Developmental regulation of CYCA2s contributes to tissue‐specific proliferation in Arabidopsis

Since CYCA2 expression is strongly associated with proliferative tissues, such as primary and lateral root meristems (Supplementary Figure S2), we probed the impact of their loss‐of‐function on root growth. Growth defects were apparent when three out of the four CYCA2s were mutated. Because the post‐embryonic growth of the quadruple mutant was extremely slow, we preferentially analysed triple mutant combinations (Supplementary Figure S3). Triple mutants cyca2;134 and cyca2;234 had shorter roots and were impaired in lateral root formation compared with WT (Supplementary Figure S4). To determine whether these growth defects arose from an abnormal cell proliferation, root meristem phenotypes were analysed. Primary root meristems of cyca2;134 and cyca2;234 were smaller (Figure 1A; Supplementary Figure S5A and B) and contained fewer dividing cells than WT, as detected by antibodies to the cytokinesis‐specific syntaxin KNOLLE (Lauber et al, 1997; Figure 1B; Supplementary Figure S5C). Similarly, developing lateral root primordia in cyca2;234 contained fewer cells than WT (Figure 1C), suggesting that cell proliferation defects underlie both reduced root length and lateral root formation.

To determine at which cell‐cycle stage CYCA2s are prominently involved, cell‐cycle progression was compared during synchronized lateral root initiation in WT versus cyca2;234 triple mutants (Figure 1D). In WT, expression of both auxin signalling and G1‐to‐S regulatory genes preceded the expression of G2‐to‐M regulatory genes (Figure 1D), as previously reported (Himanen et al, 2002, 2004; Vanneste et al, 2005). By contrast, expression of mitotic regulators, such as B‐type cyclins, was no longer induced within the same time course in cyca2;234 mutants, whereas the expression of auxin signalling and G1‐to‐S regulatory genes was unaffected (Figure 1D). This delay in activation of mitotic regulators indicates that plant CYCA2s function early in the G2/M transition, as was predicted based on sequence homology (Vandepoele et al, 2002) and on expression patterns in synchronized cell suspensions (Menges et al, 2005). Moreover, it is likely that CYCA2s also function in S‐phase, given that CYCA2;2/CDKA;1 can phosphorylate the S‐phase regulator E2Fc in vitro (del Pozo et al, 2002). However, the lack of appropriate markers hampers such determination.

To obtain its characteristic final size and shape, leaf morphogenesis depends upon a tight coordination between cell proliferation, cell‐cycle exit and differentiation. Early leaf development displays high cell division activity that is followed by a gradual tip‐to‐base deceleration of proliferation and the start of differentiation‐associated endoreduplication and cell expansion (Donnelly et al, 1999; Beemster et al, 2006). The expression pattern of several CYCA2s also showed a comparable and dynamic gradient of expression (Supplementary Figure S6; Imai et al, 2006). Dramatic increases in ploidy levels and cell sizes were observed in the mature first true leaves of cyca2 triple mutants (Figure 2A and B). To address the mechanism driving enhanced ploidy levels and cell sizes, the development of cyca2;234 leaves was analysed in greater detail. Kinematic analysis of leaf growth showed lower cell division rates in cyca2;234 leaves compared with the WT (Figure 2C; Supplementary Figure S7). In addition, as soon as the first leaf pair became macroscopically visible (after 8 days of growth, Stage 1.02; Boyes et al, 2001), DNA content was already dramatically higher than the WT (Figure 2D; %2C, %4C and %8C). Moreover, ploidy levels continued to rise in cyca2;234 (after 14 days of growth), a period when endoreduplication had already stopped in the WT (Figure 2D; %16C and %32C). Thus, enhanced ploidy levels in cyca2;234 are the combined result of an early onset and extended duration of endoreduplication. Collectively, these phenotypic and molecular analyses in roots and shoots of cyca2 triple mutants demonstrate that plant CYCA2s are fundamental elements of the plant cell cycle, and, like their animal counterparts, function in early G2‐to‐M transition. Furthermore, the enhanced endoreduplication in these mutants is consistent with the observation that low CDK activity allows yeast cells in G2 to (re)enter the G1‐to‐S program without undergoing mitosis (Coudreuse and Nurse, 2010), suggesting that plant CYCA2s contribute to CDK activities that are required for mitosis.

In addition to their expression in meristems, CYCA2s were expressed in the leaf; however, while CYCA2;2 and CYCA2;3 were expressed throughout the organ (Supplementary Figures S6, S8 and S9), the expression pattern of CYCA2;1 and CYCA2;4 remarkably mimicked the reticulate vein pattern of the leaf (Figure 3A; Supplementary Figure S9). Moreover, the promoter activities of these two genes in the leaf overlapped with one of the earliest hallmarks of the vascular precursor (‘preprocambial’) cell state (Figure 3B), the promoter activity of the homeodomain‐leucine zipper (HD‐ZIP) III gene, ATHB8 (Scarpella et al, 2004; Donner et al, 2009). The tissue‐specific expression of CYCA2;1 and CYCA2;4 suggests that these CYCA2 genes function in leaf vascular development. Indeed, cyca2;234 leaves showed fewer vascular hypertrophy zones than the WT (Figure 3C and D); however, vascular defects in cyca2;234 were seemingly associated with changes in leaf shape resulting in leaves with fewer serration tips (Figure 3C and E). Additional mutation of the vascular‐specific CYCA2;1 in the cyca2;234 background further reduced the number of vascular hypertrophy zones without additional effects on the number of serration tips (Figure 3D and E), data which are consistent with the tissue‐specific expression pattern of CYCA2;1. Thus, vascular cell proliferation defects in cyca2 mutants likely derive from tissue‐specific modulation of CYCA2 levels, rather than being secondary consequences of disrupted leaf growth.

Stomata consist of two guard cells around a pore whose regulation controls gas exchange between the shoot and the atmosphere. Their development requires at least one asymmetric division as well as a single symmetric division. After the latter division, which occurs in a guard mother cell (GMC) precursor, stomatal differentiation and morphogenesis take place (Bergmann and Sack, 2007). The leaf epidermis of cyca2;134 and cyca2;234, but not those of WT and cyca2;124, showed frequent occurrence of unpaired oval cells, displaying cell wall thickenings and plastid accumulations, trait characteristics of wild‐type guard cells (Figure 4A). As in normal stomatal guard cells, these single cells were positioned above large intercellular spaces in the subjacent mesophyll (Figure 4B). Moreover, they expressed mature guard cell identity markers, KAT1pro:GUS (Nakamura et al, 1995) and ET1728 (Gardner et al, 2009; Figure 4C). Thus, these cells correspond to aberrant, single guard cells (SGCs) that are located where stomata would normally be found. These SGCs had twice the nuclear‐DNA content (4C) of normal guard cells (2C) (Supplementary Figure S10), suggesting they are arrested in G2‐phase. Yet, the aberrant cells attained a guard cell identity and formed SGCs instead of a pair.

Strikingly, SGCs could only be found in cyca2;3 mutant alleles and derived higher order cyca2 mutant combinations (Supplementary Table SI), suggesting that CYCA2;3 is a major contributing factor to this phenotype. However, while in single mutants the frequency of SGC formation is very low, additional mutations of other CYCA2 members resulted in dramatic increases in SGC frequencies (Supplementary Table SI). Collectively, these data demonstrate that CYCA2s are synergistically required for the symmetric division that is a prerequisite for stomatal formation, and that acquisition of guard cell identity occurs independently from GMC division.

SGCs were previously reported in transgenic plants harbouring a CDKB1;1‐N161 dominant‐negative construct (Boudolf et al, 2004), as well as cdkb1;1 cdkb1;2 double mutants (Xie et al, 2010). Moreover, CDKB1;1 can form a functional complex with CYCA2;3 (Boudolf et al, 2009) and CDKB1;1 is expressed around the time of GMC symmetric division (Boudolf et al, 2004), suggesting that CYCA2s and CDKB1s directly interact in promoting the formation of a two‐celled stoma. Indeed, while cdkb1;1 single mutants only had normal stomata, cyca2;234 cdkb1;1 quadruple mutants displayed even less SGCs than cyca2;234 triple mutants (Figure 5; Supplementary Figure S11). Thus, all four genes act synergistically in promoting GMC symmetric division, and thus stomatal morphogenesis.

While for CYCA2;1 and CYCA2;4 no stomatal expression could be observed (Supplementary Figure S12), CYCA2;3 expression along with CDKB1;1 and CYCA2;2, was induced in late GMCs, remained high in young guard cells, but was strongly reduced or did even disappear in mature stomata (Figure 6A). Together with previously identified mutants that have supernumerary guard cells in stomatal complexes (Lai et al, 2005; Ohashi‐Ito and Bergmann, 2006), the observed decline in cell‐cycle gene expression at the end of stomatal development hints at the existence of an active repression mechanism. Loss‐of‐function mutations in two MYB transcription factors, FOUR LIPS (FLP/MYB124) and its paralogue MYB88 induce clusters of four or more guard cells (Lai et al, 2005). Loss‐of‐function in the basic helix‐loop‐helix protein FAMA (Ohashi‐Ito and Bergmann, 2006) also results in cell clusters, but unlike those of flp myb88, without guard cell identity. The apparent independency from the stomata differentiation process renders FLP/MYB124 and MYB88 as potential candidate CYCA2 repressors, the more because they are expressed at roughly the same stages of stomatal development as CYCA2;2, and CYCA2;3 (Figure 6A).

To determine whether CYCA2 expression is required for the extra divisions found in flpmyb88, and/or fama backgrounds, we generated cyca2;234 fama‐1 quadruple and cyca2;234 flp‐7myb88 quintuple mutants. The cyca2;234 fama‐1 plants did not show any SGCs, instead they formed clusters of cells that lacked guard cell identity; however, these clusters had fewer cells than fama‐1 suggesting that fama‐1 is only partly epistatic to cyca2;234 (Supplementary Figure S13). By contrast, the formation of stomatal clusters in a flp‐7myb88 cyca2;234 background was completely suppressed (Figure 6B), demonstrating that CYCA2 gene products are required for the flp‐7myb88 stomatal phenotype and that FLP and MYB88 might represent transcriptional regulators of CYCA2 expression. Therefore, we analysed CYCA2 expression in a flpmyb88 background. Ten days after sowing, cotyledons from flp‐7myb88 seedlings showed about five‐fold higher CYCA2;3 expression than the WT (Figure 6C). Moreover, in flp‐1myb88 stomata, CYCA2;3 promoter activity remained high after the GMC division (Figure 6D), suggesting that FLP and MYB88 repress CYCA2;3 promoter activity. To test if this was a direct effect, we performed ChIP‐Q‐PCR using polyclonal antibodies raised against FLP and MYB88 (Xie et al, 2010). In the WT, CYCA2;3 promoter chromatin fragments were enriched after ChIP, while these were lost in flp‐1myb88 mutants (Figure 6E), demonstrating a specific, direct interaction of FLP and MYB88 with CYCA2;3 chromatin. Thus, FLP and MYB88 appear to restrict CYCA2;3 transcription after GMC division via direct interaction with its promoter.

Several findings led to the initial assumption that plant A‐type cyclins function in S‐phase and in the G2‐to‐M transition, in analogy to the animal and yeast cell‐cycle model. These findings include CYCA2 expression patterns in synchronized suspension cells (Reichheld et al, 1996; Shaul et al, 1996; Menges et al, 2005), their ability to rescue the growth of yeast cyclin mutants (Setiady et al, 1995) and their ability to induce Xenopus oocytes maturation (Renaudin et al, 1994). In addition, the ectopic expression of plant cyclins is sufficient to drive cells into mitosis (Imai et al, 2006; Boudolf et al, 2009).

Recently, it was shown that engineered yeast cells arrested in G2 are able to skip mitosis and re‐acquire a G1 status when CDK activity is low (Coudreuse and Nurse, 2010). Therefore, if CYCA2s affect mitotic CDK activity, one could expect ectopic endoreduplication and reduced proliferation in the absence of CYCA2 function. Previously, single mutants in cyca2;1 and cyca2;3 were shown to have increased levels of endoreduplication (Imai et al, 2006; Yoshizumi et al, 2006). Consistent with these data, we found that cyca2 triple mutants displayed greatly increased endoreduplication levels, reduced cell proliferation in developing leaves and G2‐arrest of GMCs resulting in SGCs with 4C DNA levels. Together, these data demonstrate that CYCA2s contribute to the CDK activity that is required for mitosis.

In animal systems, it is well established that B‐type cyclins in complex with a CDK act as mitosis‐promoting factor (MPF). MPF activity is further regulated by A‐type cyclins through effects on transcription, activation, localization and stability (Lindqvist et al, 2009). In plants, ectopic expression of CYCB1;2 in differentiated cell types such as trichomes was sufficient to trigger ectopic cell divisions, suggesting a MPF‐like function of CYCBs in plants (Schnittger et al, 2002). Using an in planta synchronized cell cycle‐inducible system, we found that the onset of B‐type cyclin expression was delayed in cyca2 triple mutants. Thus, mitotic entry involves the sequential activity of CYCA2‐CDK and CYCB‐CDK complexes.

Each cell type and tissue, within complex organs such as developing leaves, needs a custom‐tailored cell‐cycle regulation for the organ to reach its typical size and shape. This complexity is reflected in the large number of cell‐cycle regulatory genes in plants. In Arabidopsis, four CYCA2 genes are encoded in its genome. Each individual CYCA2 shows its own peculiar expression patterns across developing organs, displaying tissue‐ and cell type‐specific expression, such as in vascular tissues and the stomatal lineage. Their expression patterns also show variable degrees of overlap in certain tissues, suggesting local redundancies. Striking examples are the vascular expression of CYCA2;1 and CYCA2;4 and the stomatal expression of CYCA2;2 and CYCA2;3. In both tissues, the individual genes contribute locally to proliferation in a specific tissue or cell type.

Besides the expression pattern‐dependent redundancy, the mutant analyses revealed differential contributions of individual CYCA2s to proliferation. The analysis of the phenotypes of different triple mutants allowed the estimation of their relative importance for specific processes. In the case of root meristem size, lateral root formation, endoreduplication and stomatal development, CYCA2;3 seemed to be most relevant; during stomatal formation, only single cyca2;3 mutants resulted in SGC formation. Moreover, in combination with cyca2;3, other cyca2 mutations synergistically enhanced the frequency of SGC formation.

Observed differences in penetrance can be explained in part by tissue‐specific expression and relative expression levels. However, our study does not allow to exclude effects on protein stability and differences in biochemical properties, as additional regulatory mechanisms.

Proliferation and differentiation are largely mutually exclusive processes. While some cells exit the cell cycle after mitosis and remain in G1‐phase, other differentiating cells undergo several rounds of a modified cell cycle, in which the G2‐to‐M transition is omitted and only DNA synthesis occurs (endoreduplication). In animals, some developmental programs coordinate cell‐cycle exit during differentiation through transcription factor activity (Myster and Duronio, 2000; Buttitta and Edgar, 2007). One strategy is to induce CDK inhibitory proteins, while another is to repress cell‐cycle activating proteins. Interestingly, the transcription of A‐type cyclins is often actively repressed during differentiation processes (Li and Vaessin, 2000; James et al, 2006; Martinez et al, 2006; Sebastian et al, 2009; Pan et al, 2010). In plants, it is not known how developmental signals can modulate the switch between a full cell cycle and the endocycle or cell‐cycle exit during differentiation. Previously, increased level of polyploidy1 (ILP1) was found to act as repressor of CYCA2 expression (Yoshizumi et al, 2006). Here, we show that FLP and MYB88 repress CYCA2;3 expression during cell‐cycle exit in differentiating guard cells. This mechanism resembles the PROSPERO‐dependent mechanism in Drosophila that links neuronal lineage development with the transcriptional regulation of cell‐cycle regulatory genes (Li and Vaessin, 2000).

Mutations that affect CYCA2 function display higher than normal ploidy levels (Imai et al, 2006; Yoshizumi et al, 2006), whereas CYCA2;3 overexpression strongly suppresses endoreduplication (Imai et al, 2006; Boudolf et al, 2009), indicating that CYCA2 levels are major negative determinants of endoreduplication in leaves. Early stages of leaf development involve high proliferation rates, while later stages gradually switch to differentiation‐associated endoreduplication and cell expansion (Donnelly et al, 1999; Beemster et al, 2006). Interestingly, CYCA2;3 expression is rapidly repressed during the switch from proliferation to endoreduplication in differentiating leaves (Imai et al, 2006). Similarly, antagonizing auxin signalling also enhances endoreduplication via reduced CYCA2;3 expression (Ishida et al, 2010). However, it remains to be seen whether this effect is directly mediated by differentiation‐induced transcription factors, and how auxin is involved in this.

Stomatal development ends after a single symmetric division of a GMC, each of whose daughter cells terminally differentiate into individual guard cells (Bergmann and Sack, 2007). Mutants in the stomatal transcription factors FLP and MYB88 do not stop dividing after the GMC has divided, even though guard cell identity markers are expressed (Lai et al, 2005). We found that downregulation of CYCA2;3 after the first GMC division, normally seen in wild‐type plant, was absent in flpmyb88 double mutants. Direct interaction with CYCA2;3 promoter chromatin corroborate that FLP and MYB88 act as direct repressors of CYCA2;3 expression in guard cells. Similarly, the expression of an interacting CDK (Boudolf et al, 2009; Boruc et al, 2010b), CDKB1;1 was also shown to be directly repressed by FLP and MYB88 (Xie et al, 2010). These data are consistent with a model in which FLP and MYB88 enforce cell‐cycle exit during terminal guard cell differentiation by direct repression of CYCA2/CDKB1;1 kinase complexes. This mechanism ensures that stomata consist of only two guard cells, a condition required for their proper functioning as adjustable air valves.

We used Arabidopsis seedlings of the accession Col‐0 and Ler and mutants for the various A2‐type cyclins from publicly available collections (SALK (Alonso et al, 2003), GABI‐KAT (Rosso et al, 2003) and EXOn Trapping Insert Consortium (EXOTIC; http://www.jic.bbsrc.ac.uk/science/cdb/exotic/index.htm)), and stomatal lineage mutant alleles flp‐1myb88, flp‐7myb88 (Lai et al, 2005) and fama‐1 (Ohashi‐Ito and Bergmann, 2006). Cyclin mutant alleles used are cyca2;1‐1 (SALK_121077; Yoshizumi et al, 2006), cyca2;1‐2 (SALK_123348), cyca2;2‐1 (GABI_120D03), cyca2;3‐1 (SALK_092515; Imai et al, 2006), cyca2;3‐2 (SALK_086463; Imai et al, 2006), cyca2;3‐3 (SALK_043246), cyca2;4‐1 (SALK_070301) and cyca2;4‐2 (GAT_5.10009; Supplementary Figure S1). Promoter∷reporter lines for FLP (Lai et al, 2005), CDKB1;1 (Xie et al, 2010) and CYCA2;1 (Burssens et al, 2000) have been reported previously. For detection of T‐DNA inserts, we used primers specific to the left border of the T‐DNAs used for mutagenesis (LBC1, LB_GABI and LB_EXOTIC) in combination with gene‐specific primers (Supplementary Table SII). The alleles cyca2;1‐1, cyca2;2‐1, cyca2;3‐1 and cyca2;4‐1 are representative knockout alleles and have been used for generating triple mutants. After surface sterilization, seeds were sown on half‐strength MS medium supplemented with 1% sucrose and 0.8% agar. After stratification, plates were moved to cooled benches in a growth chamber (temperature, 22°C; irradiation, 65 μE/m²/s photosynthetically active radiation; photoperiod, 16 h light/8 h dark or continuous light).

One‐week‐old seedlings, grown on 0.5 × MS medium under continuous illumination, were fixed in paraformaldehyde. Immunolocalization was performed as described (Sauer et al, 2006). The rabbit anti‐knolle antibody (1:2000; Lauber et al, 1997) and the fluorochrome‐conjugated secondary antibody anti‐rabbit‐Cy3 (1:600; Dianova) were used. Fluorescence detection was done with a confocal laser‐scanning microscope Zeiss 710.

Promoter∷GUS‐GFP fusions of MYB88, CYCA2;2, CYCA2;3 and CYCA2;4 were generated through Gateway cloning of promoter fragments into pKGWFS7. PCR fragments of CYCA2;2, CYCA2;3 and CYCA2;4 promoters were described previously (Benhamed et al, 2008). To generate the CYCA2;1 and CYCA2;4 transcriptional fusions (CYCA2;1pro:HTA6:EYFP and CYCA2;4pro:HTA6:EYFP, respectively), 1808 bp upstream of the CYCA2;1 start codon and 1963 bp upstream of the CYCA2;4 start codon were amplified from Arabidopsis (Arabidopsis thaliana) ecotype Col‐0 genomic DNA and cloned into the Gateway‐adapted pFYTAG binary vector, which contains a translational fusion between the coding region of histone 2A (HTA6; At5g59870) and that of the enhanced YFP (EYFP) (Zhang et al, 2005).

The origin of the ATHB8pro:HTA6:EYFP and the ATHB8pro:ECFP‐Nuc has been described (Sawchuk et al, 2007, 2008). Seeds were sterilized and germinated, and seedlings and plants were grown, transformed and selected as described (Sawchuk et al, 2007, 2008). For CYCA2;1pro:HTA6:EYFP and CYCA2;4pro:HTA6:EYFP, the progeny of eight independent, single insertion transgenic lines were inspected to identify the most representative expression pattern. We define ‘days after germination’ (DAG) as days following exposure of imbibed seeds to light. Dissected seedling organs were mounted and imaged as described (Sawchuk et al, 2007, 2008; Donner et al, 2009). Brightness and contrast were adjusted through linear stretching of the histogram in ImageJ (National Institutes of Health; http://rsb.info.nih.gov/ij). Signal levels and colocalization were visualized as described (Sawchuk et al, 2008).

Total RNA was extracted with the RNeasy Mini kit (Qiagen, Venlo, The Netherlands). Poly(dT) cDNA was prepared from 1 μg total RNA with the Superscript III First Strand Synthesis System for RT–PCR (Invitrogen, Carlsbad, CA) and quantified on an iCycler apparatus (Bio‐Rad, Hercules, CA) with the Platinum SYBR Green qPCR Supermix‐UDG kit (Invitrogen, Merelbeke, Belgium). PCR was carried out in 96‐well optical reaction plates heated for 10 min to 50°C to allow UNG activity, followed by 10 min of 95°C to activate hot start Taq DNA polymerase, and 40 cycles of denaturation for 20 s at 95°C and annealing–extension for 20 s at 58°C. Target quantifications were performed with specific primer pairs designed using Beacon Designer 4.0 (Premier Biosoft International, Palo Alto, CA). Expression levels were normalized to At5g25760 (Q_PEX4) and At4g16100 (Q_UNKN1), which showed constitutive expression across samples. All Q‐RT–PCR experiments were performed in triplicates and the data were processed using qBase v1.3.4 (Hellemans et al, 2007).

The β‐glucuronidase (GUS) assays were performed as described (Beeckman and Engler, 1994). For microscopic analysis, chlorophyll was removed by EtOH treatment and further cleared by mounting in 90% lactic acid (Acros Organics, Brussels, Belgium). All samples were analysed by differential interference contrast microscopy.

For anatomical sections, samples were fixed overnight in 1% glutaraldehyde and 4% paraformaldehyde in 50 mM phosphate buffer (pH 7). Samples were dehydrated and embedded in Technovit 7100 resin (Heraeus Kulzer, Wehrheim, Germany) according to the manufacturer's protocol. Sections of 5 μm were cut with a microtome (Minot 1212; Leitz, Wetzlar, Germany), dried on Vectabond‐coated object glasses, stained with Toluidine Blue for 8 min (Fluka Chemica, Buchs, Switzerland), and rinsed in tap water for 30 s. After drying, the sections were mounted in DePex medium (British Drug House, Poole, UK).

Primary leaves of 3‐week‐old seedlings were chopped with a razor blade in 300 μl of buffer (45 mM MgCl₂, 30 mM sodium citrate, 20 mM 3‐[N‐morpholino]propanesulphonic acid, pH 7, and 1% Triton X‐100). To the supernatants, 1 μl of 4′,6‐diamidino‐2‐phenylindole from a stock of 1 mg/ml was added, which was filtered over a 30‐μm mesh. The nuclei were analysed with a CyFlow^® ML (Partec) flow cytometer.

Nuclei were stained fluorescently by fixing 3‐week‐old cotyledons in a mixture of 9:1 ethanol:acetic acid (v/v). After the samples had been rinsed, they were stained for 24 h with 0.1 μg/ml of 4′,6‐diamidino‐2‐phenylindole, mounted in Vectashield mounting medium (Vector Laboratories, Burlingame, CA), and observed under a × 63 oil immersion objective on a Zeiss Axioskop equipped with an Axiocam CCD camera (Zeiss). Images were obtained using the Axiovision software and were analysed in grey scale with the public domain image analysis program ImageJ (version 1.28; http://rsb.info.nih.gov/ij/). Relative fluorescence units were reported as integrated density, which are the product of the area and the average fluorescence of the selected nucleus.

Plants of the wild‐type and the cyca2 triple mutants were sown in quarter sections of round 12 cm Petri dishes filled with 100 ml of half‐strength Murashige and Skoog medium (Duchefa, Haarlem, The Netherlands) and 0.6% plant tissue culture agar (Lab M, Bury, UK). At relevant time points after sowing, plants or primary leaves (depending on the size) of the respective genotypes were harvested. All healthy plants were placed in methanol overnight to remove chlorophyll, and subsequently they were cleared and stored in lactic acid for microscopy.

The following parameters were determined: total area of all cells in the drawing, total number of cells and number of guard cells. From these data, we calculated the average cell area. We estimated the total number of cells per leaf by dividing the leaf area by the average cell area (averaged between the apical and basal positions). Finally, average cell division rates for the whole leaf were determined as the slope of the log₂‐transformed number of cells per leaf, which was done using five‐point differentiation formulas (Erickson, 1976).

Polyclonal antibodies against the FLP/MYB88 proteins were generated by inoculating rabbits with Ni‐NTA affinity purified NHis6‐MYB88. ChIP experiments were performed essentially as before (Xie et al, 2010). In brief, 10‐day‐old shoots of wild‐type, flp‐1 myb88 (200 mg fresh weight for each) were crosslinked in 1% formaldehyde for 20 min by vacuum filtration, and the crosslinking reaction was stopped by the addition of 0.1 M glycine (final concentration) for additional 5 min. Tissues were ground to a fine powder using mortar and pestle in liquid nitrogen and then suspended in 300 μl of lysis buffer (50 mM HEPES, pH 7.5; 150 mM NaCl; 1 mM EDTA, pH 8.0; 1% Triton X‐100; 0.1% sodium deoxycholate; 0.1% SDS; 1 mM phenylmethanesulphonylfluoride; 10 mM sodium butyrate; 1 × protein protease inhibitor from Sigma), and sonicated to achieve an average DNA size of 0.3–1 kb. The sonication conditions using the Bioruptor (Diagenode) were as follows: at high power; 30 s of sonication followed by 30 s of break; change ice every 10 min; 30 min in total. After cleared using 30 μl salmon sperm DNA/protein‐A agarose (Upstate) at 4°C for at least 1 h, the supernatant fractions were incubated, respectively, with 1 μl FLP/MYB88 rabbit polyclonal antibody or 1 μg rabbit IgG (Abcam) at 4°C overnight. At the same time, 10% of the supernatant was saved as the input fraction. The chromatin–antibody complex was incubated with salmon sperm DNA/protein‐A agarose (Upstate) at 4°C for at least 3 h, washed with lysis buffer, LNDET buffer (0.25 M lithium chloride; 1% NP40; 1% sodium deoxycholate and 1 mM EDTA, pH 8.0) and TE buffer (10 mM Tris–Cl, pH 7.5; 1 mM EDTA, pH 8.0) twice, respectively, and the complex was reverse crosslinked in elution buffer (1% SDS; 0.1 M NaHCO₃; 1 mg/ml proteinase K) overnight at 65°C. DNA was extracted using the PCR Cleaning Kit (Qiagen). The presence of the promoter of CYCA2;3 gene was examined by real‐time PCR using SYBR‐Green chemistry. The housekeeping gene PDF2/PP2A (At1g13320) was used as an internal control for normalization. The fold enrichment was normalized to the internal control PDF2/PP2A using the 2^−ΔΔCt method. Two biological replicates were conducted for each real‐time PCR experiment. The ChIP‐PCR primers used are listed in Supplementary Table II.

Supplementary data are available at The EMBO Journal Online (http://www.embojournal.org).