility is that transcriptional activation machinery is only present during cold exposure and is thus not available for re-induction of COR15A and ATGOLS3 during cold attenuation. Another possibility is that the H3K27me3 present in COR15A and ATGOLS3 cannot repress these genes directly. It has been reported that H3K27me3 is not sufficient for gene repression in Arabidopsis (Schubert et al., 2006). Actually, we show that the cold-induced decrease in the H3K27me3 level does not affect the transcriptional induction of COR15A and ATGOLS3 upon re-exposure to cold temperatures. Moreover, the promoter activity of COR15A is quite stable during cold exposure, whereas the H3K27me3 level in COR15A decreases gradually depending on the length of time for which the plants are exposed to the cold. In Arabidopsis, LHP1 binds to H3K27me3 in vitro (Zhang et al., 2007c; Exner et al., 2009) and is required for silencing of genes with H3K27me3 (Mylne et al., 2006; Sung et al., 2006; Exner et al., 2009). A recent genome-wide study of LHP1 showed that COR15A and ATGOLS3 are not targets for LHP1 binding (Zhang et al., 2007c). Rather, a subset of genes, in which H3K27me3 usually extends over a 1-kb region, serves as the target for LHP1 binding (Zhang et al., 2007c). Thus, it is likely that the extent of H3K27me3 present in COR15A and ATGOLS3 is not enough for LHP1 binding, which may be essential for silencing. In FLOWERING LOCUS C, which can be repressed by vernalization treatment, the silenced state requires a threshold level of H3K27me3 and differs from the silencing phase, where there is accumulation of inactive histone modifications such as H3K27me3 (Shindo et al., 2006). In addition, the quantitative level of H3K27me3 correlates with the natural variation in the vernalization requirement (Shindo et al., 2006). We therefore speculate that H3K27me3 levels in COR15A and ATGOLS3 represent an intermediate stage (silencing phase) along the way to the silenced state, and that transcriptional activation of these genes upon short-term cold exposure leads to a decrease in H3K27me3, thereby lowering the probability that these genes will be silenced. Similarly, we observed that the PR1, PR2, and PR5 genes, which are typical marker genes induced following pathogen stress, carry H3K27me3 but are not targets for LHP1 (Zhang et al., 2007b,c). We thus envisage that previous exposure of plants to certain environmental stresses may negatively affect the level of H3K27me3 and lower the chance of stress-responsive genes being silenced. This is likely to be beneficial for plants because they are sessile and may be exposed repeatedly to the same environmental conditions. An irreversible silencing of stress-responsive genes might hinder the rapid response of plants to adverse environmental conditions.