H2A.Z promotes the transcription of MIR156A and MIR156C in Arabidopsis by facilitating the deposition of H3K4me3

Shoot development in higher plants consists of a juvenile vegetative phase, an adult vegetative phase and a reproductive phase, which differ in many morphological and physiological traits (Poethig, 2003). The transition from the juvenile to the adult phase of vegetative development is regulated by an increase in the expression of squamosa promoter binding protein (SBP) genes, which is mediated by a decrease in the level of two miRNAs – miR156 and miR157 – that repress these genes. To understand the mechanism of vegetative phase change, it is therefore important to determine how the abundance of miR156 and miR157 is regulated.

miR156 is encoded by eight genes in Arabidopsis, two of which, MIR156A and MIR156C, produce the majority of the mature miR156 transcript and are particularly important for vegetative phase change (Yang et al., 2013; Yu et al., 2013). There is increasing evidence that epigenetic factors play important roles in the regulation of these genes. In particular, the downregulation of MIR156A and MIR156C during vegetative phase change is associated with a decrease in active histone modification, histone 3 lysine 27 acetylation (H3K27ac), and with an increase in the repressive histone modification, histone 3 lysine 27 trimethylation (H3K27me3), and requires this latter modification (Picó et al., 2015; Xu et al., 2016a). In contrast, the transcription of MIR156A is promoted by the nucleosome remodeler, BRAHMA (Xu et al., 2016b), and by the SWR1 complex (SWR1-C), which exchanges the histone variant H2A.Z for H2A (Choi et al., 2016). How these ATP-dependent nucleosome remodelers regulate the expression of MIR156A and other genes that encode miR156 is unknown. Here, we explore the mechanism by which SWR1-C regulates the transcription of MIR156A and MIR156C.

SWR1-C is an evolutionarily conserved multi-protein complex first described in budding yeast (Mizuguchi et al., 2004). Three components of SWR1-C are essential for its function in yeast. The ATPase Swr1 provides the catalytic activity for the complex, and works in association with the accessory proteins Arp6 and Swc6. The orthologue of Swr1 in Arabidopsis is PHOTOPERIOD-INDEPENDENT EARLY FLOWERING1 (PIE1); the orthologues of Arp6 and Swc6 are, respectively, ACTIN-RELATED PROTEIN6 (ARP6) and SERRATED LEAVES AND EARLY FLOWERING (SEF). As in yeast, these proteins associate with each other and loss-of-function mutations in any of these proteins cause a significant decrease in H2A.Z (March-Díaz et al., 2007; Wu et al., 2005; Wu et al., 2009b). The phenotype of pie1, arp6 and sef closely resembles the phenotype produced by mutations in genes that encode H2A.Z (March-Díaz et al., 2007), strongly suggesting that – as in yeast – the primary function of SWR1-C in Arabidopsis is to deposit H2A.Z (Mizuguchi et al., 2004; Wu et al., 2005).

H2A.Z affects many processes in fungi, plants and animals, including gene expression, recombination and DNA repair (Choi et al., 2013; Lu et al., 2009; Morrison and Shen, 2009; Rosa et al., 2013; van Attikum et al., 2007; Xu et al., 2012). In plants, H2A.Z has been implicated in the response to high temperature, the phosphate starvation response, osmotic stress, the immune response, floral induction, female meiosis, recombination, thalianol metabolism and the regulation of microRNA abundance (Choi et al., 2016, 2013; Deal et al., 2005; Kumar and Wigge, 2010; March-Díaz et al., 2008; Nützmann and Osbourn, 2015; Qin et al., 2014; Smith et al., 2010; Sura et al., 2017). The presence of H2A.Z at a locus is typically correlated with a high level of responsiveness to environmental factors (Coleman-Derr and Zilberman, 2012; Sura et al., 2017). Depending on the gene, the presence of H2A.Z may be correlated with an increase or a decrease in gene expression upon induction. This observation supports the conclusion that the presence of H2A.Z increases the susceptibility of a gene to factors that promote or repress transcription, rather than conferring a specific pattern of gene expression.

The mechanism by which H2A.Z affects transcription is still unclear. Some in vitro analyses suggest that H2A.Z destabilizes nucleosomes, whereas others indicate that it increases nucleosome stability (Abbott et al., 2001; Bowerman and Wereszczynski, 2016; Hoch et al., 2007; Hong et al., 2014; Jin and Felsenfeld, 2007; Kruger et al., 1995). Similarly, in Arabidopsis, mutations that block the deposition of H2A.Z can either increase (Choi et al., 2016; Nützmann and Osbourn, 2015) or decrease (Kumar and Wigge, 2010) nucleosome occupancy. The possibility that H2A.Z affects transcription by influencing modifications of other histone proteins is suggested by the observation that H2A.Z colocalizes with H3K4me3 near the transcription start site (TSS) of many genes (Choi et al., 2013), and by the observation that the abundance of H3K4me3 at FLC is reduced in arp6 mutants (Martin-Trillo et al., 2006).

A previous study showed that arp6 reduces the abundance of miR156, in part by reducing the expression of MIR156A (Choi et al., 2016). However, the biological significance of this effect is unclear because these investigators also found that arp6 reduces the expression of MIR172B, when the expected effect of a reduction in miR156 is an increase in MIR172B expression (Wu et al., 2009a). Whether ARP6 contributes to the downregulation of MIR156A during vegetative phase change was not investigated in this study.

To address these issues and explore the mechanism by which H2A.Z regulates gene expression in Arabidopsis, we compared the effect of mutations in components of SWR1-C on the expression pattern and the chromatin structure of MIR156A and MIR156C. We found that H2A.Z contributes to the high expression of MIR156A and MIR156C early in shoot development, but does not play a major role in their downregulation during vegetative phase change. Our results also suggest that H2A.Z promotes the expression of MIR156A and MIR156C by promoting the deposition of H3K4me3, rather than through an effect on nucleosome occupancy.

In a screen for mutations that accelerate vegetative phase change, we isolated seven alleles of EARLY IN SHORT DAYS (esd1-3 to esd1-9), also known as ACTIN-RELATED PROTEIN 6 (ARP6) (Martin-Trillo et al., 2006). These mutants produce fewer juvenile leaves than normal, and flower early in both long days (LDs) and short days (SDs). The early flowering phenotype of arp6 mutants is the result of the reduced expression of the floral repressor FLOWERING LOCUS C (FLC) and of genes in the FLC-related MADS-affecting flowering (MAF) gene family (Choi et al., 2005, 2007; Deal et al., 2005, 2007; Lázaro et al., 2008; Martin-Trillo et al., 2006). However, the mechanism by which arp6 affects vegetative phase change – which is regulated independently of FLC (Willmann and Poethig, 2011) – remains unknown.

ARP6 interacts with SERRATED AND EARLY FLOWERING (SEF) to facilitate the exchange of histone variant H2A.Z for H2A (Choi et al., 2007; March-Díaz et al., 2007). To determine whether ARP6 affects vegetative phase change through its effect on H2A.Z, we compared the vegetative phenotype of arp6 and sef mutants with plants deficient for H2A.Z. In Arabidopsis, H2A.Z is encoded by HTA8, HTA9 and HTA11 (Coleman-Derr and Zilberman, 2012; Redon et al., 2002; Yi et al., 2006). The hta9 hta11 double mutant is morphologically and physiologically similar to arp6 mutants (Cheng et al., 2013; Coleman-Derr and Zilberman, 2012; Kumar and Wigge, 2010; March-Díaz et al., 2008), so we focused on this genotype. We compared the phenotype of these mutants with plants doubly mutant for mir156a-2 and mir156c-1 to determine the extent to which their vegetative phenotype may be explained by a decrease in the level of miR156. The alleles used for this analysis are null alleles and have been extensively characterized by other investigators (Cheng et al., 2013; Coleman-Derr and Zilberman, 2012; Deal et al., 2005; Kumar and Wigge, 2010; March-Díaz et al., 2008; Yang et al., 2013).

In Arabidopsis, juvenile rosette leaves are round with smooth margins and do not produce trichomes on their abaxial surface, whereas adult leaves are elongated, have a serrated margin, and produce trichomes on both sides of the leaf. Leaf 3 and leaf 4 of arp6-1, arp6-2, sef-2 and hta9-1 hta11-2 were more elongated and serrated than the corresponding leaves in Col under both long-day (LD) and short-day (SD) conditions (Fig. 1A,B). arp6-1, arp6-2, sef-2 and hta9-1 hta11-2 also accelerated the production of abaxial trichomes (Fig. 1C). In these respects, these mutants were similar to the mir156a-2 mir156c-1 double mutant (Fig. 1A,B). This result demonstrates that H2A.Z promotes the expression of the juvenile vegetative phase, and suggests that it may do so by promoting the expression of MIR156A and/or MIR156C.

To determine whether SWR1-C regulates vegetative phase change by affecting the expression of miR156, we measured the abundance of miR156 in shoot apices of wild-type Col and arp6-1 (hereafter arp6), and sef-2 (hereafter sef) mutants grown in short days (SD). Shoot apices were harvested at time points corresponding to the juvenile (L1+), juvenile and transition (L3+, L5+) and adult (L7+) phases of shoot development. The largest leaf primordium in these samples (leaf 1, 3, 5 or 7, respectively) was 2-3 mm. RT-qPCR revealed that arp6 and sef had significantly lower than normal levels of miR156 at the earliest juvenile stage (L1+), but not at later stages of shoot development (Fig. 2A). We then examined the effect of these mutations on the primary transcripts of MIR156A (pri-miR156a) and MIR156C (pri-miR156c). Both transcripts were present at 59-67% of the wild-type (Col) level in the shoot apex of 1-week-old arp6 and sef seedlings (L1+ stage) (Fig. 2B). This decrease was associated with a 1.3-to 2.1-fold increase in the transcripts of two direct targets of miR156, SPL9 and SPL15 (Fig. 2C). To determine whether the effect of arp6 on SPL9 expression is attributable to its effect on miR156, we crossed miR156-sensitive (pSPL9::SPL9-GUS) and miR156-resistant (pSPL9::rSPL9-GUS) reporters into an arp6 background. arp6 increased the GUS activity of the miR156-sensitive reporter by twofold, but had no effect on the activity of the miR156-resistant reporter (Fig. 2D,E). This result demonstrates that the effect of arp6 on SPL9 requires miR156 activity, and suggests that arp6 increases the expression of SPL9 by reducing the abundant miR156, rather than by directly promoting SPL9 transcription. To determine whether the effect of arp6 and sef on MIR156A and MIR156C expression is attributable to their role in the deposition of H2A.Z, we examined the abundance of pri-miR156a and pri-miR156c in the H2A.Z mutants hta9-1, hta11-2 and hta9-1 hta11-2. Individually, hta9-1 and hta11-2 did not produce a significant decrease in the level of these transcripts; however, hta9 hta11 mutants had ∼60% of the wild-type level of both pri-miR156a and pri-miR156c (Fig. 2F), which is approximately the same amount as in arp6 and sef (Fig. 2B). These results support the hypothesis that the effect of arp6 and sef on the expression of MIR156A and MIR156C is attributable to the loss of H2A.Z at these loci.

To test this hypothesis, we measured the level of H2A.Z at MIR156A and MIR156C in Col and arp6 using chromatin immunoprecipitation (ChIP). Chromatin was isolated from 1-week-old seedlings and was immunoprecipitated with an antibody to H2A.Z; sites from across the promoter and transcribed regions of these genes were then amplified by PCR. H2A.Z was enriched in the first 500 nt after the transcription start site of both genes, but was relatively low in the promoter and towards the 3′ end of these genes (Fig. 3A,B). arp6 seedlings had significantly reduced levels of H2A.Z at MIR156A and MIR156C, supporting the hypothesis that its effect on their expression is attributable to the loss of H2A.Z. We then asked if H2A.Z contributes to the downregulation of MIR156A and MIR156C during development by measuring the abundance of H2A.Z at these genes at 1, 2 and 3 weeks after planting. There was no significant change in the level of H2A.Z at these genes over this time period (Fig. 3C,D). The abundance of ARP6 and SEF transcripts also did not change significantly during the first five weeks of growth (Fig. S1). Thus, variation in the abundance of H2A.Z or SWR1-C does not account for the major decrease in the expression of MIR156A and MIR156C during vegetative phase change.

To determine how SWR1-C promotes the expression of MIR156A and MIR156C we examined the effect of arp6 on the active chromatin mark H3K4me3 and the repressive chromatin mark H3K27me3, and on nucleosome occupancy. We first examined the temporal pattern of H3K4me3 and H3K27me3 deposition in wild-type plants. Chromatin from 1-, 2-, 3- and 5-week-old plants was immunoprecipitated with antibodies to H3K4me3 or H3K27me3, and DNA from these samples was then amplified using primers for sites in the promoter and coding regions of MIR156A and MIR156C. The results were normalized to the results obtained using antibodies to H3. Consistent with previous results (Xu et al., 2016a), we observed low levels of H3K27me3 in the promoter and transcribed region of MIR156A/MIR156C at 1 week, and this mark increased gradually at subsequent time points; it increased more rapidly and to a higher level at MIR156A than at MIR156C (Fig. 4A,B). H3K4me3 was confined to the transcribed regions of MIR156A and MIR156C, and was most abundant in a 750 nt region downstream of the transcription start site (Fig. 4C,D). At MIR156A, H3K4me3 decreased gradually from 1 to 5 weeks, whereas at MIR156C, it started to decline 2 weeks after planting and continued to decline thereafter. In summary, at MIR156A the abundance of H3K27me3 and H3K4me3 changed in a complementary fashion, whereas at MIR156C an increase in H3K27me3 preceded a decrease in H3K4me3.

arp6 reduced the level of H3K4me3 at both MIR156A and MIR156C in 1-week-old seedlings (Fig. 5A,B), and produced elevated levels of H3K27me3 at MIR156A (Fig. 5C). However, it had no effect on the level of H3K27me3 at MIR156C (Fig. 5D). To determine the effect of arp6 on nucleosome occupancy, we examined the sensitivity of MIR156A and MIR156C chromatin to micrococcal nuclease (MNase), which preferentially cleaves naked DNA. Chromatin from 1-week-old arp6 and wild-type seedlings was treated with MNase, and sites in a 600 bp region surrounding the transcription start sites of MIR156A and MIR156C were then assayed by qPCR. arp6 increased MNase sensitivity at the site of the +1 nucleosome in MIR156A (∼50-100 nt), but had no effect on MNase sensitivity upstream or downstream of this site (Fig. 5E). It had no effect on MNase sensitivity at MIR156C (Fig. 5F). These results suggest that H2A.Z promotes the expression of MIR156A and MIR156C primarily by promoting the deposition of H3K4me3, rather than by blocking the deposition of H3K27me3 or destabilizing the +1 nucleosome.

The observation that arp6 reduces H3K4me3 at both MIR156A and MIR156C (Fig. 5A,B), but only increases H3K27me3 at MIR156A (Fig. 5C,D) prompted us to study how arp6 affects other genes that have both of these modifications. We analyzed five genes – FLC, FT, APETALA1 (AP1), NOZZLE (NZZ) and MALE STERILITY1 (MS1) – using ACT7 and STM as positive controls for H3K4me3 and H3K27me3, respectively (Berr et al., 2010, 2015; Jiang et al., 2008; Saleh et al., 2007, 2008; Shafiq et al., 2014; Tamada et al., 2009; Yang et al., 2014; Yun et al., 2012). ChIP of extracts from 1-week-old seedlings, followed by qPCR of a site close to the transcription start site (TSS) and a site downstream of the TSS (Fig. 6A) showed that arp6 significantly reduced H3K4me3 at most of these genes (Fig. 6B), but barely had any effect on H3K27me3 (Fig. 6C). Consistent with previous studies (Saleh et al., 2008; Tamada et al., 2009), the decrease in H3K4me3 at FLC was associated with significant decrease in its expression (Fig. 6D). arp6 had no significant effect on the expression of AP1, NZZ and MS1 (Fig. 6D), but this is not surprising because these genes are involved in floral morphogenesis and are expressed at extremely low levels in the 1-week-old seedlings used for this analysis; in the absence of the transcription factors necessary for the expression of these genes, a decrease in H3K4me3 is unlikely to have an effect on their expression. The expression of FT was upregulated in arp6, probably because of the downregulation of FLC, which is a strong repressor of FT (Helliwell et al., 2006; Michaels et al., 2005). To determine whether arp6 has a global effect on H3K4me3 and H3K27me3, we measured the abundance of these marks in Col and mutant seedlings using western blots. We observed no significant difference between these genotypes (Fig. 6E), suggesting that ARP6 only affects the deposition of H3K4me3 and H3K27me3 at some of the genes that possess these marks.

To determine whether the effect of SWR1-C on vegetative phase change can be explained by its effect on H3K4me3, we examined the phenotypes and genetic interactions between arp6, sef and mutations in H3K4 methyl transferases. Phylogenetic analysis demonstrates that H3K4 methyl transferases in Arabidopsis can be classified into two groups, Trithorax genes and SET1 genes (Avramova, 2009). ARABIDOPSIS TRITHORAX1 (ATX1) and ATX2 are members of the Trithorax family, and work together to promote the expression of floral homeotic genes (Alvarez-Venegas et al., 2003; Carles and Fletcher, 2009) and the expression of FLC (Pien et al., 2008; Tamada et al., 2009). ATXR7 is the only member of the SET1 subfamily in Arabidopsis, and functions along with ATX1 and ATX2 to promote the expression of FLC (Tamada et al., 2009).

The vegetative phenotypes of the single mutants atx1-1, atx2-1, atxr7-1, arp6 and sef-2, and the phenotypes of plants bearing combinations of these mutations, were compared under both LD and SD conditions. atx1-1 (hereafter, atx1), atx2-1 (hereafter, atx2) and the atx1 atx2 double mutant had no significant effect on leaf shape or on the timing of abaxial trichome production under both LD and SD (Fig. S2). In addition, neither mutation significantly enhanced the early abaxial trichome phenotype of arp6 or sef in double mutants (Fig. S2C-F). The atx1 atx2 arp6 and the atx1 atx2 sef triple mutants were embryonic lethal, making it impossible to determine the effect of these genotypes on vegetative phase change.

atxr7-1 (hereafter, atxr7), did not have an obvious effect on leaf shape (Fig. 7A), but produced leaves with abaxial trichomes approximately one leaf earlier than wild type; in this respect, it had approximately the same phenotype as arp6 and sef (Fig. 7B). The double mutants arp6 atxr7 and sef atxr7 had more elongated and serrated leaves than the single mutants (Fig. 7A), and also produced abaxial trichomes significantly earlier than the single mutants (Fig. 7B). By comparison, the timing of abaxial trichome production in the atx1 atx2 atxr7 triple mutant was not significantly different from atxr7, although this triple mutant had significantly more elongated leaves than atxr7 (Fig. S2A,B). RT-qPCR analysis of 1-week old seedlings revealed that arp6, sef and atxr7 had reduced levels of pri-miR156a and pri-miR156c, whereas arp6 atxr7 had significantly lower levels of pri-miR156a than the single mutants (Fig. 7D,E). However, the level of pri-miR156c in arp6 atxr7 was not significantly different from arp6.

To determine how ATXR7 regulates the expression of MIR156A and MIR156C, we used ChIP to measure the level of H3K4me3 and H3K27me3 in atxr7 and wild-type seedlings. atxr7 produced lower levels of H3K4me3 within the transcribed region of both MIR156A and MIR156C (Fig. 7F,G), but had a more reproducible effect at MIR156A than at MIR156C; it had no effect on the abundance of H3K27me3 at these loci (Fig. 7H,I). This result further demonstrates that the deposition of H3K4me3 and H3K27me3 is regulated independently at these genes. We then examined whether ATXR7 binds directly to MIR156A and MIR156C by taking advantage of a transgenic line expressing an ATXR7-GFP fusion protein under the regulation of the ATXR7 promoter (Tamada et al., 2009). ChIP revealed that ATXR7-GFP binds adjacent to the transcription start site of MIR156A (Fig. 8A), at a position and with an affinity that are nearly identical to that reported at FLC using the same transgenic line (Tamada et al., 2009). The results of our RT-qPCR analysis were considerably more variable for MIR156C than for MIR156A, and we did not observe significant binding of ATXR7 to MIR156C (Fig. 8B). This observation, and our observation that atxr7 has a smaller effect on the expression of MIR156C and the level of H3K4me3 at MIR156C than at MIR156A (Fig. 8D-G), indicates that ATXR7 is more important for the regulation of MIR156A than MIR156C, and suggests that ATXR7 may associate only transiently with MIR156C.

miR156 promotes juvenile development by repressing the expression of a group of SBP/SPL transcription factors that promote the expression of the adult phase (Wu et al., 2009a; Xu et al., 2016a). The transition from the juvenile to the adult phase of vegetative development occurs when miR156 levels decrease sufficiently to allow these transcription factors to be expressed. Understanding the mechanism of this decrease is therefore essential for understanding the mechanism of vegetative phase change. Previously, we have shown that MIR156A and MIR156C are the major sources of miR156 in Arabidopsis (Yang et al., 2013), and that these genes are transcriptionally downregulated during vegetative phase by a decrease in the active histone modification H3K27ac and an increase in the level of H3K27me3 (Xu et al., 2016a). Here, we show that H3K4me3 also promotes the elevated expression of MIR156A and MIR156C early in shoot development, and that deposition of this mark is promoted by the presence of H2A.Z at these genes.

The molecular basis for the effect of H2A.Z on gene expression has been difficult to decipher because this histone can have positive or negative effects on transcription. In Arabidopsis, H2A.Z stabilizes nucleosomes and represses transcription at genes involved in defense, drought and temperature responses (Coleman-Derr and Zilberman, 2012; March-Díaz et al., 2008; Sura et al., 2017), but destabilizes nucleosomes and promotes gene expression at the FLC locus (Choi et al., 2005, 2007; Deal et al., 2005, 2007; Lázaro et al., 2008; March-Díaz et al., 2007, 2008; Martin-Trillo et al., 2006), and at the thalional and merneral metabolic gene clusters (Nützmann and Osbourn, 2015). Although mutations that affect the deposition of H2A.Z can affect the expression of genes involved in developmental transitions and the response to various environmental conditions, these developmental or environmental factors do not necessarily produce changes in the abundance of this histone (Hu et al., 2011). These observations have led to the conclusion that H2A.Z sensitizes genes to the activity of other factors that regulate transcription, but does not by itself promote or repress gene expression (Deal et al., 2007; Kumar and Wigge, 2010). Our results support this hypothesis. Specifically, we found that H2A.Z levels do not change significantly at MIR156A and MIR156C during vegetative phase change, and that the temporal expression pattern of miR156 is not significantly affected by the loss of H2A.Z in arp6. The simplest interpretation of this observation is that H2A.Z sensitizes MIR156A and MIR156C to factors that promote transcription without inhibiting their accessibility to temporally regulated factors that repress their transcription. Although H2A.Z does not play a major role in the repression of MIR156A and MIR156C transcription during vegetative phase change, it may be important for the upregulation of miR156 expression during embryogenesis, when the abundance of this miRNA increases to very high levels (Nodine and Bartel, 2010). The reactivation of FLC expression during embryogenesis requires the activity of PIE1, a component of SWR1-C (Choi et al., 2009), and perhaps the reactivation of MIR156A and/or MIR156C expression during embryogenesis also depends on this complex.

A previous study of the effect of arp6 on miRNA gene expression concluded that ARP6 promotes the expression of MIR156A by reducing nucleosome occupancy at this locus (Choi et al., 2016). We confirmed the observation that arp6 increases nucleosome occupancy at MIR156A, but found that it had no effect on nucleosome occupancy at MIR156C. Similarly, arp6 increased the level of H3K27me3 at MIR156A, but had no effect on this mark at MIR156C. The only effect that was common to both genes was a reduction in the abundance of H3K4me3. We also found that arp6 reduced H3K4me3 at FLC, FT, AP1, NZZ and MS1. Importantly, the loss of H3K4me3 at MIR156C, FLC, FT, AP1, NZZ and MS1 was not accompanied by a change in the level of H3K27me3 at these genes. Furthermore, although arp6 reduced the transcription of MIR156A, MIR156C and FLC, it did not affect the transcription of AP1, NZZ and MS1 in the 1-week-old seedlings used for this analysis. AP1, NZZ and MS1 are involved in floral differentiation or gametophyte development (Berr et al., 2010; Mandel and Yanofsky, 1995; Wagner et al., 1999) and are expressed at very low levels in 1-week-old seedlings. It is therefore not surprising that the loss of H3K4me3 is not associated with an increase in H3K27me3 or further decrease in their expression at this stage. It remains to be determined whether arp6 affects their expression in the inflorescence and flowers, where they are normally expressed. Whatever the case, these observations suggest that ARP6 and, by extension, H2A.Z, directly promotes the deposition of H3K4me3, and that this is the primary mechanism by which it regulates the expression of MIR156A and MIR156C.

In mammalian embryonic stem cells, H2A.Z facilitates the binding of the H3K4 methyltransferase MLL, a homolog of the Trithorax gene in Drosophila (Creyghton et al., 2008; Hu et al., 2013). We found that mutations in the H3K4 methyltransferase gene ATXR7 decrease the level of H3K4me3 at MIR156A and MIR156C, and have a phenotype similar to arp6 and sef, making it an excellent candidate for an H2A.Z-interacting protein. Whether or not H2A.Z acts by affecting the accessibility of MIR156A and MIR156C to ATXR7, it is clear that ATXR7 is not the only H3K4 methyl transferase that operates at these loci because arp6 atxr7 and sef atxr7 double mutants have a more severe phase change phenotype than the single mutations. atx1 and atx2 did not have an obvious effect on phase change, but the atxr7 atx1 atx2 triple mutant had a stronger leaf shape phenotype than atxr7, suggesting that these TrX proteins may also operate at MIR156A and MIR156C. Although H2A.Z may facilitate the activity of these or other H3K4 methyltransferases, the evidence that H3K4me3 decreases at MIR156A and MIR156C in the absence of a change in H2A.Z or a change in the expression of ARP6 or SEF indicates that these factors are not responsible for developmental variation in the deposition of H3K4me3 at MIR156A and MIR156C.

The transcription of genes that possess the active histone modification H3K4me3 and the repressive modification H3K27me3 is thought to be regulated by the balance between these marks (Pien et al., 2008), but how the relative abundance of these modifications is regulated is still unknown. At FLC, for example, mutations in the H3K4 methyltransferases ATX1, ATXR3 and ATX7 decrease H3K4me3 and simultaneously increase the level of H3K27me3 (Pien et al., 2008; Tamada et al., 2009; Yun et al., 2012), suggesting that these marks are mutually exclusive. However, in vitro studies indicate that, whereas H3K4me3 can inhibit the binding of PRC2 to nucleosomes, this effect depends on the subunit composition of PRC2. Complexes containing EMF2 are incapable of binding nucleosomes containing H3K4me3, but complexes containing VRN2 – a paralog of EMF2 – are capable of binding such nucleosomes (Schmitges et al., 2011). If MIR156A and MIR156C are bound by different isoforms of PRC2, this may explain why arp6 decreased H3K4me3 at both MIR156A and MIR156C, but only increased H3K27me3 at MIR156A. The evidence that MIR156C expression is more significantly affected by mutations in the PRC2 component SWN than is MIR156A expression (Xu et al., 2016a) is consistent with the hypothesis that these loci are targeted by different forms of PRC2.

miR156 regulates many aspects of plant development, including the timing of vegetative phase change, flowering time, fruit development, the rate of leaf initiation, lateral root development, anthocyanin production, stress responses and shoot regeneration in tissue culture (Ferreira e Silva et al., 2014; Gou et al., 2011; Khan et al., 2014; Lei et al., 2016; Wang et al., 2009, 2008; Wu et al., 2009a; Wu and Poethig, 2006; Xing et al., 2010; Yu et al., 2015; Zhang et al., 2015). It is unknown whether all of these responses are regulated by MIR156A and MIR156C, or whether other members of this gene family are more important under certain conditions. In the context of the present study, this is an important issue because H2A.Z may have different functions at different MIR156 genes, perhaps mediated by the unique histone methylation signatures at these loci. Determining the role of this histone in the many processes regulated by miR156 will be an important subject for future research.

All of the stocks used in this study were in a Col background. arp6-1 (SAIL_599_G03), arp6-2 (SAIL_236_C07), sef-2 (SAIL_1142_C03), hta9-1 (SALK_054814C), hta11-2 (SALK_031471), atx1-2 (SALK_149002), atx2-1 (SALK_074806) and atxr7-1 (SALK_149692C) were obtained from ABRC. Primers for genotyping these mutants are listed in Table S1. mir156a-2 and mir156c-1 have been described previously (Yang et al., 2013). The ATXR7pro::ATXR7-GFP line was a gift from Dr Richard M. Amasino (University of Wisconsin, Madison, USA). Seeds were stratified at 4°C for 2 to 4 days and then transferred to 22°C at day 1. For phenotypic analysis, plants were grown under long days (16 h light:8 h dark) or short days (10 h light:14 h dark) conditions at 22°C, under a combination of cool white and Gro-Lite WS (Interelectric) fluorescent lights. All qPCR and ChIP analyses were performed with plants grown under short day conditions.

Shoot apices of plants at different developmental stages or from 1-week-old whole seedlings were used for qPCR, as indicated in the text. Total RNA was isolated using TRIzol (Invitrogen), followed by TURBO DNase (Ambion) treatment, according to the manufacturer's instructions. RNA (1.2 µg) was used in reverse transcription with SuperScriptIII Reverse Transcriptase (Invitrogen). RT-qPCR was performed on BIO-RAD CFX96 Real time System. Quantitative analysis of mature miR156 transcripts and mRNAs of protein-coding genes was performed as described previously (Xu et al., 2016a) using the primers listed in Table S1.

One-, 2-, 3- and 5-week-old Col shoot apices (0.5 g) were harvested for the analysis of H3K4me3 and H3K27me3 abundance at MIR156A and MIR156C. One-week-old seedlings were used for a comparison of H3K4me3 and H3K27me3 deposition at these genes in Col and arp6. Chromatin immunoprecipitation was performed as described previously (Xu et al., 2016a) and the results were analyzed using the PCR primers described in this paper. Antibodies to H3 and H3K27me3 were purchased from Abcam (ab1791) and Millipore (07-449), respectively. The antibody against H3K4me3 was obtained from EMD (04-745). The H2A.Z antibody was a gift from Roger Deal (Emory University, Atlanta, GA, USA). The abundance of H3K27me3, H3K4me3 and H2A.Z was calculated as the ratio of H3K27me3/H3, H3K4me3/H3 or H2A.Z/H3. The GFP antibody was purchased from Clontech (632592). The fold-enrichment was calculated as the percentage of input and normalized to TA3, and the fold in Col was set to 1. The data are presented as the average of three biological replicates.

MNase assays were performed using 0.8 g of tissue from 1-week-old seedlings, as previously described (Xu et al., 2016a). Briefly, nuclei were isolated and treated with 0.1 U/μl micrococcal nuclease (Clontech, 2910) for 3-5 min. Mononucleosome DNA was then gel-purified using a Fisher GenJET gel extraction Kit. DNA amplification and the calculation of nucleosome occupancy were performed as described previously (Xu et al., 2016a).

Nuclei from 1-week-old Col and arp6 seedlings were isolated according to Han and colleagues (2012). Nuclei proteins were released and separated by electrophoresis on a 15% SDS-PAGE gel. Immunoblots were performed using the anti-H3, anti-H3K4me3 and anti-H3K4me3 antibodies that were used for ChIP assays.

All statistical tests were conducted in R (R Core Team, 2015). The appropriateness of parametric tests were all confirmed by analyzing the distribution of the residuals and homogeneity of the variance from fitted linear models. In necessary cases, response variables were log transformed to meet parametric assumptions. Welch's t-test was used in cases where the assumption of equal variance was not met. Bonferroni adjustments were made on P values derived from two or more comparisons. The multcomp package (Hothorn et al., 2008) was used for Dunnett's test. The BSDA package (Arnholt and Evans, 2017) was used for nonparametric Sign-Tests.

We are grateful to Roger Deal for antibodies against H2A.Z and to members of the Poethig laboratory for helpful discussions.