Isolation of fsp1, a Suppressor of flp-1 in Stomatal Development.

The flp-1 mutant is featured by extra terminal divisions during stomatal development, suggesting the role of FLP in restricting cell division (3). To identify new genetic players in regulating the one-time terminal division, we created an ethylmethane sulfonate-mutagenized M2 population of flp-1 mutants and screened for mutants with altered stomatal phenotypes. flp-1 fsp1 (flp-1 suppressor1) was isolated for significantly reduced stomatal clusters, compared to flp-1 (Fig. 1 A–C and SI Appendix, Table S1). In addition, aberrant cells were occasionally found in the epidermis of the flp-1 fsp1 double mutant (Fig. 1C). Using the GC fate marker E1728, we confirmed that the aberrantly shaped single cells in fsp1 have the GC identity (Fig. 1 D, Inset), resembling the Single Guard Cell (SGC) phenotype in cdkb1;1 1;2 and cyca2s mutants (4, 5). The Stomatal Index (SI) of the fsp1 mutant was reduced as well, indicating that FSP1 promotes stomatal production (SI Appendix, Table S2).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

FSP1 encodes the RPA2a subunit required for stomatal precursor cell divisions. (A–D) Differential interference contrast images of epidermis of wild-type (A), flp-1 (B), flp-1 fsp1(C), and fsp1 (D). Brackets indicate the stomatal clusters. SGCs are highlighted with yellow color and normal GCs with blue color. (Inset in A and D) Expression of the mature GC marker E1728 in a SGC. (E) Map-based isolation and gene structure of FSP1. The mutation site in fsp1 is the same as ror1-2. rpa2a-5 (ror1-3) is a T-DNA insertion line. (F and G) RPA2a:RPA2a-GFP rescued the fsp1 SGC phenotype (F) and induced the reappearance of flp-1 clusters in flp-1 fsp1 (G). (H and I) flp-1 myb88 cluster numbers and size are repressed by fsp1 mutation. (J) RPA2a-GFP expression profile in the stomatal linage cells. (Scale bars, 20 µm.)

FSP1 Encodes the Arabidopsis RPA2a Subunit of the RPA Complex.

Using map-based cloning, we determined that mutation of the FSP1 gene was located within the bacterial artificial chromosome clone T28124 on chromosome 2. Based on growth defects reminiscent of rpa2a/ror1-1 mutants, such as dwarf seedlings, narrow leaves, and early flowering (SI Appendix, Fig. S1A) (19, 20), we amplified and sequenced the open reading frames of RPA2a from fsp1 and found that fsp1 possessed the same point mutation as in ror1-2, ag (1343) changed to aa, which induces altered splicing events (Fig. 1E). The expression of RPA2a (cDNA) fused with GFP under the control of the native promoter fully rescued the stomatal defects of fsp1 mutants (Fig. 1F and SI Appendix, Table S2). The functionality of RPA2a was further confirmed by the reappearance of flp-1–featured stomatal clusters in flp-1 fsp1 mutants carrying RPA2a:RPA2a-GFP (Fig. 1G and SI Appendix, Table S1). In addition, the introgression of the fsp1 mutation in flp-1 myb88 double mutants largely repressed the size and number of stomatal clusters (Fig. 1 H and I and SI Appendix, Fig. S1B). We also revisited the T-DNA allele rpa2-5/ror1-3 (Fig. 1E and SI Appendix, Fig. S1A) (19, 20), which displayed a similar stomatal phenotype as fsp1. The flp-1 stomatal phenotype was suppressed in flp-1 rpa2-5 double mutants (SI Appendix, Fig. S1 C and D and Table S2). Benefiting from easy genotyping, rpa2-5 was then used in most of the later experiments (renamed as rpa2a-5).

RPA2a Is Expressed in Specific Stomatal Lineage Stages.

A complemented ror1-1 transgenic line harboring ROR1:gROR1-GUS-GFP (19) (here referred to as RPA2a-GFP) was used to investigate the RPA2a expression pattern and localization. RPA2a-GFP fluorescent signals were observed in a subset of stomatal lineage cells and compared with those of three translational reporters of SPEECHLESS (SPCH), MUTE, and FLP, which have distinct and sequential expression patterns during stomatal development (Fig. 1J). The expression of RPA2a-GFP overlapped with SPCH-GFP (22) in MMCs prior to asymmetric entry divisions but not in both newly formed meristemoids (M) and stomatal lineage ground cells after division. In late meristemoids and early GMCs (EGMCs), where MUTE-GFP was turned on (23), diffuse RPA2a-GFP signals reappeared. RPA2a-GFP persisted in late GMCs (LGMCs), but disappeared in young GCs (YGCs) after the terminal symmetric cell division. The expression of RPA2a overlapped with FLP-GFP (3, 24) only at the LGMC stage, prior to the GMC division, but not in YGCs. Taken together, RPA2a shows a cell type- and time-specific expression profile, with the preferences in the precursor cells (actively dividing) prior to either asymmetric or symmetric divisions during stomatal development.

RPA2a Functions Together with FLP Downstream Target Genes CDKB1s and CYCA2s.

The CDKB1;1 gene is one of the direct transcriptional targets of FLP in controlling stomatal terminal symmetric divisions. In agreement with this, expression of CDKB1;1-GFP (24) overlaps with FLP-GFP in LGMCs and YGCs and partially with RPA2a-GFP in LGMCs (Fig. 1J). In comparison to SGC frequencies of 42.0 ± 2.6% in cdkb1;1 1;2 and 8.0 ± 2.1% in rpa2a-5, the rpa2a-5 cdkb1;1 1;2 triple mutant produced dramatically more SGCs (64.0 ± 1.6%, Fig. 2 A–D). GMC divisions also require CYCA2 activity (5). While SGCs in cyca2;3 mutants were found at a low frequency of 4.0 ± 0.8% (Fig. 2E), the occurrence of SGCs in rpa2a-5 cyca2;3 double mutants was elevated to 47.0 ± 2.0% (Fig. 2F). Consistently, the frequency of SGCs in cyca2;3 2;4 (27.6 ± 3.4%) was greatly elevated to 78.0 ± 8.5% in rpa2a-5 cyca2;3 2;4 triple mutants (Fig. 2 G and H), indicating that interaction between RPA2a and CDKB1s/CYCA2s functions in GMC divisions.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

RPA2a functions synergistically with CYCA2s and CDKB1s in GMC division. (A–G) Micrographs of epidermis of wild-type (A), rpa2a-5 (B), cdkb1;1 1;2 (C), rpa2a-5 cdkb1;1 1;2 (D), cyca2;3 (E), rpa2a-5 cyca2;3 (F), cyca2;3 2;4 (G), and rpa2a-5 cyca2;3 2;4 (H). SGCs are highlighted with yellow color and normal GCs with blue color. The numbers at the right-upper corner indicate the percentage of SGCs in total stomata. (I–N) DAPI fluorescence micrographs. Red arrowheads indicate the nuclei. (Scale bar, 20 µm.)

An ectopic and prolonged expression of RPA2a-GFP was occasionally found in two daughter cells after a GMC division in flp-1, suggesting that RPA2a is required for subsequent GMC divisions that end up with two stomata next to each other (SI Appendix, Fig. S2 A–H). RPA2a transcription level slightly but not significantly increased in flp-1 myb88 (SI Appendix, Fig. S2I). Although 12 putative FLP/MYB88 binding sequences are present within the RPA2a promoter (4⇓–6), neither FLP nor MYB88 showed clear binding activity to RPA2a promoter according to the results of yeast one-hybrid assays (SI Appendix, Fig. S2 J and K), indicating that RPA2a might not be a direct transcriptional target of FLP/MYB88.

Combined Loss of RPA2a and CDKB1s/CYCA2s Function Induces Abnormal Cell Enlargement and Endoreduplication.

Differing from the interlocking, continuously expanding pavement cells, the mature stomatal GCs, once matured, retain their shape and size for the life time in leaves. The area of the paired GCs ranged mostly from 200 to 500 µm² in Arabidopsis (SI Appendix, Fig. S3). However, about 45 and 71% stomata in rpa2a-5 cdkb1;1 1;2 and rpa2a-5 cyca2;3 2;4 triple mutants, respectively, are larger than 500 µm² (Fig. 2 A–H and SI Appendix, Fig. S3). It has been widely accepted that cell size and differentiation are closely correlated to the nuclear DNA levels (25). We therefore measured the nuclear DNA levels in SGCs after DAPI (4′,6-diamidino-2-phenylindole) staining. The DNA content in wild-type GCs was defined as 2C-DNA level. Our results showed that SGCs in the rpa2a-5, cdkb1;1 1;2, and cyca2;3 2;4 mutants all displayed a similar DAPI fluorescence intensity (4C-DNA level), indicating that RPA2a, similar to CDKB1s and CYCA2s, is required for the mitosis of GMC divisions. But in the rpa2a-5 cdkb1;1 1;2 or rpa2a-5 cyca2;3 2;4 triple mutants, the enlarged SGCs contained single giant nuclei with a mean DNA level of 6C-8C (Fig. 2 I–N). The analysis of DNA ploidy distribution in 9- and 14-d-old cotyledons using flow cytometry confirmed that RPA2a may coordinately act with CDKB1s/CYCA2s to restrict the onset of endocycle (SI Appendix, Fig. S4).

The above results indicated that the mitosis was blocked but the endoreplication was inappropriately unrestricted in the rpa2a-5 cdkb1;1 1;2 triple mutants. In the rpa2a-5 single mutant, the expression levels of the G1-to-S genes CDT1, MCM2, ORC3, and PCNA1 were significantly increased compared to those in wild type. Most of these elevations, as well as CDC6a, CCS52A, ORC1a, CDKA;1, were additively enhanced in rpa2a-5 cdkb1;1 1;2 triple mutants (SI Appendix, Fig. S5A). In addition, the expression levels of genes during endoreplication initiation, such as E2Fa, E2Fc, and E2Ff, were largely increased in the triple mutants (SI Appendix, Fig. S5B). These results demonstrated that RPA2a and CDKB1s act together to suppress the expression of endocycling genes that restrict endoreduplication.

RPA2a Function Is Conservatively Regulated by CDKs through Protein Phosphorylation.

Supported by our genetic data above, we tested the hypothesis that RPA2a may directly interact with CDKs in regulating cell division and cell cycle. Results of Bimolecular Fluorescent Complementation (BiFC) assays using the tobacco transient expression system and of the Co-ImmunoPrecipitation (Co-IP) assay using the proteins that were extracted from the infiltrated tobacco leaves suggested that RPA2a directly interacts with CDKB1;1 and CDKA;1 (Fig. 3 A–D and SI Appendix, Fig. S6).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Phosphorylation of RPA2a by CDK complexes. (A–D) The RPA2a protein binds CDKB1;1 protein in BiFC assays (A–C) and Co-IP assays (D). YFP, yellow fluorescent protein; BF, bright field. (Scale bars, 20 µm.) (E) In vitro phosphorylation assays. RPA2a is phosphorylated by the Cak1-CDKA;1-CYCD3;1 complex in the form of the RPA1b-RPA2a-RPA3a complex. (F) RPA2a is also phosphorylated by Cak1-CDKB1;1-CYCB1;2 complexes. The bottom box shows the loading controls after Coomassie Brilliant Blue (CBB) gel staining. MBP and GST proteins were used as positive and negative controls, respectively. Red arrows indicate the RPA2a phosphorylation band in phosphorylation autoradiographs. (G) Phosphorylation of RPA2a by Cak1-CDKB1;1-CYCB1;2 complexes is confirmed by finding a delayed band in 12.5% SDS/PAGE. The extent of phosphorylation of the RPA2a proteins was reduced when Ser-11 and Ser-21 were substituted or was barely detected when the N terminus was deleted.

It was reported that coexpression of the RPA2 subunit with other RPA subunits could enhance the solubility and function of RPA protein complexes (26). When RPA2a-GST was coexpressed with RPA1b-HIS in Escherichia coli, RPA1b proteins failed to be pulled down by RPA2a. Only when the RPA3a was present, in which RPA2a-GST-RPA3a-S coexpressed with RPA1b-HIS, the whole RPA complex could be pulled down by GST beads, indicating that the preformation of the RPA2-RPA3 subcomplex is essential for the assembly of the RPA complex (SI Appendix, Fig. S7A). However, regardless of the presence of RPA3 and RPA1, RPA2a proteins could be phosphorylated in vitro by active CDK-CYCLIN protein complexes (27) (Fig. 3 E and F and SI Appendix, Fig. S7B). Additionally, a faint delayed band next to the RPA2a-GST was observed when the products of the CDK phosphorylation reaction were separated on SDS/PAGE gels (Fig. 3G). Once alkaline phosphatases were added to the products, the delayed band disappeared, indicating that the phosphorylation is the cause of the slower mobility of RPA2a-GST proteins in gels (SI Appendix, Fig. S7B). Elimination of the N-terminal 32 amino acids abolished the delayed band, suggesting that this region contains the primary phosphorylation target sites (Fig. 3G).

To predict the CDK-phosphorylation target residues in RPA2a protein, the sequence of the N terminus of RPA2a protein was compared with those in human, yeast, and rice RPA2 proteins. Among the 12 Serine residues within the first 32 amino acids of the RPA2a N terminus, Ser-11 is the consensus CDK phosphorylation site, corresponding to Ser-23 of human RPA2 protein. Ser-21 is another predicted phosphorylation target site of CDKs (www.cbs.dtu.dk/services/) that may be conserved in plants (SI Appendix, Fig. S7 C–F). When Ser-11 and Ser-21 were substituted with nonphosphorylatable Glycines, the extent of the phosphorylated band was reduced (Fig. 3G), indicating that these two residues are predominant CDK-phosphorylation targets, but other sites in RPA2a might become phosphorylated in vitro as well.

To assess the function of phosphorylation sites in the RPA2a protein in planta, RPA2a variants were generated and introduced into the rpa2a-5 mutant to evaluate their ability to rescue stomatal defects. RPA2a-GFP and individual variants with a single mutation, RPA2a^S11/G-GFP, and RPA2a^S21/G-GFP could fully rescue the stomatal production and GMC division defects of the rpa2a-5 mutant (SI Appendix, Table S2). By contrast, further mutation of RPA2a^S11,21/G-GFP or the N-terminal deletion variant RPA2a^△32-GFP showed a reduced rescuing ability (Fig. 4 A and B and SI Appendix, Table S2). However, the phosphorylation-mimic variant RPA2a:RPA2a^S11,21/D-GFP (Ser-11 and Ser-21 were substituted with aspartic acids) rescued rpa2a-5 GMC division defects (Fig. 4C and SI Appendix, Table S2).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Rescue stomatal phenotypes by RPA2a variants. (A–C) The formation of SGCs in the rpa2a-5 mutant was partially rescued by the RPA2a^△32-GFP construct (A) or RPA2a^S11,21/G-GFP (B) and was mostly rescued by RPA2a^S11,21/D-GFP (C). (D and E) The formation of SGCs in flp-1 fsp1 was barely affected by RRPA2a^△32-GFP (D) or RPA2a^S11,21/G-GFP (E). (F–H) The formation of SGCs in the cdkb1;1 1;2 mutant is partially rescued by RPA2a-GFP (G), but is greatly repressed by RPA2a^S11,21/D-GFP (H). The numbers at the right-upper corner indicate the percentage of SGCs in total stomata. SGCs are highlighted with yellow color and normal GCs with blue color. (Scale bars, 10 µm.)

We further examined how RPA2a phosphorylation variants perform in flp-1 fsp1 double mutants. While the wild-type RPA2a-GFP fully complemented the loss function of RPA2a and restored the flp-1–like stomatal clustering, the expressions RPA2a:RPA2a^△32-GFP and RPA2a:RPA2a^S11,21/G-GFP led to a mild stomatal clustering phenotype (Fig. 4 D and E and SI Appendix, Table S1). RPA2a^S11/G-GFP and RPA2a^S21/G-GFP fully complemented the stomatal phenotype of flp-1 fsp1 (SI Appendix, Table S1). Therefore, the phosphorylation status of Ser-11 and Ser-21 within the N-terminal 32 residues is critical for the full function of RPA2a.

We introduced RPA2a-GFP, RPA2a:RPA2a^S11/D-GFP, RPA2a^S21/D-GFP, and RPA2a^S11,21/D-GFP into cdkb1;1 1;2 double mutants to assess the genetic interaction between RPA2a and CDKB1s in vivo. RPA2a^S11,21/D-GFP showed strong effects in reducing the formation of SGCs in cdkb1;1 1;2, while the remaining 3 transgenes had certain effects (Fig. 4 F–H and SI Appendix, Table S3). Therefore, these results suggest that the phosphorylation status of RPA2a is epistatic to CDKB1 mutations.

CDKB1 Activity Promotes RPA2a Nuclear Localization.

The strong expression of complementary RPA2a-GFP was found mainly in the nucleus, but also was present in the cytoplasm. RPA2a-GFP signals were more diffuse in cdkb1;1 1;2 epidermal cells (Fig. 5 A and B and SI Appendix, Fig. S8 A and B). Western blot analysis confirmed that the abundance of nuclear RPA2a-GFP proteins decreased in cdkb1;1 1;2 (Fig. 5C). RPA2a^Δ32-GFP and the phosphorylation-deficient version RPA2a^S11,21/G-GFP displayed pronounced fluorescent signals in the cytoplasm of epidermal cells (Fig. 5 D and E and SI Appendix, Fig. S8 C and D). By contrast, the signals of RPA2a^S11,21/D-GFP, which fully complemented the stomatal defects of rpa2a-5, displayed a high degree of colocalization with DAPI in the nucleus, with a Pearson’s Rr-value of 0.89 (Fig. 5F and SI Appendix, Fig. S8E). Quantitative analysis of relative GFP fluorescent intensity in the nucleus consistently supported that CDK-mediated phosphorylation of RPA2a promotes the nuclear accumulation of RPA2a (SI Appendix, Fig. S8K).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

The function of RPA2a is associated with its phosphorylation status and subcellular localization. (A) Nuclear expression of RPA2a-GFP in a complemented rap2a-5. (B) Diffuse RPA2a-GFP in cdkb1;1 1;2. (C) RPA2a-GFP proteins were found in both the cytoplasm and the nucleus by Western blot analysis. The amount of RPA2a-GFP in the nucleus is reduced in cdkb1;1 1;2. (D and E) Diffuse GFP signals in RPA2a^△32-GFP (D) and RPA2a^S11,21/G-GFP (E). (F) Predominant GFP signals in the nucleus of RPA2a^S11,21/D-GFP. (G–K) The RPA2a subcellular localization in RPA2a-NNLS-GFP (G), RPA2a-NLS-GFP (H), RPA2a-NNES-GFP (I), and two RPA2a-NES-GFP transgenic lines [line #12 (J) and line #5 (K)]. (Insets) The enlarged region highlighted by the white dashed-line boxes. (Scale bars, 20 µm.) (L) The retarded overall growth of rpa2a-5 was not fully rescued by RPA2a-NES-GFP. (Scale bar, 1 cm.)

We fused the Nuclear Localization Sequence (NLS) or Nuclear Export Signal (NES) to the C terminus of CDKB1;1 and coexpressed with RPA2a-mCherry in tobacco leaves. A Non-Nuclear Localization Sequence (NNLS, a control sequence to the NLS) and a Non-Nuclear Export Signal (NNES, a control sequence to the NES) (28) were also fused with CDKB1;1 as controls. When individually expressed, RPA2a-mCherry remained in the nucleus and cytoplasm (SI Appendix, Fig. S9A). However, RPA2a-mCherry signals were predominantly retained in the nucleus when coexpressed with CDKB1;1-NLS-GFP (SI Appendix, Fig. S9B). By contrast, when coexpressed with CDKB1;1-NES-GFP, in addition to the nucleus, abundant RPA2a-mCherry signals were accumulated in the cytoplasm (SI Appendix, Fig. S9C). CDKB1;1-NNLS-GFP or CDKB1;1-NNES-GFP did not change the subcellular localization pattern of RPA2a-mCherry (SI Appendix, Fig. S9 D and E). Thus, the physical interaction with CDKB1;1 either in the nucleus or in the cytoplasm seems to affect the subcellular localization of RPA2a.

To further clarify whether the subcellular localization is associated with the function of RPA2a, we generated NLS-fused RPA2a:RPA2a-NLS-GFP and NES-fused RPA2a:RPA2a-NES-GFP and transformed them into a rpa2a-5 mutant. The fluorescent signals of RPA2a-NLS-GFP were exclusively accumulated in the nuclei (Fig. 5H and SI Appendix, Fig. S8 G and L), in which the defective overall growth and GMC division of rpa2a-5 mutants were fully recovered (Fig. 5L and SI Appendix, Table S4). However, a transgenic line carrying RPA2a-NES-GFP line #12, in which the diffuse GFP signals were detected in both cytoplasm and nucleus (Fig. 5J and SI Appendix, Fig. S8 I and L), the growth and stomatal defects in rpa2a-5 were partially rescued (Fig. 5L and SI Appendix, Table S4). Whereas RPA2a-NES-GFP transgenic line #5, in which GFP signals were restricted mostly to the cytoplasm, barely showed a colocalization with DAPI in the nucleus and a representative low value of Pearson’s R at −0.23 (Fig. 5K and SI Appendix, Fig. S8 J and L). Neither the plant growth defect nor the stomatal defect of the rpa2a-5 mutant was rescued in line #5 (Fig. 5L and SI Appendix, Table S4). The high expression level of RPA2a-NES-GFP in line #5 (SI Appendix, Fig. S10) might provide higher nuclear export signals (from the NES), which drive RPA2a proteins into the cytoplasm against the endogenous nuclear “stay force,” such as being phosphorylated by CDKs. The control lines RPA2a-NNLS-GFP and RPA2a-NNES-GFP phenocopied the protein localization and complementation ability of the wild-type RPA2a-GFP (Fig. 5 G, I, and L and SI Appendix, Fig. S8 F and H and Table S4).

The N-Terminal Domain Is also Required for RPA2a’s Function in Response to DNA Damage.

It has been reported that Arabidopsis rpa2a alleles exhibited hypersensitivity to the genotoxic agent Methyl Methane Sulfonate (MMS) treatment (19⇓–21). However, how the RPA2a protein and its subdomains are related to the DNA damage responses in plants remains unknown. The 5-d MMS treatment generally arrested plant growth and induced seedling chlorosis. The introduction of RPA2a-GFP or RPA2a^S11,21/D-GFP recovered rpa2a-5 growth to the wild-type level. However, the N-terminal defective RPA2a^△32-GFP and the phosphorylation-deficient RPA2a^S11,21/G-GFP only partially complemented the chlorosis phenotype of rpa2a-5 (SI Appendix, Fig. S11). Furthermore, cdkb1;1 1;2 and rpa2a-5 cdkb1;1 1;2 displayed hypersensitivity to MMS treatment as well. The phosphorylation-mimic variant RPA2a^S11,21/D alleviated the sensitivity of cdkb1;1 1;2 to MMS (Fig. 6 A and B and SI Appendix, Fig. S12), which is consistent with its stronger capability than wild-type RPA2a in rescuing cdkb1;1 1;2 GMC divisions (SI Appendix, Table S3). Therefore, our finding observation is consistent with the previous findings in yeast and humans that the phosphorylation of the N terminus is important for RPA2 function in response to DNA damage.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 6.

Phosphorylation-mimic RPA2a^S11,21/D alleviated the MMS effects in cdkb1;1 1;2. (A and B) Seven-day-old seedlings were transferred to liquid Murashige–Skoog (MS) medium (A) or liquid MS medium containing 0.01% MMS (B) for 5 d before imaging. Seedlings of rpa2a-5, cdkb1;1 1;2, and rpa2a-5 cdkb1;1 1;2 displayed hypersensitivity to MMS. Phosphorylation-mimic RPA2a^S11,21/D alleviated the MMS effects on cdkb1;1 1;2. (Scale bar, 1 cm.) (C) Model for RPA2 function in stomatal terminal division regulation and DNA repair progression. T-bars and dashed lines indicate negative and positive regulation, respectively.