Regulation of Cell Proliferation in the Stomatal Lineage by the Arabidopsis MYB FOUR LIPS via Direct Targeting of Core Cell Cycle Genes

The MYB protein FOUR LIPS (FLP) promotes Arabidopsis stomatal patterning by suppressing cell division before differentiation. FLP direct targets were found to be enriched in cell cycle genes that function in both S-G1 and G2-M phase, indicating that this transcription factor acts as a developmental integrator.

Abstract

Stomata, which are epidermal pores surrounded by two guard cells, develop from a specialized stem cell lineage and function in shoot gas exchange. The Arabidopsis thaliana FOUR LIPS (FLP) and MYB88 genes encode closely related and atypical two-MYB-repeat proteins, which when mutated result in excess divisions and abnormal groups of stomata in contact. Consistent with a role in transcription, we show here that FLP and MYB88 are nuclear proteins with DNA binding preferences distinct from other known MYBs. To identify possible FLP/MYB88 transcriptional targets, we used chromatin immunoprecitation (ChIP) followed by hybridization to Arabidopsis whole genome tiling arrays. These ChIP-chip data indicate that FLP/MYB88 target the upstream regions especially of cell cycle genes, including cyclins, cyclin-dependent kinases (CDKs), and components of the prereplication complex. In particular, we show that FLP represses the expression of the mitosis-inducing factor CDKB1;1, which, along with CDKB1;2, is specifically required both for the last division in the stomatal pathway and for cell overproliferation in flp mutants. We propose that FLP and MYB88 together integrate patterning with the control of cell cycle progression and terminal differentiation through multiple and direct cell cycle targets. FLP recognizes a distinct cis-regulatory element that overlaps with that of the cell cycle activator E2F-DP in the CDKB1;1 promoter, suggesting that these MYBs may also modulate E2F-DP pathways.

INTRODUCTION

Stomata are turgor-operated valves essential for plant gas exchange, carbon assimilation, and water use efficiency. Since each stoma consists of two facing guard cells around a pore, and stomata are rarely found in contact (Figures 1A and 1B), these traits are presumably adaptive for plant survival and productivity (Bergmann and Sack, 2007). Stomatal number and distribution depend upon a balance between cell proliferation and differentiation. As in most plants, Arabidopsis thaliana stomata differentiate after at least one asymmetric and one final symmetric division (Figure 1A). The initial division creates a smaller precursor cell, a meristemoid, which later develops into an oval guard mother cell (GMC). Meristemoids and their sister cells usually undergo additional asymmetric divisions (Geisler et al., 2000). The number of epidermal cells produced, including stomata, depends upon the number of unequal divisions. In contrast with the extensive cell proliferation via asymmetric divisions that occurs earlier in the stomatal cell lineage, each GMC divides only once, dividing symmetrically to produce two cells of equal size and fate. Moreover, stomatal guard cells (GCs) do not divide, suggesting that proliferation is repressed in young as well as in mature GCs.

Figure 1.

Stomatal Development and Mutant Phenotypes in Pro35S:CDKB1;1-N161 (N161) and flp-1 Plants.

(A) Stomata form via asymmetric division(s) and one symmetric division. Late GMCs develop cell type–specific end wall thickenings. Pad-like wall thickenings form the stomatal pore. M, meristemoid.

(B) Living, wild-type stoma with pore. Differential interference contrast optics (DIC).

(C) Single guard cell in N161 background lacks dividing wall and pore (DIC).

(D) Stomatal cluster in flp-1 with two stomata in lateral contact (DIC).

(E) In flp mutants, each GMC daughter cell can undergo an ectopic symmetric division, producing four adjacent guard cells.

The control of stomatal formation and patterning in Arabidopsis is enforced by a series of developmental checkpoints in a dedicated stem cell lineage (Nadeau, 2009). These checkpoints are controlled by different, but sometimes functionally overlapping, gene sets. The first set regulates asymmetric divisions and includes genes encoding putative receptors (e.g., TOO MANY MOUTHS and ERECTA), ligands (EPIDERMAL PATTERNING FACTORS1 and 2, CHALLAH, and STOMAGEN), and members of a downstream mitogen-activated protein kinase cascade (Nadeau and Sack, 2002; Shpak et al., 2005; Bergmann and Sack, 2007; Lampard et al., 2008; Abrash and Bergmann, 2009; Hara et al., 2009; Hunt and Gray, 2009; Nadeau, 2009; Sugano et al., 2010). These signaling proteins consist of positive and negative regulators of stem cell proliferation, and they also orient the plane of a subset of asymmetric divisions, processes that minimize the occurrence of pattern violations. The intrinsic polarity of these unequal divisions has recently been shown to require BASL, a novel protein (Dong et al., 2009).

A second gene set encodes basic helix-loop-helix (bHLH) transcription factors such as SPEECHLESS (SPCH), MUTE, and FAMA, which regulate key fate transitions that include entry into the stomatal lineage, the meristemoid to GMC conversion, and guard cell fate specification, respectively (Ohashi-Ito and Bergmann, 2006; MacAlister et al., 2007; Pillitteri et al., 2007). These genes also promote (SPCH) or repress (MUTE and FAMA) cell proliferation at specific stages. Additional bHLH encoding genes, bHLH116 (also known as INDUCER OF CBF EXPRESSION and SCREAM1) and bHLH33 (SCREAM2), also regulate fate transitions and cell proliferation (Kanaoka et al., 2008).

FOUR LIPS (FLP; MYB124) and MYB88 comprise a third gene set (Lai et al., 2005). Like FAMA, these MYB proteins restrict GMCs to a single division. Unlike FAMA, however, they are not required for a stomatal fate (Lai et al., 2005; Ohashi-Ito and Bergmann, 2006). Mutations in FLP induce one or more rounds of ectopic symmetric divisions leading to the formation of two (flp-1 allele) or more (flp-7 allele) stomata in direct contact, forming clusters (Figures 1D and 1E). FLP thus prevents new GMC daughter cells from perpetuating mother cell–like divisions. Mutations in MYB88 show no stomatal phenotype, but flp myb88 double mutants have larger stomatal clusters than flp alone due to extra symmetric divisions.

Despite the importance of cell proliferation in stomatal development, relatively few cell cycle genes are known to act directly in this pathway. Indirect positive regulators of stomatal development include CYCLIN D4;1 and CYCLIN D4;2, which when ectopically expressed can promote the proliferation of hypocotyl epidermal cells from which cells that initiate stomatal formation are selected (Kono et al., 2007). Altered levels of CELL DIVISION CYCLE6 (CDC6) and cdc10-DEPENDENT TRANSCRIPT1 also affect stomatal number, suggesting a link between the extent of licensing of origins of DNA replication and the number of asymmetric divisions in the stomatal pathway (Castellano et al., 2004).

A cell cycle gene directly implicated in stomatal formation is CYCLIN DEPENDENT KINASE B1;1 (CDKB1;1), which in a dominant-negative form (N161) blocks the GMC symmetric division, resulting in the formation of single guard cells that express correct GC fate markers and display many traits of differentiated stomata (Boudolf et al., 2004a) (Figure 1C). This is a type of plant-specific CDK, whose expression in the cell cycle is reported to be limited to S-G2 phases (Fobert et al., 1996; Sorrell et al., 2001). Arabidopsis CDKB1;1 is expressed specifically in the stomatal cell lineage and might promote the symmetric division of GMCs. However, because the N161 dominant-negative protein might also interfere with the activity of similar kinases, it remains to be seen whether CDKB1;1 alone is required for GMC division.

Thus, although cell cycle regulators have been extensively characterized in plants (Menges et al., 2005; Inze and De Veylder, 2006; Gutierrez, 2009), those that function in stomatal development are still poorly defined. Similarly, while many stomatal pathway genes restrict cell proliferation, their molecular targets are mostly unknown (Lampard et al., 2008). In particular, FLP and MYB88 might restrict proliferation directly by regulating the expression of the cell cycle machinery or indirectly, such as by controlling a switch in cell fate.

To probe how division is downregulated before differentiation, we analyzed FLP and MYB88 function. Using in vitro selective enrichment techniques, we show that FLP has a novel DNA binding preference. We further identified potential in vivo FLP targets using chromatin immunoprecipitation coupled with microarray analysis (ChIP-chip), which revealed that FLP and MYB88 directly bind promoters of a spectrum of cell cycle genes, including components of the prereplication complex, cyclins, and cyclin-dependent kinases (CDKs), findings that are consistent with a role of FLP/MYB88 in suppressing DNA replication and cell cycle progression. Among these genes, FLP negatively regulates the mitosis-inducing factor CDKB1;1, which is required for cell overproliferation in flp mutants. The FLP binding cis-regulatory element in the CDKB1;1 promoter overlaps with that of the cell cycle transcriptional activator E2F-DP, suggesting that FLP/MYB88 integrate cell cycle progression and terminal differentiation to form and pattern stomata.

RESULTS

FLP Is a Nuclear Protein with Unique DNA Binding Preferences

FLP and MYB88 encode atypical two-repeat R2R3-MYB proteins with unusual substitution patterns of key residues (Lai et al., 2005), raising the question whether FLP and MYB88 control stomatal development through specific promoter–DNA interactions or by some other non-DNA binding mechanism. To first assess the subcellular localization of FLP and MYB88, the FLP-GFP (for green fluorescent protein) and MYB88-GFP fusions, driven by the constitutive cauliflower mosaic virus 35S promoter (Pro35S), were transiently expressed in 3-week-old tobacco (Nicotiana benthamiana) plants by Agrobacterium tumefaciens–mediated infiltration. The FLP and MYB88 proteins show strong nuclear localization, in contrast with the presence of free GFP in both the nucleus and cytoplasm (see Supplemental Figure 1 online), suggesting that these MYBs can act in the nucleus.

From a phylogenetic perspective, FLP/MYB88 are positioned between the three-repeat 3R-MYB proteins that control cell division and the typical R2R3-MYB proteins that regulate a variety of plant-specific cellular processes (Lai et al., 2005). To determine whether FLP/MYB88 bind DNA, we investigated the ability of a bacterially expressed N-terminal histidine-tagged version of the MYB domain of FLP (NHis₆-FLP^MYB) to recognize the MSA motif (ATAACGG), present in the promoters of genes regulated by 3R-MYB factors (Ito et al., 2001), or the CC^T/_AACC motif, a characteristic DNA binding sequence of typical R2R3-MYB proteins. DNA binding analyses, performed by electrophoretic mobility shift assays (EMSAs), showed that NHis₆-FLP^MYB was unable to bind the MSA motif (FLP in Figure 2A, lane 1), present in the promoter of the animal mim-1 gene and recognized by the MYB domain of the v-MYB version of the c-MYB proto-oncoprotein (García et al., 1991) (v-MYB in Figure 2A, lane 3), but not by the MYB domain of P1, a typical maize (Zea mays) R2R3-MYB factor (Grotewold et al., 1994) (Figure 2A, lane 2). NHis₆-FLP^MYB did not bind either to the CC^T/_AACC motif (FLP in Figure 2A, lane 4), represented by the high-affinity P1 binding sites present in the promoter of the maize A1 flavonoid biosynthetic gene (Grotewold et al., 1994) and robustly recognized by the MYB domain of P1 (P1 in Figure 2A, lane 5), but not by the MYB domain of v-MYB (Figure 2A, lane 6). These results suggest either that FLP does not bind DNA or that it has a DNA binding specificity distinct from most other characterized MYB factors.

Figure 2.

FLP/MYB88 Bind DNA with Novel Sequence Preferences.

(A) EMSAs were performed with the NHis₆-FLP^MYB (FLP), NHis₆-P1^MYB (P1), and NHis₆-v-MYB (v-MYB) proteins on the mim-1, APB1, and SS-30 probes. FLP does not bind to the MSA element present in the promoter of the animal mim-1 gene nor does NHis₆-FLP^MYB recognize the canonical R2R3-MYB consensus site recognized by P1 in the APB1 probe. FLP, but not v-MYB or P1, recognizes the SS-30 sequence identified by SELEX. Black arrows indicate protein-DNA complexes. The white arrows show the positions of the free probe (bottom left) and of the likely nonspecific binding of bacterial proteins (right). The sizes of FLP^MYB, P1, and v-MYB are 12.5, 14.7, and 18.4 kD, respectively.

(B) Nucleotide frequency distribution of the FLP binding consensus deduced from the SELEX experiments, displayed using WebLogo where the sizes of the characters represent the frequency of occurrence. The vertical black arrow references a conserved C nucleotide in the consensus sequence.

(C) The SS-30 oligonucleotide (top sequence) corresponds to one of the sequences identified in the SELEX experiments using NHis₆-FLP^MYB, a bacterially expressed histidine-tagged (NHis₆-) FLP MYB domain (amino acids 23 to 125) protein. The vertical black arrow indicates the same conserved nucleotide in the consensus sequence (underlined) shown in (B). Nonradiolabeled double-stranded oligonucleotides containing mutations in the SS-30 sequence (m2-m7 and mm2-mm9) were used as competitors in 50 or 500 molar excess in the binding of NHis₆-FLP^MYB to SS-30. Mutations that interfere with NHis₆-FLP^MYB binding are unable to compete with SS-30 binding and thus allow excess SS-30 probe (radiolabeled) signal to be present in the latter two lanes.

To test whether FLP/MYB88 bind DNA and to define their optimal DNA binding sequences, we performed systematic evolution of ligands by exponential enrichment (SELEX) experiments with NHis₆-FLP^MYB. The DNA fragments contained 26 random bases flanked by conserved linker regions (Grotewold et al., 1994). After seven rounds of selection, 87 sequences were identified (see Supplemental Figure 2 online), all containing the [A/T/G][A/T/G]C[C/G][C/G] core consensus sequence (Figure 2B).

SS-30, one of the selected sequences, contains the consensus-fitting GGCGCGC motif (Figure 2C). To establish the significance of each nucleotide in this sequence for FLP/MYB88 recognition, EMSA was performed with a His-tagged Escherichia coli–expressed FLP-MYB domain protein, NHis₆-FLP^MYB, using SS-30 as a probe and mutant SS-30 versions as competitors. Consistent with the GCGC core being absolutely required for FLP/MYB88 binding, mutants m3, m4, and mm5-mm8 failed to compete with the binding of NHis₆-FLP^MYB to radiolabeled SS-30 (Figure 2C). By contrast, all the other mutants, including unlabeled SS-30 itself, competed for binding when in significant molar excess (Figure 2C). Binding of NHis₆-MYB88 to most of these sequences was also verified by EMSA.

Highlighting the distinct DNA binding consensus of FLP/MYB88, compared with other MYB proteins, neither P1 nor v-MYB recognized the SS-30 sequence, which is robustly recognized by NHis₆-FLP^MYB (Figure 2A, compare lane 7 with 8 and 9). Several mutant flp alleles result in early termination or frame-shift mutations within the MYB domain (Lai et al., 2005). Most informative for understanding the significance of the MYB domain for FLP function is the recessive flp-8 allele, which expresses a protein containing a Lys replacement of the Glu-84 residue, which is conserved in plant and animal MYB proteins (Lai et al., 2005). Like the flp-7 allele, flp-8 displays a more severe stomatal phenotype than flp-1 (Lai et al., 2005). To determine whether DNA binding is required for the GMC division function of FLP, we tested by EMSA whether an abnormal NHis₆-FLP^MYB protein that harbors the E84K mutation (FLP^MYBE84K) binds SS-30. Revealing the essential role of DNA binding for FLP function in vivo, FLP^MYBE84K did not bind DNA in vitro (see Supplemental Figure 3 online). Taken together, these results demonstrate that FLP and MYB88 are nuclear proteins with DNA binding preferences distinct from other known MYB proteins and that DNA binding is required for FLP/MYB88 function.

CDKB1;1 Is an Immediate Direct Target of FLP/MYB88

The role of FLP/MYB88 in restricting cell division, along with the demonstration that FLP binds DNA in a sequence-specific fashion, led us to investigate whether FLP/MYB88 target cell cycle genes. One candidate is CDKB1;1 (At3g54180), which is expressed in the stomatal cell lineage and which in a dominant-negative kinase form (N161) blocks GMC symmetric divisions (Boudolf et al., 2004a) (Figure 1C). Using a transcriptional fusion of the CDKB1;1 promoter with GFP (ProCDKB1;1:GFP), and similar to the expression of FLP (Figures 3A to 3F) (Lai et al., 2005), we detected ProCDKB1;1-driven fluorescence from late GMCs, newly formed guard cells, and from immature stomata (Figures 3G to 3L). GFP expression driven by a translational fusion, ProCDKB1;1:CDKB1;1-GFP, showed a similar pattern, except that the fluorescence was localized to the nucleus and was only very bright in newly formed guard cells (Figures 3M to 3R). No fluorescence was found in meristemoids, young GMCs (identified as those that had not yet developed specialized end wall thickenings), or in mature stomata. Thus, CDKB1;1, like FLP, is expressed just before and after the symmetric division, a developmental window that includes the G2-to-M transition in GMCs, the specification of guard cell fate, and the arrest or exit from cycling in young guard cells.

Figure 3.

GFP Expression Driven by the FLP and CDKB1;1 Promoters Was Detected Specifically around the Time of GMC Symmetric Division.

(A) to (F) Transcriptional fusion of the native FLP promoter driving GFP expression (ProFLP:GFP).

(G) to (L) ProCDKB1;1:GFP transcriptional fusion.

(M) to (R) Although GFP fluorescence was found in young guards using all three markers (D, J, and P), the ProCDKB1;1:CDKB1;1-GFP translational fusion, which was found in the nucleus, was especially bright (P).

Confocal laser scanning micrographs of living epidermis. Cell walls visualized using propidium iodide fluorescence (red channel converted to magenta). M, meristemoid; SD, symmetric division complete. All bars = 10 μm with each stage/column of images at same magnification.

To investigate whether CDKB1;1 is regulated by FLP/MYB88, we used quantitative RT-PCR (qRT-PCR) to compare the expression of CDKB1;1 between flp-1, the flp-1 myb88 double mutant, and wild-type plants. The accumulation level of the CDKB1;1 mRNA was normalized to KAT1, which is marker for mature guard cells (Lai et al., 2005). CDKB1;1 mRNA levels were significantly higher (P value < 0.01, two-sided t test) in flp-1 myb88 than in wild-type controls, as well in the flp-1 mutant (P value < 0.05, two-sided t test), which has a weaker phenotype than the double mutant (Figure 4). By contrast, the mRNA levels of CDKB1;2, a gene with high sequence identity to CDKB1;1, but with a much wider expression pattern (Wang and Yang, 2007), is similar between flp-1 myb88 and wild-type plants (Figure 4). Together, these results suggest that FLP and MYB88 delimit the temporal and spatial window of CDKB1;1 expression.

Figure 4.

FLP/MYB88 Negatively Control CDKB1;1 Expression.

qRT-PCR results of CDKB1;1 and CDKB1;2 mRNA accumulation in wild-type (WT), flp-1, and flp-1 myb88 mutants. Expression was normalized to KAT1, a marker for mature guard cells (Nakamura et al., 1995), to correct for the increased number of guard cells in flp-1 and flp-1 myb88 mutants. Error bars correspond to the sd on biological triplicates. Differences among Col-0, flp-1, and flp-1 myb88 plants were statistically significant for CDKB1;1 (P value < 0.0002) but not for CDKB1;2 (P value = 0.4101).

To determine whether FLP/MYB88 directly control CDKB1;1 expression, we generated a rabbit polyclonal antibody against E. coli–expressed and affinity-purified full-length NHis₆-MYB88. This antibody also recognizes NHis₆-FLP (data not shown) as would be expected since these genes are paralogous and the corresponding proteins are highly similar in sequence (Lai et al., 2005). Both MYB88 and FLP appear to have been generated by the last polyploidy event during the evolutionary history of Arabidopsis (i.e., at the base of the Brassicaceae family roughly 25 million years ago; Keith Adams, personal communication). Moreover, an extra, transgenically induced genomic MYB88 fragment (gAtMYB88) complements the flp phenotype even though flp mutants already have two wild-type copies of MYB88 (Lai et al., 2005).

The FLP/MYB88 antibody was used for ChIP experiments on 10-d-old wild-type (Columbia-0 [Col-0]) and flp-1 mutant plants (Figure 5A). An antibody recognizing acetylated histone H3 (Figure 5A, AcH3) at Lys-9 and Lys-14 served as a control for the ChIP experiments. Unlike the idiotypic IgG negative control, which showed no bands after amplification (Figure 5A, IgG), the FLP antibodies (Figure 5A, αFLP) immunoprecipitated a CDKB1;1 promoter fragment in wild-type plants (Figure 5A, Col-0). The faint band sometimes observed in flp-1 is likely due to native MYB88 expression, which is insufficient to complement the flp-1 phenotype and is expressed at very low levels. The absence of any band in flp-1 myb88 (Figure 5A) is consistent with the enhanced stomatal phenotype of the double mutant (Lai et al., 2005). By contrast, flp-1 mutants harboring gAtMYB88, which results in increased MYB88 expression (Figure 5B), complements the flp-1 phenotype and shows robust in vivo MYB88 binding to the CDKB1;1 promoter fragment in ChIP experiments (Figure 5A, flp-1 gAtMYB88). Taken together, these results indicate that FLP/MYB88 directly bind the CDKB1;1 promoter.

Figure 5.

FLP/MYB88 Bind the CDKB1;1 Promoter in Vivo.

(A) ChIP experiments used antibodies against FLP/MYB88 (αFLP), which were then followed by qPCR to determine the extent of enrichment of CDKB1;1 promoter sequences in Col-0 (wild type), flp-1, flp-1 myb88, and flp-1/gAtMYB88 plants. The flp-1/gAtMYB88 plants carry a transgene containing a genomic fragment of MYB88 that complements the flp-1 myb88 mutant phenotype (Lai et al., 2005). An antibody against acetylated (at Lys-9 and Lys-14) histone H3 (AcH3) was used as a positive control. Immunoprecipitated DNA from biological triplicates was quantified for fold enrichment using qPCR. Error bars correspond to the sd of biological triplicates. The differences among all four genotypes were statistically significant (P value < 0.0001).

(B) qRT-PCR levels of MYB88 mRNA accumulation with mRNA levels normalized to those of KAT1, a stomatal specific gene, in 12-d-old flp-1 plants that harbor the gAtMYB88 transgene. The difference between flp-1 and flp-1/gMYB88 is statistically significant (P value < 0.01).

Identification of FLP/MYB88 Binding Sites in the CDKB1;1 Promoter

To identify the FLP/MYB88 binding sites in CDKB1;1, we cloned the 5′ region of CDKB1;1 from the transcription start site (TSS) to the stop codon of the immediately upstream gene (At3g54170). This 399-bp fragment contains all the sequences previously used to investigate the regulation of CDKB1;1 by E2F-DPa (Boudolf et al., 2004a), with the exception of the CDKB1;1 5′-untranslated region (UTR). To test binding by EMSA, the 399-bp promoter region was split into a 256-bp upstream region and a 143-bp fragment most proximal to the CDKB1;1 TSS (Figure 6A). EMSA experiments were conducted with NHis₆-FLP^MYB, NHis₆-FLP, and NHis₆-MYB88 on both fragments, but a protein-DNA complex was only observed with the proximal 143-bp region (Figure 6B). Using the FLP/MYB88 DNA binding preference (Figure 2B), we scanned this 143-bp fragment and identified two putative FLP/MYB88 binding sites (Figure 6A). Each of these sites was mutated (mutations m1 and m2, Figure 6A), and the corresponding fragments were tested for binding by NHis₆-MYB88. The m1, but not the m2, mutation abolished NHis₆-MYB88 binding (Figure 6C, compare m1 and m2), indicating that the sequence at position −66 (with respect to the TSS) corresponds to a FLP/MYB88 binding site. To validate these results in vivo, the wild-type CDKB1;1 promoter or promoter versions harboring the m1 or m2 mutation (CDKB1;1^m1 and CDKB1;1^m2, respectively) were transformed into wild-type Arabidopsis plants. We performed ChIP experiments using the FLP/MYB88 antibodies (αFLP) and PCR with a primer set that distinguished between the endogenous and transformed CDKB1;1 promoters. Consistent with the EMSA experiments (Figure 6C), the ChIP results showed that, while the m1 mutation completely abolished FLP/MYB88 binding, the m2 mutation had no effect (Figure 6D).

Figure 6.

FLP/MYB88 Binding Sites in the CDKB1;1 Promoter.

(A) Structure of the CDKB1;1 promoter showing the position of two putative FLP/MYB88 binding sites.

(B) EMSA with NHis₆-FLP^MYB (FLP^MYB), NHis₆-FLP (FLP), and NHis₆-MYB88 (MYB88) using the proximal region of the CDKB1;1 promoter as probe. Black arrows, protein-DNA complexes; white arrow, free probe; gray arrow, likely degradation products of FLP. The sizes of FLP^MYB, FLP, and MYB88 are 12.5, 51, and 56 kD, respectively.

(C) EMSA with NHis₆-MYB88 and the wild type (WT) or mutant (m1 and m2) versions of the CDKB1;1 promoter. Black arrow, protein-DNA complexes; white arrow, position of the free probe.

(D) PCR performed on ChIP DNA obtained with antibodies against FLP/MYB88 (αFLP) from transgenic Arabidopsis plants harboring the wild-type (ProCDKB1;1^WT), m1 (ProCDKB1;1^m1), or m2 (ProCDKB1;1^m2) versions of the CDKB1;1 promoter. The black triangles (top) represent serial fourfold dilutions used for PCR.

(E) FLP/MYB88 repress CDKB1;1 transcription. Transient expression experiments were performed by bombarding Arabidopsis seedlings with the Pro35S:proCDKB1;1^WT:Luc and Pro35S:proCDKB1;1^m1:Luc reporter plasmids together with Pro35S:MYB88 with or without the genomic FLP minigene (gFLP) and with Pro35S:Renilla as a normalization control. The x axis represents the percentage of activation compared with the corresponding reporters alone (100%). Error bars represent the sd of three independent biological replicates. The differences between gFLP, gFLP + Pro35S:MYB88, and the wild-type reporter (Pro35S:proCDKB1;1^WT:Luc) were statistically significant (F test, P value <0.0001) but not compared with the mutated Pro35S:proCDKB1;1^m1:Luc construct (P value = 0.7482).

qRT-PCR experiments comparing wild-type and flp myb88 mutant plants indicated that FLP/MYB88 negatively control CDKB1;1 mRNA accumulation (Figure 4). To further establish the importance of the verified FLP/MYB88 binding site in the CDKB1;1 promoter, we cloned the wild-type and the m1 mutant 399-bp CDKB1;1 promoter fragment between a double Pro35S enhancer and the luciferase (Luc) reporter, resulting in the Pro35S:proCDKB1;1^WT-Luc and Pro35S:proCDKB1;1^m1:Luc plasmids. Using biolistic transformation, we bombarded these plasmids with and without Pro35S:MYB88 into young Arabidopsis seedlings, together with the genomic FLP (gFLP) that complements the flp mutant phenotype (Lai et al., 2005). A Pro35S:Renilla reporter was included in all bombardments as a normalization control. The activity of luciferase (normalized to Renilla) was compared between samples bombarded with only Pro35S:proCDKB1;1^WT:Luc or Pro35S:proCDKB1;1^m1:Luc reporters (Figure 6E) and samples cobombarded with the FLP/MYB88 regulators. Consistent with a role of FLP and MYB88 in repressing CDKB1;1 expression by binding to the m1 site, gFLP with or without MYB88 significantly repressed the expression of Pro35S:proCDKB1;1^WT-Luc but not of Pro35S:proCDKB1;1^m1-Luc (Figure 6E). To confirm the importance of this cis-regulatory element for the transcriptional repression of CDKB1;1, qRT-PCR was conducted on transgenic Arabidopsis plants harboring the ProCDKB1;1^WT:GUS (for β-glucuronidase) or ProCDKB1;1^m1:GUS constructs. Consistent with the m1 mutation abolishing the repressive activity of FLP/MYB88 on CDKB1;1 expression, GUS expression was significantly higher in plants harboring the ProCDKB1;1^m1:GUS constructs (see Supplemental Figure 4 online). We interpret the higher expression of ProCDKB1;1^m1:GUS to indicate that other positive regulators, besides E2Fa, must contribute to CDKB1;1 expression.

FLP/MYB88 and E2F Binding Sites Overlap in the CDKB1;1 Promoter

We noted that the m1 site overlaps with a cis-regulatory element that was previously proposed to be recognized by the E2F-DPa regulatory complex for CDKB1;1 activation (Boudolf et al., 2004b). Because these earlier experiments were performed by transient activation in tobacco BY2 cells, they could not establish whether E2Fa directly interacts with the CDKB1;1 promoter. Thus, we expressed the E2Fa gene under the control of a 35S promoter fused to GFP (Pro35S:E2Fa-GFP) in transgenic Arabidopsis plants. ChIP experiments were performed using an antibody against GFP (αGFP), as previously described (Morohashi and Grotewold, 2009), and the presence of the CDKB1;1 promoter in the immunoprecipitate was determined by PCR. A product was identified only in the Pro35S:E2Fa-GFP transgenic plants and not in control plants (Figure 7A), confirming that CDKB1;1 can be an immediate direct target of E2Fa in Arabidopsis. To determine whether E2Fa specifically recognizes the m1 site in the CDKB1;1 promoter, as previously proposed (Boudolf et al., 2004b), wild-type and mutant versions of the CDKB1;1 promoter were cloned upstream of GUS, cotransformed with Pro35S:E2Fa-GFP into N. benthamiana leaves by agroinfiltration, and tested for E2Fa binding by ChIP. As shown in Figure 7B, E2Fa-GFP was only tethered to the wild-type version of the promoter (ProCDKB1;1^WT:GUS, Figure 6A) and not to the promoter harboring the m1 mutation (ProCDKB1;1^m1:GUS, Figure 6A). These results demonstrate that FLP/MYB88 and E2Fa, and perhaps other E2Fs, bind to the same or overlapping cis-regulatory elements in the CDKB1;1 promoter.

Figure 7.

An Intact FLP/MYB88 Binding Site in the CDKB1;1 Promoter Is Needed for E2Fa Binding.

(A) PCR-based amplification of E2Fa from ChIP DNA obtained using an anti-GFP antibody (αGFP) in Arabidopsis plants that harbor Pro35S:E2Fa-GFP (top row) but not from untransformed controls. The black triangles in (A) and (B) represent serial fourfold dilutions used for PCR.

(B) Amplification of E2Fa-GFP from ChIP DNA obtained from N. benthamiana agroinfiltrated leaves using antibodies against GFP (αGFP).

PCR performed on ChIP DNA with Pro35S:E2Fa-GFP and the wild-type or mutant (m1) CDKB1;1 promoters driving GUS.

CDKB1;1 and CDKB1;2 Redundantly Promote Stomatal Development

Our results show that CDKB1;1 is a direct target of FLP/MYB88, raising the question of whether this regulation is sufficient to explain the flp myb88 clustered stomata mutant phenotype. While the N161 dominant-negative version of CDKB1;1 blocks GMC division (Figure 1C) (Boudolf et al., 2004a), this phenotype might result from N161 interfering with the function of related kinases, such as CDKB1;2. Moreover, the loss-of function stomatal phenotypes of neither CDKB1;1 nor CDKB1;2 individually or together have been previously reported.

Heterozygous or homozygous T-DNA insertional alleles in either the CDKB1;1 or the CDKB1;2 locus (see Methods) showed no apparent stomatal abnormalities in cotyledons and first leaves (Figures 8A to 8C; see Supplemental Table 1 and Supplemental Figure 5 online). By contrast, in a cdkb1;1 cdkb1;2 double mutant (SALK_073457 and SALK_33560), 46 and 17% of all stomata consisted of single GCs in cotyledons and first leaves, respectively (Figure 8D). That these cells are single GCs was confirmed by their expression of a stomatal fate marker ET1728 (Gardner et al., 2009) (Figures 8F and 8G). Thus, neither locus on its own is essential for the last division in the stomatal lineage, but together they are required for GMC cytokinesis in many stomata. This redundancy is further highlighted by the absence of single GCs in the cdkb1;1 cdkb1;2 double mutant transformed with a ProCDKB1;1:CDKB1;1-GFP translational fusion. Unlike CDKB1;1 expression, no epidermal fluorescence was detected using a ProCDKB1;2:GFP transcriptional fusion, consistent with the lower abundance of CDKB1;2 transcripts compared with CDKB1;1 in synchronized Arabidopsis cell suspension cultures (Menges et al., 2005).

Figure 8.

CDKB1;1 and CKDB1;2 Promote Symmetric Division and Are Downstream of FLP.

(A) to (C) Single guard cells are absent from the wild type and from cdkb1;1 (SALK_073457) and cdkb1;2 (SALK_133560) insertional, knockdown alleles (evaluated by qPCR). Cell walls in 4-d-old developing cotyledons were visualized using propidium iodide and laser scanning confocal microscopy.

(D) to (E) Single guard cells (black stars) are present in the cdkb1;1 cdkb1;2 double mutant as well as in N161 transformants.

(F) GFP fluorescence from stomatal identity marker ET1728 in wild-type stomata

(G) ET1728 fluorescence showing stomatal identity in two single guard cells in the cdkb1;1 cdkb1;2 double mutant.

(H) to (J) cdkb1;1 cdkb1;2 is epistatic to flp-7 as well as to flp-1 myb88

(K) Pro35S:CDKB1;1•N161 transformant showing single guard cells (black stars) and a normal stoma. DIC micrograph of cleared cotyledon.

(L) The presence of single guard cells in the Pro35S:CDKB1;1•N161 flp-1 double mutant indicates that N161 is epistatic to flp-1.

The cdkb1;1 cdkb1;2 phenotypes resemble those produced by dominant-negative CDKB1;1 (Pro35S:N161) quantitatively as well as qualitatively (Figures 8D, 8E, and 8K). Both genotypes had roughly comparable percentages of stomata with failed cytokinesis in cotyledons (i.e., 46.3% ± 1.5% [± se] for the double mutant and 51.9% ± 2.5% for N161). In addition to inducing single guard cells, Pro35S:N161 also reduces stomatal number (Boudolf et al., 2004a) (Figure 8E; see Supplemental Table 1 online). The cdkb1;1 cdkb1;2 double mutant also showed reduced densities of both normal stomata and of single GCs, a phenotype not found in either cdkb1;1 or cdkb1;2 single mutants (Figures 8B to 8D; see Supplemental Table 1 online). Thus, CDKB1;1 and CDKB1;2 together promote both the last division in the stomatal cell lineage as well as the number of stomata and stomatal cell lineages that form, and they act redundantly in both roles. These data indicate that the kinase dead form of N161 likely mimics the cdkb1;1 cdkb1;2 phenotype by interfering with the combined functions of CDKB1;1 and CDKB1;2 during stomatal development.

The flp myb88 Stomatal Cluster Phenotype Requires CDKB1 Function

Because FLP/MYB88 restrict and CDKB1;1/CDKB1;2 promote symmetric divisions, we asked whether they act in the same or different genetic pathways. A flp-7 cdkb1;1 cdkb1;2 triple mutant completely lacked stomatal clusters and most stomata were normal, but 11.6% (±0.9 se) were aberrant single guard cells (Figures 8H and 8I; see Supplemental Table 1 online). Similarly, the introgression of the N161 dominant-negative construct into flp-1 produced single guard cells (19.5% of all stomata) and virtually suppressed stomatal clusters (0.0012% out of a total of 12,730 stomatal units scored) (Figures 8K and 8L). Comparable phenotypes were also observed in the quadruple loss-of-function mutant (flp-1 myb88 cdkb1;1 cdkb1;2) (Figure 8J). The suppression of the flp myb88 phenotype is not simply due to a cytokinesis defect or to the production of fewer stomata because the vast majority of GMCs divided normally in these backgrounds and thus provided a large pool of cells susceptible to possible flp myb88–induced ectopic divisions. We conclude that the cdkb1;1 cdkb1;2 double mutant is epistatic to flp myb88, that both CDKB1 genes are likely to be downstream of FLP/MYB88 in stomatal development, and that CDKB1 genes are redundantly required for the flp myb88 phenotype.

ChIP-Chip Experiments Identify Additional Cell Cycle Genes Targeted by FLP/MYB88

Our results indicate that FLP/MYB88 directly target CDKB1;1, resulting in reduced CDKB1;1 expression (Figure 4). Consistent with its role in promoting nuclear and cell division, CDKB1;1 function has been primarily associated with the G2/M transition (Inze and De Veylder, 2006). However, CDKB1;1 is also expressed during S phase of the cell cycle, as well as during G2 and early M. In addition, the stomatal cluster phenotype of flp myb88 mutants is consistent with FLP/MYB88 acting during G1/S as well as G2/M.

To better define how FLP/MYB88 repress extra symmetric divisions, we searched for additional FLP/MYB88 direct target genes. We adapted ChIP methods coupled with the hybridization of the whole-genome Arabidopsis Affymetrix tiling 1.0R array, comprising >32 million 25-mer tiled features spaced every 10 nucleotides (Zhang et al., 2006) (ChIP-chip). We used the FLP/MYB88 antibody to immunoprecipitate the chromatin fragments associated with these regulators obtained from formaldehyde cross-linked green tissues of 10-d-old wild-type or flp-1 myb88 seedlings. Three biological ChIP-chip replicates for the wild type and flp-1 myb88 were performed. ChIP-chip analyses were performed essentially as previously described (Keles et al., 2006; Morohashi and Grotewold, 2009), using a sliding window in scan statistics of ~350 bp (corresponding to 10 probes) to conform to the spatial structure of the data (see Methods). Therefore, a peak corresponds to the integration of individual probe signals over multiple probes, minimizing artifacts resulting from spurious hybridizations to individual probes. Unless otherwise indicated, only peaks overlapping in at least two biological ChIP-chip replicates and within 3 kb of genes were considered putative target genes. According to these criteria, a total of 226 intergenic regions adjacent to 241 FLP/MYB88 putative target genes were identified (see Supplemental Data Set 1 online), and these were distributed over the five Arabidopsis chromosomes (see Supplemental Figure 6 online). Twenty-two genomic regions that were significantly enriched (compared with input DNA) in at least one FLP/MYB88 ChIP-chip experiment were targeted for validation by ChIP-PCR. Of these 22 regions, 19 showed robust and reproducibly different signals in ChIP experiments compared with wild-type and flp-1 myb88 plants (Figure 9; see Supplemental Figure 7 online); this suggests an experimentally validated (rather than predicted) false positive discovery rate of 14%. Because nine of the genes validated by ChIP came from positives identified in just one of the three ChIP-chip experiments, it is likely that the number of FLP/MYB88 targets is significantly larger than 241.

Figure 9.

Validation of Selected Cell Cycle Genes Identified in ChIP-Chip as FLP/MYB88 Direct Targets.

An antibody (αFLP) recognizing FLP/MYB88 was used to evaluate the ChIP-PCR binding of FLP/MYB88 to the respective gene promoters in 10-d-old Col-0 and flp-1 myb88 double mutant (dm) plants. Actin was used as a negative control. The parentheses under each gene abbreviation indicate the number of times that genes were identified in separate ChIP-chip experiments. The black triangles (top) represent serial fourfold dilutions used for PCR.

To uncover possible patterns in the biological functions of the genes targeted by FLP/MYB88, we used the Gene Ontology (GO) tools provided by the Munich Information Center for Protein Sequence. A single GO class was significantly (P = 2.17e⁻⁰⁴) enriched among these 226 FLP/MYB88 targeted regions and corresponded to cell cycle genes (see Supplemental Figure 8 online). CDKB1;1 was very significantly enriched for FLP/MB88 binding in just one of the three ChIP-chip experiments. However, as indicated, FLP/MYB88 binding to ProCDKB1;1 was confirmed using ChIP. Other cell cycle–related genes validated by ChIP binding of FLP/MYB88 (Figure 9) include CDKA;1 (At3g48750), CELL DIVISION CYCLE6a and 6b (CDC6a At2g29680 and 6b At1g07270), CYCLIND4;1 (At5g65420), and a cyclin-like gene, CYCLINT;1 (CYCT;1, At1g35440). In addition, CDKD1;3 (At1g8040) and CYCB1;3 (At3g11520) were found in only one of three ChIP-chip experiments, whereas CYCT;1 was identified in all three replicates.

CDKA;1 encodes the archetypal CDK that is related to the yeast cdc2 and human CDKs (Iwakawa et al., 2006; Nowack et al., 2006). While detection of a stomatal phenotype in cdka;1 mutants was precluded because of embryonic lethality (Nowack et al., 2006), a CDKA;1 transcriptional fusion (ProCDKA;1:YFP-DB) was strongly expressed before but not after GMC symmetric division (Figures 10A and 10B). This sharp developmental cutoff in expression was likely detected because the construct included a cyclin B destruction box (DB) fused to yellow fluorescent protein (YFP) that in transgenic plants targets CDKA;1 for anaphase-promoting complex-mediated degradation. Thus, despite an earlier report of ProCKDA;1:GUS staining in stomata (Serna and Fenoll, 1997), CDKA;1 expression appears to shut off as guard cells form, consistent with the observed timing of FLP activity.

Figure 10.

Reporter-Based Expression of Two FLP/MYB88 Target Cell Cycle Genes.

(A) and (B) The native CDKA;1 promoter (ProCDKA;1) drives expression of YFP (fused to a DB) in GMCs (white asterisks) in (A) but not in newly formed guard cells shown in (B) by white arrows. Confocal laser scanning micrographs. Cell walls visualized using propidium iodide fluorescence with red channel digitally converted to magenta.

(C) The native CDC6a promoter (ProCDC6a) drives GUS expression before (white star) and just after (white arrow) GMC symmetric division but not in older but still developing stomata (black arrow). DIC optics.

Both CDC6 proteins are recruited to prereplication complexes where they promote the licensing of DNA replication (Masuda et al., 2004). We evaluated expression driven by ProCDC6a and found GUS staining in late (but not early) GMCs as well as in young stomata (Figure 10C). While persistent staining might explain the presence of GUS in young stomata, no staining was found in mature guard cells. Thus, CDKA;1 and CDC6a are expressed when FLP and MYB88 act during stomatal development consistent with these cell cycle genes being transcriptional targets of FLP/MYB88.

DISCUSSION

Stomatal function depends developmentally upon correct control of cell division to generate two guard cells and new cell walls that then separate to form a pore. To achieve this, the GMC must divide once to form the two young guard cells, which must then cease further division to avoid stomatal clustering or stacking. FLP/MYB88 function is required to restrain further GC divisions, since loss of FLP and MYB88 functions lead to extra symmetric cell divisions (Lai et al., 2005). Previous analysis has also shown that GMC division can be blocked by overexpression of a kinase-dead dominant-negative form of CDKB1;1 known as N161 (Boudolf et al., 2004a). Here, we link the developmental phenotypes seen in flp mutants and cell cycle regulation by CDKB1, showing genetically that CDKB1 function, provided by either CDKB1;1 or CDKB1;2, is required both for the symmetric division of the GMC and for the flp clustering phenotype to be manifest. FLP binds to and negatively regulates expression of CDKB1;1 and of a number of other cell cycle regulatory genes to arrest further division activity within newly formed GCs.

CDKB1 Function Is Required for Stomatal Pore Formation and the flp Phenotype

Our analysis shows that loss of CDKB1 function in cdkb1;1 cdkb1;2 double mutants results in frequent formation of stomata composed of a single, undivided guard cell, recapitulating the phenotype produced by overexpression of N161 (Boudolf et al., 2004a). We conclude therefore that the N161 phenotype results from the disruption of the activities of both CDKB1;1 and CDKB1;2. These genes promote both nuclear and cell division in GMCs, since N161 (Boudolf et al., 2004a) and double loss-of-function mutants both show abnormal stomata that consist of a single cell with guard cell identity and also show a single 4C nucleus (data not shown).

Most CDKs in yeast and animals are expressed throughout the cell cycle with their activity controlled primarily at a posttranscriptional level. By contrast, the expression (Fobert et al., 1996) and activity (Sorrell et al., 2001) of plant CDKB genes varies during the mitotic cell cycle, suggesting that they are transcriptionally controlled (Menges et al., 2005). A cell division-promoting function for CDKB1;1 and CDKB1;2 is consistent with their expression peaking in early G2 phase and continuing into M phase in cell suspension (Menges et al., 2005) and with functional analysis (Boudolf et al., 2004b, 2009). This function is also supported by the pattern of CDKB1;1-driven GFP expression in the stomatal cell lineage that is first detected before symmetric division in late GMCs. In contrast with a previous report (Boudolf et al., 2004a), which employed a GUS reporter, we did not detect ProCDKB1;1-driven GFP expression in early stomatal precursor cells (meristemoids). Although CDKB1;1 was also reported to be expressed in mature stomata using ProCDKB1;1-driven GUS and in situ hybridization (Boudolf et al., 2004a), we could not detect GFP fluorescence in mature guard cells. Our data are therefore consistent with the conclusion that CDKB1;1 activity is necessary for the transition from G2 to M in GMCs (Inze and De Veylder, 2006).

We also find that the double mutant cdkb1;1 cdkb1;2 is epistatic to flp; hence, CDKB1 function is required for the repeated divisions of newly formed GCs in flp that lead to stomatal clusters. We show that CDKB1;1 expression is directly repressed by FLP/MYB88, so it is possible that FLP normally blocks ectopic divisions by directly repressing CDKB1;1 expression in newly divided GMCs or in very young guard cells. However, we found high levels of ProCDKB1;1:CDKB1;1-GFP fluorescence in newly formed guard cells in wild-type plants despite FLP/MYB88 presence. This could be explained if the normally operating proteasomal destruction mechanisms were overwhelmed, but it was also seen in cdkb1;1 cdkb1;2 plants complemented by this translational fusion, suggesting that FLP/MYB88 probably do not block cell cycle progression by acting on CDKB1;1 alone. Additionally, negative regulation of the G2/M transition alone through the repression of CDKB1;1 expression by FLP/MYB88 is insufficient to explain fully the clustered stomata phenotype of flp mutants, since this proliferation suggests an involvement of FLP/MYB88 in regulating other cell cycle checkpoints including the G1/S phase transition.

Further Cell Cycle–Related Targets of FLP/MYB88

Together, these data suggest that FLP/MYB88 regulation of CDKB1;1 is insufficient to explain its ability to restrain additional GC divisions. We therefore used ChIP-chip to search for further FLP/MYB88 targets and showed that the most significantly enriched GO category of genes bound by FLP encode cell cycle regulators, including CDC6a, CDC6b, CDKA;1, and CYCD4;1. Notably, these genes operate at the G1/S phase transition and several are expressed in GMCs (Figure 10), suggesting that FLP/MYB88 can modulate both the G2/M and G1/S phase transitions. CDKA;1 and CYCD4;1 can form protein complexes that trigger the G1/S transition, and mutations in CYCD4;1 reduce stomatal formation in the Arabidopsis hypocotyl (Kono et al., 2007). CDC6 function is required for DNA prereplication complex assembly, and ectopic expression increased the stomatal index (Castellano et al., 2004).

CYCT1;1, the activating cyclin partner of CDKC (Fulop et al., 2005; Kitsios et al., 2008), was identified in all three ChIP-chip experiments. Arabidopsis CDKC2 combines the functions of the human CDK9/CYCT complex of the positive transcription elongation factor-b (P-TEFb) (Kohoutek, 2009) with those of the mammalian CDK-related CRK7 (Kitsios et al., 2008), which is proposed to link transcription with splicing. Since P-TEFb plays a critical role in converting RNA polymerase II (RNAP II) to a processive enzyme through phosphorylation of the C-terminal domain of RNAP II and negative elongation factors, this raises the intriguing possibility that FLP/MYB88 further control gene expression through CYCT;1-mediated modulation of transcription or splicing activity.

We therefore propose that FLP/MYB88 allow GMC division and guard cell differentiation to take place and then stop further cell cycle progression by inhibiting the genes involved in the G1-to-S phase transition and essential components of the prereplication complex (CDC6a and CDC6b), possibly acting also through modulation of RNAP II transcription and/or splicing. Taken together, our results show that FLP/MYB88 target multiple cell cycle–related genes for regulation, in addition to CDKB1;1.

Modulation of CDKB1;1 Expression

The cell cycle–regulated expression of CDKB1;1 is activated by E2F (Boudolf et al., 2004b) and presumably by other regulators as well. We show that CDKB1;1 is an immediate direct target of the FLP/MYB88 proteins and identify a specific cis-regulatory element in the CDKB1;1 promoter (m1 in Figure 6) that conforms to the FLP/MYB88 DNA binding consensus. This cis-regulatory element is necessary for FLP/MYB88-mediated downregulation of the CDKB1;1 promoter. Strikingly, the FLP/MYB88 binding site in CDKB1;1 overlaps with a regulatory element previously proposed to be involved in the E2Fa-DP–mediated transcriptional activation of CDKB1;1 (Boudolf et al., 2004b). We confirmed by ChIP that E2Fa, and potentially other similar E2F factors, can recognize this element in vivo (Figures 7A and 7B).

Since the FLP and E2Fa binding sites overlap in the CDKB1;1 promoter, it is possible that binding of FLP may interfere with E2Fa-mediated CDKB1;1 activation (Boudolf et al., 2004b), or they might bind together at the CDKB1;1 promoter. In Drosophila melanogaster, a complex called dREAM-Myb/MuvB, containing E2F, the retinoblastoma protein, and MYB-interacting proteins, represses transcription of S phase genes by binding to deacetylated histone H4 associated with the corresponding promoters (Korenjak et al., 2004). During G0, the human equivalent of the dREAM complex associates with the E2F4 repressor so that the complex represses cell cycle genes, while in S phase it recruits B-MYB, converting the complex to an activator of cell cycle genes (Blais and Dynlacht, 2007). It is possible that FLP/MYB88 participate in a similar complex with E2Fa or other Arabidopsis E2F proteins to activate or repress cell cycle genes at different cell cycle stages. Alternatively, FLP/MYB88 might compete with E2F for binding to overlapping sites in the corresponding promoters, such as those identified in CDKB1;1.

Wider CDKB1 Functions in Stomatal Development

In addition to promoting symmetric division, CDKB1;1 and CDKB1;2 positively regulate stomatal number since the cdkb1;1 cdkb1;2 double mutant, like N161, has fewer stomata. Stomatal number depends partly upon the size of the pool of 2C epidermal cells capable of entering the cell lineage (Bergmann and Sack, 2007). N161 promotes endoreduplication in leaf epidermal cells (Boudolf et al., 2004b); thus, cdkb1;1 cdkb1;2 might reduce stomatal number by decreasing the pool of available 2C cells due to precocious endoreduplication. While neither ProCDKB1;1- nor CDKB1;2-driven GFP fluorescence was observed in 2C epidermal cells, both might still be expressed at low levels as previously reported for ProCDKB1;1-GUS (Boudolf et al., 2004a). This possibility is supported indirectly by the observation that cdkb1;2 potentiates GMC cytokinesis defects (in a cdkb1;1 cdkb1;2 double mutant background) even though GMCs show no ProCDKB1;2:GFP fluorescence. Thus, the reduction in stomatal number in N161 and in the double mutant might result if CDKB1 genes were normally expressed in young nonstomatal epidermal cells at low but functional levels sufficient to restrict endoreplication.

In addition to promoting stomatal number, CDKB1;1 and CDKB1;2 are also required for ectopic symmetric divisions in a flp/myb88 background, since a flp-1 myb88 cdkb1;1 cdkb1;2 quadruple mutant mostly lacks stomatal clusters. One possible mechanism is that the GMC precursor cells that fail to divide instead develop into single guard cells and represent the same pool of precursors that would otherwise be susceptible to cluster formation. If so, then the primary role of these CDKs might be to promote G2-to-M phase and a normal GMC division. However, given the number of normal symmetric divisions that still take place in this quadruple mutant, it is perhaps more likely that these CDKs play additional roles, such as in promoting the entry into S phase, as well as G2/M (Inze and De Veylder, 2006).

We also note that, although the cdkb1;1 cdkb1;2 double mutant shows many defective stomata, others appear normal, suggesting additional partial redundancy, perhaps with CDKA;1 or the two genes of the related CDKB2 type, themselves required for shoot meristem activity (Andersen et al., 2008). Thus, unlike the CDKB2 genes, which are important for normal cell cycle progression, CDKB1 genes might not be essential for wider Arabidopsis development.

FLP/MYB88 DNA Binding Preferences and MYB Evolution

FLP/MYB88 belong to an atypical subgroup of R2R3-MYB proteins (Lai et al., 2005). FLP and MYB88 are positioned phylogenetically between a group of broadly distributed R1R2R3-MYB proteins (also known as 3R-MYBs) and plant-specific typical R2R3-MYB proteins (Braun and Grotewold, 1999; Nakagawa et al., 2007), and FLP/MYB88 share characteristics with both groups of MYBs. Despite having two MYB repeats, FLP/MYB88 resemble 3R-MYB proteins in lacking the Leu residue located between the second and third α-helices of the second repeat (R2), a residue present in typical plant R2R3-MYB proteins. Unlike both 3R- and R2R3-MYB proteins, the third α-helix of the third MYB repeat (R3), which is likely to be involved in making DNA contacts, is not conserved in FLP/MYB88. While this might suggest that FLP/MYB88 do not bind DNA, our results show that FLP/MYB88 bind DNA in a sequence-specific fashion (Figure 2). However, the DNA binding specificity of FLP/MYB88, characterized by a [A/T/G][A/T/G]C[C/G][C/G] core consensus (Figure 2B), is distinct from that of any other reported plant, fungal, or animal MYB protein. Together with the finding that the E84K mutation in R3 abolishes DNA binding, these results suggest that, as is the case for 3R- and R2R3-MYB domains, at least two MYB repeats are necessary to efficiently contact DNA. The Arabidopsis ASYMMETRIC LEAVES1 (AS1) protein and its orthologs, PHANTASTICA and ROUGH SHEATH2, also harbor atypical R2R3-MYB domains with divergent DNA contact α-helices. However, unlike FLP/MYB88, AS1 only binds DNA in cooperation with the LOB domain protein AS2 (Guo et al., 2008), suggesting that these atypical R2R3MYBs control gene expression by distinct mechanisms.

CONCLUSION

The final stage of stomatal formation is a highly coordinated symmetric division that follows on from several asymmetric divisions. It is preceded by the differentiation of the meristemoid cell into the GMC. This symmetric division is dependent on CDKB1 function, and following this, the resulting guard cells must undergo cell cycle arrest and differentiate, which requires FLP function. FLP is a developmental regulator that integrates stage-specific cell cycle and cell fate cues in time and space, and it does so in a specific, critical phase of the stomatal cell lineage by controlling multiple cell cycle genes, including CDKB1;1. It therefore regulates the proliferation/exit balance and acts as an overall coordinator of cell cycle exit in a precisely timed manner.

METHODS

Plant Materials

The mutants flp-1, Pro35S:N161, and myb88 (SALK_068691 T-DNA insertion line) are all in a Col-0 background. The myb88 allele was obtained from the Arabidopsis Biological Resource Center (ABRC; Columbus, OH). Plants of flp-1 are also homozygous for glabrous1-1. Seeds of the enhancer trap line ET1728 were obtained from the ABRC, as were seeds harboring T-DNA insertions in CDKB1;1 and CDKB1;2.

Seeds harboring ProCDKB1;1:GUS and Pro35S:N161 were gifts of Lieven De Veylder and Dirk Inze (VIB, Ghent, Belgium). Seeds harboring ProCDC6a:GUS and of ProCDKA;1:YFP-DB were obtained from Crisanto Gutierrez and Arp Schnittger, respectively. The latter construct was fused to the CYCLINB1;1 DB, which normally confers CYCLIN degradation during cell division, and the CDKA promoter sequence in this construct is sufficient to drive expression of the CDKA;1 cDNA that rescues the cdka;1 mutant.

Reporter Constructs and Histology

Fluorescence images of living tissues were obtained with Nikon Eclipse 80i and Zeiss Pascal LSM5 Axioimager laser scanning confocal microscopes using 488- and 543-nm laser lines to excite GFP and propidium iodide, respectively. ProCDC6A:GUS staining was detected as by Malamy and Benfey (1997) except that leaves were cleared for 2 to 7 d.

To make Pro35S:E2Fa-GFP transgenic lines, E2Fa (At2g36010) cDNA was cloned into a pENTR Gateway vector (Invitrogen), mobilized into a pGWB5 binary destination vector, and then Agrobacterium tumefaciens was transformed into plants using a floral dip method (Karimi et al., 2002).

The ProCDKB1;1:GFP and ProCDKB1;2:GFP transcriptional fusions were generated via PCR amplification of ~0.5 and 2.9 kb, respectively, of sequence upstream of the start codon (primers shown in Supplemental Table 2 online). This was followed by cloning the PCR products into pENTR/TOPO (Invitrogen) and then recombination into the pGWB4 destination binary vector (Karimi et al., 2002). To generate the ProCDKB1;1:CDKB1;1-GFP construct, the protein-coding sequence of CDKB1;1 was amplified by PCR using the RT-PCR product as template and cCDKB1;1-F and cCDKB1;1-R as primers (see Supplemental Table 2 online). The 0.93 kb of the CDKB1;1 cDNA was cloned into NcoI and AscI sites at the 3′ region of the promoter of CDKB1;1 in pENTR/TOPO. The GFP transcriptional fusions of CDKB1;1 and CDKB1;2 and the translational fusion of CDKB1;1 were transformed into wild-type and cdkb1;1 cdkb1;2 double mutant plants (see below; Agrobacterium strain GV3101). Transgenic lines were selected on half-strength Murashige and Skoog medium containing 25 μg/mL hygromycin. Reporter expression and transgene complementation were verified in at least 15 independent transformants for each construct in segregating T2 lines.

Clones containing the full-length coding regions of FLP (ATTS5837) and MYB88 (U84152) were obtained from the ABRC. They were introduced into the pENTR vector using the GATEWAY system (Clontech) to make pENTR-FLP and pENTR-MYB88 and subsequently mobilized into the destination vector pGWB5 (Karimi et al., 2002) containing an N-terminal GFP tag using Gateway LR clonase (Invitrogen) to obtain Pro35S:FLP-GFP (pGWB5-FLP) and Pro35S:MYB88-GFP (pGWB5-MYB88). These constructs were then transformed into Agrobacterium strain GV3101.

To make NHis₆-FLP, NHis₆-MYB88, NHis₆-FLP^MYB, NHis₆-FLP^MYBE84K, FLP, and MYB88, the MYB domain of FLP (amino acids 23 to 125) and the MYB domain of FLP^E84K cDNA were amplified using PCR. After digestion with EcoRI/SalI and gel purification, the correct fragments were ligated into the Escherichia coli protein expression vector pET24b(+) predigested with EcoRI/SalI and transformed into E. coli DH5α competent cells (Invitrogen), and proteins were expressed in BL21-DE3 cells.

Verification of Mutants and Mutant Combinations

The flp-1 myb88 and the flp-7 myb88 double mutants were generated and verified as described (Lai et al., 2005). To make flp-1 Pro35S:CDKB1;1•N161 double mutants, transgenic Pro35S:CDKB1;1•N161 plants were confirmed by PCR using specific primers corresponding to the cauliflower mosaic virus 35S promoter and to the CDKB1;1 gene, and flp-1 homozygous plants were confirmed by derived cleaved amplified polymorphic sequences (Neff et al., 1998, 2002). After PCR using FLP-specific primers FLP-2F and FLP-20R, the 162-bp PCR product was restricted with PvuII, which cuts the wild-type FLP sequence into 135- and 27-bp fragments (see Supplemental Figure 5C online) but does not affect flp-1 due to a mutated restriction site. Thus, flp-1 homozygous plants show only a 126-bp fragment, whereas heterozygous plants show two smaller ones.

The cdkb1;1 alleles analyzed were T-DNA lines SALK_120549, SALK_044766, and SALK_073457. The latter insertion is in exon 4 and showed a 90% downregulation of expression using qRT-PCR (see Supplemental Figure 5A online). The cdkb1;2 alleles used were lines SALK_133560, SALK_069377, and SALK_046497 (see Supplemental Figure 5B online). A double mutant was selected from F2 progeny from a cross between SALK_073457 (cdkb1;1) and SALK_133560 (cdkb1;2), which has an insertion in exon 1. The double mutant was confirmed by PCR-based genotyping using primer sequences obtained from iSECT tools (www.salk.signal.edu). To determine the expression levels of cdkb1;1 and cdkb1;2 alleles, gene-specific primers were designed from the 5′-UTR region for CDKB1;1 and the 3′-UTR region for CDKB1;2 (see Supplemental Table 2 online). Total RNA was extracted (using an RNeasy plant mini kit; Qiagen) from 12-d-old Col-0 gl1, cdkb1;1, and cdkb1;2 seedlings.

Transient Expression in Nicotiana benthamiana Leaves

Three-week-old N. benthamiana plants grown at 30°C in greenhouse conditions were used for transient expression, and the young leaves were used for agroinfiltration. Agrobacterium containing Pro35S:FLP-GFP and Pro35S:MYB88-GFP were infiltrated into the tobacco leaves following established methods (Llave et al., 2000). Briefly, agrobacteria containing different binary constructs were grown at 29°C overnight in Luria-Bertani medium with 20 μM acetosyringone and 10 mM MES. After harvesting by centrifugation, the cells were resuspended in media containing MgCl₂, MES, and acetosyringone to an OD₆₀₀ ~1.0. After 3 to 4 h at room temperature, the bacteria containing either Pro35S:FLP-GFP, Pro35S:MYB88-GFP, or Pro35S:GFP (corresponding to the pGWB5 vector) were infiltrated into leaves from 3- to 4-week-old Nicotiana plants. Pro35S:GFP was used as a positive control for cytoplasmic and nuclear localization. Three days after agroinfiltration, the inoculated leaves were observed for green fluorescence signal using confocal laser scanning microscopy using ×40 and ×100 magnifications with the same gain.

SELEX

SELEX was performed, with some modifications, as previously described (Grotewold et al., 1994). Briefly, a double-stranded oligonucleotide library containing a 26-nucleotide random sequence in the center that was flanked by two 27-nucleotide constant sequences was incubated with NHis₆-FLP^MYB and bound to Ni-NTA agarose beads. After washing, the protein-DNA complex was eluted from the beads using 50 mM imidazole. The eluted DNA was amplified by PCR using the primers corresponding to the flanking sequences and the selection was repeated. After seven rounds of selection, the DNA was 5′ radiolabeled with [γ-³²P]ATP and used in EMSA experiments. The DNA in the shifted band was cut and cloned into pBluescript, and individual clones were confirmed by EMSA and sequenced. The sequences were aligned manually and using MotifSampler (Thijs et al., 2001). The Weblogo package (Crooks et al., 2004; http://weblogo.berkeley.edu/logo.cgi) was used for visualization.

A series of SS-30 mutants were designed as indicated in Figure 2C. To make double-stranded DNA, equal molar ratios of both strands for each oligo were mixed and boiled for 5 min, followed by cooling down to room temperature. For EMSA experiments, 200 ng of the purified MYB domain of FLP (FLP^MYB) were incubated with radioactive-labeled SS-30 at 4°C for 30 min, followed by the addition of 0×, 50×, or 500× (molar ratio to radioactive-labeled SS-30 probe) cold SS-30 mutant probes and incubated at 4°C for another 30 min. Each sample was run side-by-side on 8% acrylamide PAGE, dried, and exposed at −70°C overnight.

For EMSA and the mutation assay in CDKB1;1, NHis₆-FLP^MYB was expressed in E. coli and affinity purified using Ni-NTA beads. The CDKB1;1 promoter fragments used for EMSA were generated by PCR using the following primer pairs: P1, 5′-GAATTGTAGCTTTCAAAAAATG-3′ and 5′-GAAAACTAATTGGACTCACTT-3′; P2, 5′-CCCAAAAACATTCACAGAG-3′ and 5′-TTCTGAGAGGTTTCGTAAAATTG-3′, in which one of the primers was radioactively labeled with [γ-³²P]ATP. NHis₆-FLP^MYB (30 ng) was used for each EMSA reaction. The EMSA experiments were as described (Grotewold et al., 1994). The MYB domain of v-MYB from vertebrates and the MYB domain of P1 from maize (Zea mays) were expressed in E. coli and induced by 1 mM isopropyl β-d-1-thiogalactopyranoside. After purification, 30 ng of each protein was used for EMSA. The APB1 and Mim-1 oligo probes were radiolabeled.

To test the binding of the MYB domain of the flp-8-like protein to DNA (SS-30), equal amounts (30 ng) of MYB domain proteins from the wild type and flp-8 were used and quantified by protein gel blot using a FLP/MYB88 rabbit polyclonal antibody.

ChIP

Antibodies against the FLP/MYB88 proteins were generated by inoculating rabbits with Ni-NTA affinity-purified NHis₆-MYB88. ChIP experiments were performed as described (Morohashi et al., 2007) with some modifications. Briefly, 10-d-old seedlings were fixed with 1% formaldehyde for 30 min, and then tissues were ground and suspended in lysis buffer (50 mM HEPES, pH 7.5, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, 0.1% deoxycholate, 0.1% SDS, 1 mM PMSF, 10 mM Na-butyrate, and protein protease inhibitor). An average DNA size of 0.3 to 1 kb was obtained using sonication. After clearing using salmon sperm DNA/protein-A agarose for 1 h, the aliquot supernatant fractions were incubated with IgG, FLP antibody, or acetylated histone H3 at K9 and K14 antibodies at 4°C overnight. The chromatin antibody complex was precipitated with salmon sperm DNA/protein-A agarose, washed with lysis buffer, LNDET buffer (0.25 M LiCl, 1% Nonidet P-40, 1% deoxycholate, and 1 mM EDTA), and TE buffer, and the complex was reverse cross-linked by elution buffer (1% SDS, 0.1 M NaHCO₃, and 1 mg/mL proteinase K) overnight at 65°C. DNA was extracted by a Qiagen PCR purification kit. The primers used for PCR were ProCDKB1;1-F and ProCDKB1;1-R.

To test whether E2Fa binds to the promoter of CDKB1;1, a ChIP assay was performed using an anti-GFP polyclonal antibody (Ab290; Abcam) on extracts from 10-d-old wild-type (control) plants and from plants harboring a Pro35S:E2Fa-GFP transgene. To test whether E2Fa binds to the wild-type CDKB1;1 promoter or to a mutated (m1) CDKB1;1 promoter, an anti-GFP antibody was used in ChIP experiments with N. benthamiana plants that were agroinfiltrated with Pro35S:E2Fa-GFP along with either ProCDKB1;1-WT:GUS or ProCDKB1;1-m1:GUS. PCR was performed using pCDKB1;1-F and pCDKB1;1-R primers (see Supplemental Table 2 online).

To validate candidate target genes obtained from ChIP-chip experiments, PCR primers were designed that recognized putative binding sites by FLP based upon ChIP-chip data (see Supplemental Table 3 online).

Arabidopsis thaliana Bombardment

To create the Pro35S:proCDKB1;1WT:Luc and Pro35S:proCDKB1;1m1:Luc constructs, PCR was used to clone ProCDKB1;1 (wild type or m1) into destination vector PGWB5, and then Pro35S:proCDKB1;1WT or Pro35S:proCDKB1;1m1 was cloned into the vector PJD301 using BamHI/NcoI. One microgram of plasmid ProFLP:gFLP (genomic FLP DNA) in PCAMBIA 2300 or 1 μg of Pro35S:MYB88 and 3 μg Pro35S:proCDKB1;1WT:Luc or Pro35S:proCDKB1;1m1:Luc were used for bombardment into 10-d-old wild-type Arabidopsis plants grown on the half-strength Murashige and Skoog plates. Three micrograms of Pro35S:Rellina were used in each sample as an internal control for normalization. The bombardment protocol was as described (Hernandez et al., 2004). Three biological replicates were conducted, and each treatment was repeated at least twice. Data analysis and assays for luciferase and Rellina were as described by Grotewold et al. (1994).

ChIP-Chip Experiments and Analysis

ChIP-chip analyses were performed as previously described (Keles et al., 2006) with some modifications. ChIP-chip high-density oligonucleotide array signal intensities were first normalized using quantile normalization. The logarithms of the enrichment (compared with input) were calculated for both the wild type and flp-1 myb88. The scan statistics based on the differences between these genotypes were used in multiple testing to identify peaks. A sliding window with an approximate size of ~350 bp (corresponding to 10 probes) was used in scan statistics to account for the spatial structure of the data. Simulations based on array signal intensities identified critical values that control the family-wise error rate at 5%. A nested-Bonferroni multiplicity adjustment and twofold change criteria were used to infer the location of transcription factor binding sites. Only peaks overlapping in at least two biological replicates were used. FLP/MYB88 binding positions on the five chromosomes were visualized using the Integrated Genome Browser from Affymetrix; binding positions were calculated relative to the transcription start sites. FLP/MYB88 putative direct target genes from ChIP-chip were classified according to their putative functions using the Munich Information Center for Protein Sequence. Conserved motifs selected in the FLP/MYB88 ChIP-chip experiments (see Supplemental Table 4 online) were subjected to the enriched motif discovery algorithm MEME at http://meme.sdsc.edu/meme/intro.html (Bailey et al., 2006). ChIP-chip results have been deposited in the Genome Expression Omnibus of the National Center for Biotechnology Information under accession number GSE19763 at http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?token=hhilveaoqkmgatgandacc=GSE19763.

Real-Time PCR Analysis and RT-PCR

Total RNA was extracted from 10-d-old Col-0 gl1, flp-1, and flp-1 myb88 shoots using an RNeasy plant mini kit. RNA was treated with DNase (Promega) to remove any remnant genomic DNA. Using 1 μg total RNA, cDNAs were synthesized using an oligo (dT)₂₀ primer (Thermoscript RT-PCR dystem from Invitrogen). Real-time PCR based on SYBR-Green was performed using 36 ng cDNA as described. Real-time quantitative PCR was monitored over 50 cycles using a Bio-Rad iCycler. Gene expression levels are presented as the ratio relative to the wild type control using the 2^(−delta delta Ct) method and normalized to KAT1 expression, a marker for mature guard cells and therefore correcting for the increased number of guard cells present in flp and myb88 mutants. CDKB1;1 and CDKB1;2 gene-specific primers were used for semiquantitative RT-PCR using the same RNA material as in real-time PCR, under optimized RNA concentrations and PCR conditions. Three biological replicates were conducted for each real-time PCR experiment.

Quantitative PCR experiments were performed with DNA obtained by ChIP using SYBR-Green chemistry and with primers At3g54180-F and At3g54180-R (see Supplemental Table 3 online), which include the CDKB1;1 (−214; −31) promoter region. Each quantitative PCR reaction was performed in duplicate with the values averaged, and data were obtained from biological replicates, with DNA extracted from different plants of the same genotype. The fold enrichment for each sample was calculated by dividing the quantitative PCR/ChIP results using FLP/MYB88 antibodies by those using the IgG control.

Accession Numbers

Sequence data from this article can be found in the Arabidopsis Genome Initiative or GenBank/EMBL databases under the following accession numbers: FLP (At1g14350), MYB88 (At2g02820), CDKB1;1 (At3g54180), CDKB1;2 (At2g38620), E2Fa (At2g36010), CDC6a At2g29680), CDC6b (At1g07270), CDKA;1 (At3g48750), CDKD1;3 (At1g18040), CYCB1;3 (At3g11520), CYCD4;1 (At5g65420), CYCT;1 (At1g35440), and full-length coding regions for FLP (ATTS5837) and MYB88 (U84152). Online data showing likely direct targets of FLP/MYB88 are available at AGRIS (Arabidopsis Gene Regulatory Information Server; http://Arabidopsis.med.ohio-state.edu/) (Palaniswamy et al., 2006) by following the “Direct Targets” link at http://Arabidopsis.med.ohio-state.edu/AtTFDB/tfsummary.html?locusid=At1g14350.

Supplemental Data

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

• Supplemental Figure 1. Nuclear Localization of FLP and MYB88.

• Supplemental Figure 2. FLP Binding Motif Identified by SELEX.

• Supplemental Figure 3. Residue in FLP Required for DNA Binding.

• Supplemental Figure 4. FLP/MYB88 Negatively Regulate CDKB1;1 Expression by Binding to a Specific Promoter Sequence.

• Supplemental Figure 5. Positions of T-DNA Insertions in CKDB1;1 and CDKB1;2, Expression Levels, and Double Mutant Genotyping.

• Supplemental Figure 6. Genome-Wide Distribution of Binding Sites and Target Genes Identified Using ChIP-Chip.

• Supplemental Figure 7. ChIP Validation of Non-Cell Cycle, Putative FLP/MYB88 Target Genes.

• Supplemental Figure 8. Cell Cycle Genes Are Enriched among FLP/MYB88 Targets.

• Supplemental Table 1. Single Guard Cell Phenotypes and Genetic Relationships: CDKB1 Genes Promote Symmetric Division and Are Genetically Downstream of FLP.

• Supplemental Table 2. Primers Related to CDKB1;1 and CDKB1;2.

• Supplemental Table 3. Primers Used in ChIP Validation of Targets from ChIP-Chip.

• Supplemental Table 4. Conserved Motifs in FLP/MYB88 ChIP-Chip–Enriched Sequences.

• Supplemental Data Set 1. FLP/MYB88 Putative Direct Targets.