Research Article

Timing Mechanism Dependent on Cell Division Is Invoked by Polycomb Eviction in Plant Stem Cells

See allHide authors and affiliations

Science  31 Jan 2014:
Vol. 343, Issue 6170, 1248559
DOI: 10.1126/science.1248559

Structured Abstract


In plants, leaves and flowers originate from the shoot apical meristem. In an indeterminate shoot apical meristem, stem cells persist for the life of the plant. In a determinate meristem, a certain number of organs are produced before the meristem is terminated; this characterizes the floral meristem derived from the shoot apical meristem. In Arabidopsis, stem cell identity is sustained by expression of the gene WUSCHEL. Expression of WUSCHEL can be terminated by the zinc finger protein KNUCKLES (KNU), with the result that stem cell identity is inactivated. KNU expression is induced by the floral homeotic protein AGAMOUS (AG), but that induction process requires ~2 days and invokes modification of histones resident at the KNU locus. Here, we show that the 2-day time lag is a consequence of a regulated molecular mechanism and that this mechanism can be embedded in a synthetic regulatory system to invoke a similar time lag.

Embedded Image

Induction of KNU in Arabidopsis floral meristems. Synchronized inflorescences imaged by confocal microscopy and reconstructed into three-dimensional projections (red stains by a fluorescence dye show the shapes of developing flowers). In an Arabidopsis line that has been engineered so that its floral development is both inducible and synchronized, KNU expression (green) begins 1 to 3 days after the activation of flower development. The delay is mediated by repressive histone methylation at the KNU locus. Upon activation, the transcription factor AG displaces Polycomb proteins, and the repressive histone marks are lost with cell cycle progression. Scale bar, 100 μm.


For transgenic Arabidopsis plants, we accelerated or inhibited cell cycles with pharmacological agents and studied the resulting KNU expression in response to AG induction. We used chromatin immunoprecipitation to study the presence of Polycomb proteins on the KNU locus at specific times during flower development. We used insertional mutagenesis to alter the function of the Polycomb response element (PRE) and analyzed the response from a heterologous promoter in Arabidopsis cell cultures. We constructed a synthetic mimic in Arabidopsis floral buds of the AG function by using the DNA binding domain of the lactose operon repressor (lacI) with its cognate binding sites. To test the logic that the delay in downstream gene induction was caused by the need to evict Polycomb group (PcG) proteins from their residence, we simulated the competition between PcG proteins and DNA binding proteins by using lacI, designed transcription activator–like effector DNA binding proteins, and synthesized promoters in Arabidopsis cell lines.


AG induces KNU with a time delay regulated by epigenetic modification. In wild-type plants, KNU expression begins in the center of the floral meristem and follows cell cycle progression. The binding sites for AG in the KNU upstream region are located within the PRE sequences required for the repressive histone modification. Binding of AG displaces PcG proteins, leading to the failure to maintain the repressive histone methylation. The combination of lacI operator sequences with a chimeric protein that contained the lacI DNA binding domain but lacked the activation domain was able to mimic the AG activity in Arabidopsis floral buds. We also reconstituted the cell division–dependent delayed-induction circuit in cell lines.


Our results indicate that flower development in Arabidopsis employs cell division to provide stem cells with a window of opportunity to change fate. The competition we observed between repressive PcG proteins and an activating transcription factor may reflect a general mechanism. The logic of the molecular circuit we have uncovered here may impose timing control on diverse growth and differentiation pathways in plants and animals.

A Matter of Timing

Plants flower only when their developmental programs give the go-ahead; otherwise floral genes remain repressed. Sun et al. (10.1126/science.1248559; see the Perspective by Zhang) analyzed the regulatory program that controls expression of the transcription factor KNUCKLES (KNU), which is required in the control of floral genes. KNU expression was silenced by the presence of Polycomb group (PcG) proteins. The floral homeotic protein AGAMOUS competed for control of KNU and activated its expression, but with a 2-day lag time. Thus, eviction of PcG by activating DNA binding proteins can insert a lag time before a switch in gene expression takes place.


Plant floral stem cells divide a limited number of times before they stop and terminally differentiate, but the mechanisms that control this timing remain unclear. The precise temporal induction of the Arabidopsis zinc finger repressor KNUCKLES (KNU) is essential for the coordinated growth and differentiation of floral stem cells. We identify an epigenetic mechanism in which the floral homeotic protein AGAMOUS (AG) induces KNU at ~2 days of delay. AG binding sites colocalize with a Polycomb response element in the KNU upstream region. AG binding to the KNU promoter causes the eviction of the Polycomb group proteins from the locus, leading to cell division–dependent induction. These analyses demonstrate that floral stem cells measure developmental timing by a division-dependent epigenetic timer triggered by Polycomb eviction.

Multicellular developmental processes require the precise coordination of growth and differentiation. The growth and development of the aerial part of the plant depends on the continuous activity of stem cells that are maintained at the growing tips called the shoot apical meristem (SAM) (1). After floral transition, the SAM usually becomes an inflorescence meristem (IM) that generates floral meristems (FMs). In contrast to the continuous SAM and IM, the FM gives rise to a certain number of organs and loses its pool of stem cells during differentiation. Thus, in the vast majority of land plants, FMs are considered determinate meristems. However, the mechanisms that control when the cells stop dividing and terminally differentiate remain largely mysterious. In Arabidopsis, the homeodomain protein WUSCHEL (WUS) is expressed in a small group of cells at the center of the meristems (SAMs, IMs, and FMs) and is essential for maintenance of the stem cell pool (2). WUS expression is repressed by the C2H2-type zinc finger protein KNUCKLES (KNU) in the FM (3, 4). The floral homeotic protein AGAMOUS (AG), which is directly induced by WUS in young floral buds, directly induces KNU ~2 days later to terminate FMs at the precise timing (Fig. 1, A and B) (48). Delayed or precocious KNU expression leads to the formation of extra or fewer organs, respectively (4). Here, we study the molecular mechanisms of these time-regulated delays in developmental programs.

Fig. 1 KNU induction timing is cell division–dependent.

(A) Confocal observations of the inflorescence doubly transgenic for pAG::GFP (blue) (8) and pKNU::KNU-VENUS (green), which rescues knu-1. *, SAM. Numbers, floral stages (39). (Inset) A higher magnification of the KNU-VENUS signal in a stage 5 to 6 floral bud. (B) Schematic diagram showing the developmental expression patterns of AG, KNU, and WUS in different floral stages. (C to F) GUS staining in ag-1 35S::AG-GR pKNU::KNU-GUS flowers at 0, 2, 3, and 4 days (C and D) or 0, 24, 36, and 48 hours (E and F) after treatment with 10 μM DEX alone (C and E), 10 μM DEX and 100 μM olomoucine (OLO) (D), or 10 μM DEX and 50 μM gibberellic acid 3 (GA) (F). For better penetration of the cell cycle inhibitors and phytohormones, we pretreated the lines with these compounds twice in total, 1 day before the DEX treatment and again in combination with DEX. (G to I) Confocal observation of ap1 cal 35S::AP1-GR pKNU::KNU-VENUS, 2 days after treatment with 1 μM DEX (G), 1 μM DEX and 100 μM OLO (H), or 1 μM DEX and 50 μM GA (I). Cells were stained with FM4-64 dye in (A) and (G) to (I). Scale bars, 50 μm (A and inset), 100 μm (C and D), 200 μm (E and F), and 20 μm (G to I).

KNU Induction Is Cell Division–Dependent

To examine the basis of the 2-day delay between AG and KNU induction, we used ag-1 35S::AG-GR pKNU::KNU-GUS, a KNU reporter line that expresses an AG construct that can be induced posttranslationally with dexamethasone (DEX) in the ag-1 mutant (4). Roscovitine and olomoucine (cyclin-dependent kinase inhibitors) block cell cycle progression at the G1-S and G2-M phases, whereas aphidicolin (an inhibitor of DNA polymerase) blocks the cell cycle at the early S phase (9). Whereas the KNU-GUS reporter was induced ~2 days after DEX treatment in the AG inducible line, addition of cell cycle inhibitors prevented the KNU induction on day 2 (Fig. 1, C and D, and fig. S1). AG also induces SPOROCYTELESS (SPL/NOZZLE) at about the same developmental stage when KNU starts to express (10). However, SPL reporter expression was unaffected by olomoucine treatment (which had the strongest effect to KNU) in the AG inducible line (figs. S1 and S2). Thus, division-dependent time lag is required for AG’s induction of KNU but not of SPL.

Fig. 2 Polycomb group protein binding on KNU and cis activities.

(A) Schematic diagram of the KNU locus and primer sets P1 to P6 used for the ChIP assays in (B) and (C). (B and C) ChIP assays using ap1 cal 35S::AP1-GR pFIE::FIE-VENUS (B) and ag-1 ap1 cal 35S::AP1-GR pFIE::FIE-VENUS (C) inflorescences harvested 0, 1, and 2 days after a single 1 μM DEX treatment. Nuclear protein complexes were immunoprecipitated with an anti-GFP antibody, and the enriched DNA was used for quantitative PCR analysis. The y axis shows the relative enrichment with immunoglobulin G (IgG) as a control. The error bars represent SD based on three biological replicates. ACTIN (ACT) was used as a control gene for calibration. (D) Sequences of the KNU promoter, encompassing the P2 region from −1048 to −810 from the transcription start site (+1), which is bound by AG and PcG proteins. The 6-bp fragment (CATATG) was inserted at the position −966/−965 on the KNU promoter (red arrowhead). The three AG half-perfect binding consensus sequences are underlined (the full-length sequences are marked by thin lines). The boxed sequences show the target site for the TAL designer protein shown in Fig. 4. (E to H) GUS staining of the wild-type pKNU::KNU-GUS (E). The 6-bp insertion at the position −966/−965 caused ectopic and precocious KNU expression (shown by #) (F), whereas the same insertion at the −976/−975 [a small black arrowhead in (D)], 10 bp upstream (G) and −1066/−1065, 100 bp upstream sites (H) showed the same KNU expression pattern as the wild-type pKNU::KNU-GUS plants. *, IMs. Scale bar, 100 μm.

Phytohormones gibberellin and cytokinin, which accelerate cell cycle progression (11), increased the expression of a cyclin reporter construct in developing flowers (11) (fig. S3, A to C). In the inducible line of AG, treatment with these phytohormones alone did not induce the KNU reporter (fig. S3, D to F), but AG induction (through DEX) combined with phytohormone treatment showed precocious KNU-GUS activity within 24 to 36 hours (Fig. 1, E and F, and fig. S4).

To visualize KNU expression at the cellular level, we established a KNU fluorescence reporter in the inducible line of flower development, ap1 cal 35S::AP1-GR pKNU::KNU-VENUS (12), which enabled us to synchronize flower development of multiple floral buds. After the induction of flower development, KNU started to be expressed only in a limited number of cells with a time window of 1 and 3 days (fig. S5). The treatment with olomoucine or gibberellin to this line delayed or accelerated the induction timing of KNU, respectively, in a way that correlated with cell cycle progression (fig. S6). Cellular-level observation showed that KNU was induced in more cells by gibberellin treatment than the control on days 2 and 3, whereas no (on day 2) or fewer cells (on day 3) showed KNU induction when treated with olomoucine (Fig. 1, G to I, and fig. S7). Although we did not see any obvious phenotypic differences by the cell cycle manipulation until day 2, olomoucine and gibberellin delayed or accelerated the emergence of floral organ primordia on day 3, suggesting that cell division plays an important role in morphological changes in flower development (fig. S8; see supplementary text for details). The cell division of floral stem cells is not synchronized and occurs once in 18 to 36 hours in most of the cells (13). Together, these data indicate that KNU regulation may be controlled by an intrinsic cell division–based timer.

AG Displaces Polycomb Group Proteins from KNU

The KNU transcribed region carries the repressive histone modification trimethylation of lysine 27 of histone H3 (H3K27me3), which is established and maintained by Polycomb group (PcG) proteins (14, 15). KNU activation by AG is associated with the loss of H3K27me3 (4). In PcG mutants, KNU is precociously expressed, and so is the KNU reporter line with a deletion in the KNU coding region carrying H3K27me3, suggesting that the mark has the commanding role for temporal regulation (4). Furthermore, the reduction of the H3K27me3 levels is AG-dependent (4), and the above-mentioned truncated reporter is expressed in ag-1 (fig. S9). Thus, AG appears to temporal-specifically remove H3K27me3. To examine the mechanism underlying how AG affects the repressive marks on KNU in a cell division–dependent manner, we created tagged lines for the essential components of PcG, the ESC homolog FERTILIZATION-INDEPENT ENDOSPERM (FIE) and the structural homolog of Su(z)12, EMBRYONIC FLOWER2 (EMF2) (16, 17). We first transformed the pFIE::FIE-VENUS and pEMF2::EMF2-VENUS constructs into the fie-11/+ and emf2-1/+ heterozygous mutant lines, respectively, and obtained the rescued lines in the homozygous background for each mutation. Next, we introduced them into ap1 cal 35S::AP1-GR (12), which enabled us to harvest stage-specific floral buds (fig. S10, A to C). After one time of DEX treatment, floral buds were harvested to perform chromatin immunoprecipitation (ChIP) assays for FIE, EMF2, and AG [with an AG antibody (4)] at days 0, 1, and 2. We detected AG binding in the samples on days 1 and 2 at a region ~900 base pairs (bp) from the transcriptional start site, but not on day 0 (4) (fig. S11). In contrast, we detected moderate binding of FIE and EMF2 around a wider region of the KNU locus, including the AG binding sites on day 0 (Fig. 2, A and B, and fig. S12, A and B). From day 1 when AG binding started to be detected, the relative binding levels of FIE and EMF2 decreased to the basal levels (Fig. 2B and figs. S11 and S12B). In the backgrounds containing the ag-1 mutation, we detected continuous binding of FIE and EMF2 (Fig. 2C and fig. S12C). These results suggest that AG binds to the upstream region of KNU in a competitive manner with PcG proteins, evicting them from the locus, which would lead to the loss of H3K27me3 at the coding region.

KNU Upstream Regions Contain a Polycomb Response Element

PcG proteins are recruited to a specific site of target gene loci and then mediate the spread of H3K27me3 (18, 19). Originally found in Drosophila, the Polycomb response element (PRE) is responsible for the recruitment of PcG to specific genes, possibly through DNA binding proteins or noncoding RNAs (14, 15). The antagonistic localization of AG and PcG proteins on the same upstream region of KNU indicates that the region is necessary for the initial recruitment of PcG and, thus, for the prevention of ectopic KNU expression. To examine the function of PcG binding, we performed insertional mutagenesis of the PcG binding region of the KNU-GUS reporter construct (Fig. 2, D to H). When we inserted 6-bp unrelated nucleotide sequences immediately upstream of the first AG half-binding consensus sequence, the KNU reporter was ectopically expressed in the IMs of most of the independent transgenic lines (Fig. 2F and table S1). However, we observed no effects by the insertion of the same 6-bp fragment in 10- or 100-bp upstream regions (Fig. 2, G and H, and table S1). These results indicate that the region immediately adjacent to the first AG binding site may contain a PRE-like activity, which was at least partially perturbed by the insertion.

To further characterize the PRE-like activity of the KNU locus, we cloned the KNU upstream region containing the AG binding sites into a heterologous promoter, pF3H, with ubiquitous activity (20) and transformed the resulting constructs into Arabidopsis cultured cells (Fig. 3A). The shortest 153-bp fragment containing three AG binding sites retained the ability to almost fully silence the reporter expression, as did a longer fragment in the KNU upstream and the positive control LEAFY COTYLEDON2 (LEC2) upstream region containing PRE-like activity (20) (Fig. 3, A to C, and fig. S13, A and B). High levels of H3K27me3 were deposited at the reporter coding region in transgenic lines containing the KNU 153-bp fragment but not in the negative control lines without any inserts (fig. S13C). These results suggest that the 153-bp fragment containing AG binding sites is sufficient to silence the heterologous promoter through the recruitment of PcG and deposition of H3K27me3 at the reporter coding region. Furthermore, the disruptive insertion, which caused the ectopic expression of the KNU reporter in plants, abolished the PRE activity in the cell line (Fig. 3, A and D).

Fig. 3 Polycomb response element and simulation of AG.

(A) Schematic diagram of the constructs to test the PRE activity of the fragment containing three AG binding sites (gray bars) on the KNU promoter and the disruptive insertion site that caused ectopic expression of the reporter in Fig. 2D (red arrowhead) and the summary of the assay in the Arabidopsis T87 culture cells. (B to D) T87 cells transgenic for the construct of pF3H::YFP alone (B) and the pF3H::YFP constructs conjugated with the 153-bp fragment, pF3H-KNU PRE 153 bp::YFP (C), and with the 153-bp fragment with the disruptive 6-bp insert, pF3H-KNU PRE 153 bp+6 bp::YFP (D). (E) Schematic diagram of the experiment to show that a chimeric protein with the lacI DNA binding domain (LacI) and glucocorticoid hormone binding domain (GR) partially mimics the function of AG to induce KNU expression by binding to the operator sequences (OP) upon DEX treatment and eviction of the PcG proteins. (F and G) GUS reporter staining 3 days after a single DEX treatment in T1 transgenic plants of 35S:: LacI-GR pKNU-OP::KNU-GUS treated with mock (F) and 10 μM DEX (G). (H to M) Time course observation of T2 35S::LacI-GR pKNU-OP::KNU-GUS flowers treated with mock (H to J) or DEX (K to M) harvested at days 0 (H and K), 0.5 (I and L), and 3 (J and M) after a single 10 μM DEX treatment. Scale bars, 25 μm (B to D), 200 μm (F and G), and 50 μm (H to M).

To test whether the KNU PRE activity is transcription-dependent, we deleted the core promoter from the original constructs, abolishing the reporter expression regardless of the KNU PRE (fig. S13, D and E). Then, we tested the accumulation of the repressive marks by ChIP assay. We detected PRE-dependent accumulation of H3K27me3 in the reporter coding region (fig. S13F). This result indicates that the KNU PRE may have a transcription-independent activity in this context to recruit PcG and deposit the repressive marks.

Mimicking AG Functions by an Artificial Protein

The KNU PRE contains AG binding sites, and the binding of PcG and AG are complementary in flower development (Figs. 2, A to D, and 3A and figs. S11 and S12), leading to the AG-dependent reduction of the H3K27me3 marks (4). The Trithorax group (TrxG) transcriptional activator complexes counteract PcG activity and mediate the active H3K4 methylation marks (14). We examined the active marks at KNU in flower development by ChIP assays using the ap1 cal 35S::AP1-GR inflorescences. The KNU coding region contains both the H3K4me2 and H3K4me3 marks, and our sequential ChIP assays with antibodies for H3K27me3 and H3K4me3 showed that the KNU locus contains the bivalent marks (fig. S14). During flower development, we noticed a progressive decrease in the H3K27me3 levels consistent with KNU activation (4), whereas the H3K4me2 and H3K4me3 levels remained unchanged (fig. S14, B and C). Therefore, AG does not affect H3K4me2 and H3K4me3, but H3K27me3. The simplest explanation for these results may be that the AG functions to physically block the binding of the PcG proteins from PRE, causing the dilution of the epigenetic repressive status through cell division. A similar sustained silencing by PcG has been suggested in assays of the Drosophila heat shock–induced clonal analysis of PcG (21, 22). Deletion of Enhancer of Zeste showed 2 to 3 days of delay before HOX misexpression was detected (22). To further examine the nature of the competitive binding between AG and PcG on the KNU locus, we tried to mimic the behavior of AG in vivo using an artificial protein (LacI-GR), which has the lactose operon repressor (lacI) DNA binding domain and a glucocorticoid hormone binding domain but no detectable transactivation activity (2325) (Fig. 3E). In particular, two of the three AG half-binding consensus sequences on the KNU promoter were replaced by two lac operator (op) sequences, and the remaining one was mutated to prevent AG binding (Fig. 3E). We detected a clear DEX-dependent induction of KNU expression with a time lag of 3 days in most of the primary T1 transgenic plants and in selected T2 plants (Fig. 3, F to M, and tables S2 and S3). We noticed a weak basal level of expression without DEX treatment (Fig. 3, F and H to J), which may be caused by the weaker PRE-like activity due to the mutagenesis and/or the leaky effect of 35S::LacI-GR. We further detected high levels of H3K27me3 on the coding region of the GUS reporter gene before the DEX treatment, and this level dropped on day 3 after the treatment (fig. S15). These results suggest that binding of the chimeric protein on the 2× op sequences can block the PcG complex binding and enable the expression of KNU through the removal of the repressive marks, which mimics the endogenous function of AG in KNU induction.

Creation of Cell Division–Dependent Epigenetic Timers

To test the logic of the timed induction of KNU based on the eviction of PcG proteins, we decided to reconstitute the “epigenetic timer” by simplifying the systems. We created a synthetic promoter combining the ubiquitous pF3H promoter and the short 50-bp PRE-like element of LEC2 (20) (Fig. 3A) conjugated with two copies of op sequences in the 5′ and 3′ ends (fig. S16A). This chimeric reporter was silenced because of the PRE-like activity, but when the LacI-GR was induced, it should bind to the op sequences and physically interfere with factors binding to the PRE-like element, leading to their eviction and the subsequent induction of the reporter. After the DEX treatment, the fluorescence reporter was induced by 18 hours in the culture cells (fig. S16B). The reporter was resilenced in 24 hours after a single DEX treatment. When the cells were pretreated with olomoucine, the induction was delayed and observed by 24 hours (fig. S16C).

We further tested the inducibility of the silenced reporter containing the KNU PRE (Fig. 3, A to C) by cotransforming the AG protein, or a TAL (transcription activator–like) effector–based synthetic DNA binding protein (26), that is designed to recognize the sequences around the first AG binding site. Overexpression of AG by 35S::AG-GR in the cultured cells could not activate the reporter (fig. S17), possibly because of the low binding affinity of the ectopically expressed AG to the KNU promoter. We then created and assayed the specific binding of the designer TAL by fusing the TAL DNA binding domain with a transcriptional activation domain (AD) and nuclear localization signal (NLS) (26). The transient assay in the leaf protoplasts revealed the specific induction of the endogenous KNU, showing that the TAL protein can bind to the upstream target sequence of KNU in vivo (fig. S18). We next induced the fusion protein between the TAL DNA binding domain, NLS, and the GR domain (which has no transcriptional activation domain) in the cultured cell lines containing the silenced reporter by KNU PRE (Fig. 4A). After a single DEX treatment, the reporter started to be induced in 20 hours, reached higher levels in 40 hours, and was resilenced in 60 hours (Fig. 4B). The induction was clearly delayed by the olomoucine pretreatment and observed by 60 to 70 hours (Fig. 4C). Accordingly, the ChIP assay in these cells showed that H3K27me3 was reduced by 72 hours in the reporter coding region (fig. S19). When we fully blocked the cell cycle progression by the continuous olomoucine treatment, we did not observe the reporter induction (fig. S20). These results indicate that competitive eviction of PcG from PREs by DNA binding proteins leads to cell division–dependent delayed induction (Fig. 4D). The different timing in induction and resilencing in these two systems may be due to the different PRE-like elements and different affinities of LacI and TAL DNA binding domains to the target sites. In Drosophila flp–mediated recombination assay of PREs in different transgenes, excision of PREs from the reporter genes led to loss of silencing at various timing from 12 hours to 2 to 3 days of considerable longer delay (27, 28). In Arabidopsis, LEAFY induces AG concomitantly through the recruitment of chromatin-remodeling adenosine triphosphatase by antagonizing PcG activities (29). Thus, the displacement of PcG proteins by competitive transcription factor binding may be a general mechanism with different cell division dependence (see supplementary text for PcG and cell division). An interesting future subject would be to determine the factors controlling the duration of this epigenetic timer and link with cell cycle phase. This epigenetic timer can be used as a module to build time-responsive molecular circuits in synthetic biology.

Fig. 4 Synthetic epigenetic timer and eviction model.

(A) Schematic diagram of a synthetic epigenetic timer. pF3H-KNU PRE 153 bp::YFP was cotransformed with the 35S::TAL DNA binding domain-GR (TAL-GR) construct, which targets the region around the first AG binding site (boxed sequences in Fig. 2D) in the Arabidopsis T87 culture cells. (B and C) Time course confocal microscopy observation of the T87 cells transgenic for the constructs shown in (A). The cells were observed at 0, 10, 20, 40, 60, and 70 hours after a single 10 μM DEX treatment. The cells were pretreated with mock (B) or 50 μM olomoucine (OLO) (C) 1 day before the DEX treatment, and the inhibitor was washed away at the time of the DEX treatment. Scale bar, 25 μm. (D) Schematic diagram of the KNU regulation. Earlier than stage 3 (corresponding to day 0 of ap1 cal 35S::AP1-GR floral buds), the KNU locus is covered by the H3K27me3 repressive mark, which is maintained by PcG. The KNU transcript cannot be induced because of the repressive marks. At stage 3, the AG protein, which is induced by WUS (6, 7), directly binds to the KNU promoter competitively with PcG, leading to eviction of PcG from the KNU locus. The eviction of PcG leads to cell division–dependent loss of the repressive status of KNU and the timed induction. Subsequently, WUS transcription is repressed by KNU (4), and floral stem cell activity is terminated with perfect timing.

Materials and Methods

Plant Materials and Chemical Treatments

All plants had the Landsberg erecta (Ler) background and were grown at 22°C under continuous light. Plant photographs were taken with a stereomicroscope (Carl Zeiss MicroImaging GmbH). DEX (Sigma) treatments were conducted by inverting the plants and submerging the inflorescences for 1 min in a solution containing either 1 μM DEX (for 35S::AP1-GR) or 10 μM DEX (for 35S::AG-GR, 35S::LacI-GR and 35S::TAL-GR) together with 0.015% Silwet L-77. The time of the initial DEX treatment was taken as day 0 or 0 hour. Cell cycle inhibitors and phytohormones, applied at final concentrations of 50 μM for roscovitine, 100 μM for olomoucine, 35 μM for aphidicolin, 500 μM for cytokinin [benzylaminopurine (BAP)], and 50 μM for gibberellic acid 3 (GA), were treated one time 24 hours before the DEX treatment for better penetration, as well as in combination with DEX at 0 hour. The mock treatments were conducted using the same method with the dipping solution without the chemicals.

For cultured cells, 10 μM DEX was applied directly into the liquid cell culture medium at 0 hour. For cell cycle manipulation, the cells were pretreated with dimethyl sulfoxide for mock or 50 μM olomoucine 1 day before the DEX treatment. Olomoucine was washed away at the time of DEX treatment or kept in the cultured medium for continuous treatment.

GUS Staining

GUS staining was performed as previously described (30) and observed in whole-mount samples in clearing solution with a Carl Zeiss MicroImaging stereomicroscope, or in paraffin sections with a Carl Zeiss Axioplan 2 microscope.

ChIP Assay

The ChIP experiments were performed as previously described (31) with slight modification. Inflorescences from ap1 cal 35S::AP1-GR were ground in liquid nitrogen and postfixed with 1% formaldehyde for 10 min. The chromatin was isolated and solubilized by sonication to generate DNA fragments with an average length of 400 bp. After incubation with salmon sperm DNA–protein A (for polyclonal antibody) or sperm DNA–protein G (for monoclonal antibody) agarose beads (Millipore), the solubilized chromatin was incubated overnight with anti-GFP (green fluorescent protein) antibody (Santa Cruz Biotechnology or Invitrogen), normal rabbit IgG (Santa Cruz Biotechnology), or anti-H3K27me3 antibody (Millipore or Abcam). The DNA fragments were recovered from the purified DNA-protein complexes and then used for enrichment tests by real-time polymerase chain reaction (PCR) analysis in triplicate. For the H3K27me3 ChIP, the ratio between the bound DNA after immunoprecipitation and the input DNA before immunoprecipitation was calculated for all the representative primer sets spanning the KNU genomic region, or for the primer sets on GUS or yellow fluorescent protein (YFP) reporters, and the ratios were plotted to show the relative changes in the levels of epigenetic marks. The relative enrichment for FIE and EMF2 proteins on the KNU locus was the ratio between the bound DNA after immunoprecipitation by anti-GFP over that by IgG. For sequential ChIP assay, cross-linked chromatin from the inflorescences was immunoprecipitated with monoclonal anti-H3K27me3 antibody (Abcam) as described above, except that chromatin was eluted in a solution of 30 mM dithiothreitol, 500 mM NaCl, and 0.1% SDS at 37°C (32). Eluted chromatin was subject to a second immunoprecipitation with monoclonal anti-H3K4me3 antibody (Abcam). In addition, a reversed sequential ChIP assay was performed in the sequence of application of anti-H3K4me3 antibody first, followed by immunoprecipitation with anti-H3K27me3 antibody. For all ChIP experiments, the primers for the Mu-like transposon or the ACT gene were included as negative controls.

Vector Construction and Transgenic Selection

1) pEMF2::EMF2-VENUS, pFIE::FIE-VENUS, and pKNU::KNU-VENUS (fig. S21A) were produced as follows: The EMF2 and FIE genomic regions were amplified from Col wild-type genomic DNA with the primer sets PSOKK131EMF2F/ PSOKK132EMF2R and PSOKK162FIEF/ PSOKK165FIER, respectively. The amplified EMF2 and FIE genomic fragments were TA-cloned into pENTR-D TOPO vectors (Invitrogen). Subsequently, an Sfo I restriction site was introduced immediately before the stop codons of EMF2 and FIE with the QuikChange Site-Directed Mutagenesis Kit (Stratagene). The VENUS fragment was generated by digestion of the vector pRS316-VENUS (33) with Sfo I. Then, the VENUS fragment flanked by Sfo I was introduced into the Sfo I sites in the above-mentioned EMF2 or FIE genomic region, or Sfo I–digested pKNU::KNU-GUS (4) to replace the GUS fragment, to generate pEMF2::EMF2-VENUS, pFIE::FIE-VENUS, or pKNU::KNU-VENUS. Through LR reaction by Gateway LR Clonase II (Invitrogen), pEMF2::EMF2-VENUS or pFIE::FIE-VENUS was exchanged from the pENTR-D TOPO backbone to the destination vector pKGW (Invitrogen). pKNU::KNU-VENUS was exchanged from the pCR8 backbone to the destination vector CD3-694, pEarleyGate 303 (4).

2) Three pKNU::KNU-GUS lines with the 6-bp insertion at three different positions (fig. S21B) were produced as follows: First, a 996-bp KNU promoter fragment (from −1249 to −253 bp) flanked by Nde I sites, which contains the three AG binding sites, was cloned with the primer set PSOXYF68-KNU-NdeI-FP/ PSOXYF69-KNU-NdeI-RP into pCR8 GW TOPO vector by TA cloning (Invitrogen). To test the positional effect of 6-bp insertion, PCR mutagenesis using the primer sets PSOXYF393/394, PSOXYF397/398, and PSOZM204/205 were performed to introduce one extra Nde I site at three different positions inside the above-mentioned 996-bp KNU promoter fragment and named pKNUM3 (−966/−965), pKNUM5 (−976/−975), and pKNUM7(−1066/−1065), respectively. The resulting 1002-bp promoter fragments were digested from pKNUM3, pKNUM5, and pKNUM7 by partial digestion with Nde I (to keep the internal Nde I site intact but not the flanking ones) and introduced into a pCR8-based modified pKNU::KNU-GUS vector (by Nde I) (4) to replace the specific endogenous KNU promoter region. Subsequently, the entire cassettes of pKNUM3::KNU-GUS, pKNUM5::KNU-GUS, and pKNUM7::KNU-GUS were recombined into the destination vector pKGW by the LR reaction (Invitrogen).

3) 35S::LacI-GR pKNU-OP::KNU-GUS (fig. S21C) was constructed as follows: To prevent AG binding, a 996-bp KNU promoter fragment (from −1249 to −253 bp) flanked by Nde I sites was cloned from pMutated KNU with all three AG half binding sites mutated (from part ii) (4) into pCRII GW/TOPO vector and named pCRII pMutated KNU NdeI. Two copies of the op sequences were introduced to replace the second and third mutated AG half binding sites on pCRII pMutated KNU NdeI by four rounds of PCR mutagenesis with the primer sets PSOXYF399-OP1-F/ PSOXYF400-OP1-R, PSOXYF401-OP2-F/ PSOXYF402-OP2-R, PSOZM8MOP3_F/ PSOZM9MOP3_R, and PSOZM10MOP4_F/ PSOZM11 MOP4-R. Subsequently, the Nde I–flanked fragment harboring two copies of the op sequences was digested with Nde I and introduced into an Nde I–digested pCR8 pKNU::KNU-GUS to form the pCR8-based pKNU-OP::KNU-GUS. The resulting construct was further mutated by PCR with PSOZM50 KNUGUS NotI F/ PSOZM51 KNUGUS NotI R to create a Not I site after the KNU 3′ untranslated region for further cloning. The 35S::LacI-GR fragment was obtained as follows: LacI-GR flanked by Sal I and Spe I was amplified by PCR with PSOZM12 LhGR_F/ PSOZM13 LhGR DB_R from vector pBJ36 (34). The LacI-GR was later introduced into the modified pGreen binary vector 0280 (35) with the same sites. The resulting pGreen-based product with a promoter and a terminator was further digested by Not I to obtain the Not I–flanked 35S::LacI-GR fragment. Subsequently, the fragment was introduced into the Not I–digested vector pCR8 pKNU-OP::KNU-GUS, resulting in the entry vector pCR8 35S::LacI-GR pKNU-OP::KNU-GUS. The whole cassette was then recombined into pEarlygate303 CD3-694 (36) by the LR reaction (Invitrogen).

4) pF3H::YFP and pF3H-RLE::YFP (RLE from LEC2) (fig. S21D) were constructed as follows: The YFP fragment and terminator (YFP CDS followed by the 3myc and CaMV ter fragments) was amplified by the primer set PSOZM161-YFP-F/ PSOZM133-CaMV-terR from p1002-35S-YFP-3myc-CaMV to generate the cassette 35S-YFP-3myc-CaMV. The resulting cassette was then digested with Asc I and introduced into the modified pENTR-D TOPO p1002 to create the p1002-YFP. The F3H fragment was amplified from Arabidopsis Ler genomic DNA by PSOZM153 F3HpF/ PSOZM154 F3Hp-35SR, which contains the sequences for the 35S core promoter. The fragment was then digested with Bgl II and Xho I and introduced into the p1002-YFP vector to create pF3H::YFP (fig. S21D), and F3H is followed by the 35S core promoter. pF3H-RLE::YFP (fig. S21D) was constructed by one round of PCR mutagenesis based on the template of pF3H::YFP with the primer set PSOZM159 F3H_iRLE F/ PSOZM160 F3H_iRLE R, resulting in the insertion of an RLE fragment (20) into pF3H::YFP to generate pF3H-RLE::YFP.

pF3H-KNU PRE 354bp::YFP and pF3H-KNU PRE 153bp::YFP were constructed based on pF3H::YFP as follows: To create pF3H-KNU PRE 354bp::YFP (fig. S21E), the 354-bp fragment flanked by Bgl II was first amplified with PSOZM142 KNU pro BglII F/ PSOZM206 KNU pro BglII R from pKNU::KNU-GUS (4). Then, the fragment was introduced into p1002 pF3H::YFP to generate pF3H-KNU PRE 354bp::YFP. To create pF3H-KNU PRE 153bp::YFP (fig. S21E), the 153-bp fragment flanked by Bgl II was first amplified with PSOZM142 KNU pro BglII F/ PSOZM143 KNU pro XhoI R from pKNU::KNU-GUS (4). Then, the fragment was introduced into p1002 pF3H::YFP to generate pF3H-KNU PRE 153bp::YFP.

5) pF3H-KNU PRE 153bp+6bp insert::YFP; pOp-RLE-Op-F3H::YFP 35S::Lac-GR (fig. S21F) were constructed as follows: For the pF3H-KNU PRE 153bp+6bp insert::YFP (fig. S21E), the 153-bp fragment was amplified from pKNUM7, which had an Nde I site at −1066/−1065. The fragment was introduced into p1002 pF3H::YFP to generate pF3H-KNU PRE 153bp+6bp insert::YFP. pOp-RLE-Op-F3H::YFP (fig. S21F) was created by two rounds of PCR mutagenesis with primer sets PSOZM175 F3H SopL F/ PSOZM176 F3H SopL R and PSOZM177 F3H SopR F/ PSOZM178 F3H SopR R sequentially to insert 1× op each into the 5′ and 3′ ends of the RLE sequences of pF3H-RLE::YFP to generate pOp-RLE-Op-F3H::YFP. 35S::LacI-GR was constructed as follows: First, the 35S::LacI-GR-CaMVter fragment was amplified by the primer set PSOZM187_35S_F/ PSOZM133CaMtR with pCR8 pKNU-Op::KNU-GUS 35S::LacI-GR-CaMVter as the template. Subsequently, the amplified fragment was digested with Asc I and introduced into pENTR-D TOPO p1002. The CaMV terminator was then replaced with the Nos terminator by mega-primer mutagenesis with the primer set PSOZM195_CaMV_NosF/ PSOZM196_CaMV_NosR, resulting in the construct p1002 35S::LacI-GR-Nos ter. Next, the pOp-RLE-Op-F3H::YFP fragment was amplified with the primer set PSOZM153_F3Hp_F/ PSOZM193_CaMV_ter_NotIR. The fragment was later introduced into p1002 35S::LacI-GR-Nos ter through Bgl II and Not I sites to generate pOp-RLE-Op-F3H::YFP 35S::LacI-GR (fig. S21F).

6) 35S::TAL-GR and 35S::TAL-AD (fig. S21G) were created as follows: The KNU promoter 996-bp fragment, −1249 to −253 bp with an Nde I site on both ends, containing the three AG half binding sites, was input into an online tool (TAL Effector-Nucleotide Targeter, TALE-NT; A 25-bp TALE sequence targeting the first AG half binding site was chosen as a good target site (Fig. 2D). TAL’C of pTAL1 vector (Addgene) contains NLSs and transactivation domain VP64 (AD). To create 35S::TAL-GR (fig. S21G), a GR fragment including a hemagglutinin (HA) tag (861 bp) was first amplified from pGreen0281 plasmid (35) with the primer set PSOWY66_talBglIINdeIGR/ PSOZM192_talAsc1HA. Subsequently, the amplified fragment was used as a primer to anneal to pTAL1 vector (Addgene), amplifying the whole plasmid, resulting in replacement of the 105-bp transactivation domain VP64 (AD) with GR to generate the pTAL1-NLS-GR. Next, the chosen 25-bp TALE sequence was created by golden gate reaction (Addgene) into the pTAL1-NLS-GR destination vector to generate pTAL1-TALE-NLS-GR, followed by exchanging the cassette of TALE-NLS-GR into binary vector pMDC32 by LR reaction (Invitrogen) to generate 35S::TAL-GR (fig. S21G). 35S::TAL-AD was constructed by using golden gate reaction (Addgene) of the chosen 25-bp TALE sequence into pTAL1 (Addgene) to generate pTAL1-25bpTALE-NLS-AD, followed by exchanging the cassette of 25bpTALE-NLS-AD into binary vector pMDC32 by LR reaction (Invitrogen) to generate 35S::TAL-AD (fig. S21G).

7) 35S core promoter deletion (Δcore promoter) (fig. S21H) was created as follows: Both pF3H::YFP and pF3H-KNU PRE 153bp::YFP constructs were mutated by the KAPA PCR mutagenesis kit to delete 10 bp in the 35S core promoter sequence using the PSOLS257-35sdel F/ PSOLS258-35sdelR primers. Subsequently, the transgenes were recombined into the destination vector KGW.

All clones from 1) to 7) (fig. S21) were fully sequenced for confirmation, and all the primers used for construction are listed in table S4. pEMF2::EMF2-VENUS and pFIE::FIE-VENUS (fig. S21A) were introduced into ap1 cal 35S::AP1-GR. All the GUS constructs from 2) and 3) (fig. S21, B and C) were introduced into wild-type Ler plants using the floral dipping method mediated by Agrobacterium. pEMF2::EMF2-VENUS and pFIE::FIE-VENUS T1 transgenic plants were selected on an MS solid medium plate with the antibiotic kanamycin at a concentration of 50 μg/ml. All the GUS reporter transgenic plants were selected on soil with the herbicide Basta 15 (Bayer; 0.2% of the commercial solution). Constructs from 4) to 7) (fig. S21, D to H) were introduced into T87 cultured cells, and transformation was performed as described below.

Arabidopsis Protoplast and Culture Cell Assay

The transient gene expression assay using Arabidopsis mesophyll protoplasts was performed as previously described (37). The stable Arabidopsis transgenic culture cells were established using T87 suspensions cells (38). Briefly, the cells were maintained in JPL3 medium under continuous illumination at 22°C with rotary shaking at 120 rpm as previously described (38). To generate transgenic lines, 2-week-old T87 cells were sieved through a 500-μm stainless mesh and resuspended in B5 medium supplemented with 1 μM 1-naphthaleneacetic acid and sucrose (30 g/liter). Cell suspension was cultured under continuous illumination at 22°C with shaking at 120 rpm for 1 day. Then, 2.5 μl of overnight cultured Agrobacterium transformed with the appropriate vectors was added to the cell suspension and cultured for a further 2 days. After cocultivation, the cell suspension was washed twice with 10 ml of JPL3 medium supplemented with carbenicillin (200 μg/ml) by centrifugation at 100g for 2 min. Finally, cells were resuspended in 3 ml of JPL3 medium and spread over a selection JPL3 agar plate supplemented with carbenicillin (250 μg/ml) and selection drugs. After 2 weeks of culture, drug-resistant calli were transferred to fresh selection medium and maintained by subculture fortnightly. Kanamycin (30 μg/ml) and/or hygromycin B (12 μg/ml) were used for single/cotransformation on a plate. Kanamycin (25 μg/ml) and/or hygromycin B (7.5 μg/ml) were used for single/cotransformation in liquid medium. For reverse transcription PCR and ChIP assays, we used 1- to 2-week-old transgenic T87 suspension lines and collected ~0.2 ml of dry weight pellet by centrifugation at 100g for 2 min. For expression assay of the reporter lines, confocal microscopy images were taken with a Leica SP5 (as described below).

Confocal Microscopy Imaging

For the observation of the reporter lines in Arabidopsis inflorescences, the transgenic seeds were sowed on soil, and inflorescences were plucked and mounted on slides. The older floral buds were then carefully removed or spaced out to expose the SAM and early-stage floral buds. Dissected inflorescence was incubated with FM4-64 dye (50 μg/ml) for ~45 min on slides. Plants were imaged using a Zeiss LSM 510 upright (with motorized stage) confocal microscope with EC Plan-Neofluar 40×/1.30 oil differential interference contrast or Plan-Apochromat 20×/0.8 objective lens. GFP was stimulated with an argon laser at 488 nm at 60 to 70% of its output, with emission filtered using a 505- to 530-nm band-pass filter. VENUS was stimulated with an argon laser at 514 nm at 65 to 80% of its output, with emission filtered using a 530- to 600-nm band-pass filter. FM4-64 dye emission was filtered with a 585-nm long-pass filter. The z-stack was acquired using a 512-by-512–pixel frame, and the three-dimensional projections of the obtained z-stacks were then made with Zeiss LSM Image Brower version 4 and adjusted with Adobe Photoshop.

For the observation of T87 transgenic lines, 10 μl of cell suspension was mounted on slides and imaged using a Leica SP5 inverted confocal microscope with HCX PL APO 40×/1.25 objective lens. YFP was stimulated with an argon laser at 514 nm at 80% of its output, with emission filtered using a 520- to 550-nm band-pass filter. Images were acquired using 512 pixels by 512 pixels. Subsequently, images were made with Leica Application Suite Advanced Fluorescence v2.6.0 and adjusted with ImageJ v1.44 and Adobe Photoshop.

Supplementary Materials

Supplementary Text

Figs. S1 to S21

Tables S1 to S4

References (4044)

References and Notes

  1. Acknowledgments: We thank H. Li and K. Kanehara for their experimental help, G. C. Angenent for the pAG::GFP seeds, D. Voytas for Golden Gate TAL effector kit (through Addgene), and RIKEN BRC for the Arabidopsis T87 culture cells. We thank H. Yu and F. Berger for their comments on the manuscript. This work was supported by a research grant to T.I. from Temasek Life Sciences Laboratory (TLL), PRESTO (Japan Science and Technology Agency, 4-1-8 Honcho Kawaguchi, Saitama, Japan), and the National Research Foundation Singapore under its Competitive Research Programme (CRP Award NRFCRP001-108). All the constructs made in this work are available from T.I. under a material transfer agreement with TLL.
View Abstract

Stay Connected to Science

Navigate This Article