DNA methylation is a reversible epigenetic mark regulating genome stability and function in many eukaryotes. In Arabidopsis, active DNA demethylation depends on the function of the ROS1 subfamily of genes that encode 5-methylcytosine DNA glycosylases/lyases. ROS1-mediated DNA demethylation plays a critical role in the regulation of transgenes, transposable elements and some endogenous genes; however, there have been no reports of clear developmental phenotypes in ros1 mutant plants. Here we report that, in the ros1 mutant, the promoter region of the peptide ligand gene EPF2 is hypermethylated, which greatly reduces EPF2 expression and thereby leads to a phenotype of overproduction of stomatal lineage cells. EPF2 gene expression in ros1 is restored and the defective epidermal cell patterning is suppressed by mutations in genes in the RNA-directed DNA methylation pathway. Our results show that active DNA demethylation combats the activity of RNA-directed DNA methylation to influence the initiation of stomatal lineage cells.
A network of genes that regulates stomatal development in Arabidopsis has been identified and established as a model for addressing fundamental questions such as how specific cell lineages are initiated and established, how stem cell-like asymmetric divisions are temporally maintained and how precursor cells ultimately differentiate into functional, mature structures1,2. Earlier studies have shown that three master bHLH genes, SPEECHLESS (SPCH), MUTE and FAMA, act sequentially in stomatal fate transition and determination. The function of SPCH is especially important for the first step in the initiation of stomatal lineage cells3,4. Another important factor controlling the stomatal lineage cell population is EPIDERMAL PATTERNING FACTOR 2 (EPF2). EPF2 belongs to a family of plant-specific, cysteine-rich peptides that is secreted by the early-stage lineage cells and acts as a negative regulator in stomata formation5,6.
ROS1 is a bifunctional 5-methylcytosine DNA glycosylase/lyase critical for active DNA demethylation in most tissues of Arabidopsis plants7,8. ROS1 and its two paralogues, DEMETER-like 2 (DML2) and DEMETER-like 3 (DML3), are required for the prevention of DNA hypermethylation at thousands of genomic regions9,10,11. Loss-of-function mutations in these enzymes cause transcriptional silencing of some transgenes and endogenous genes7,12. Unlike another paralogue, DEMETER (DME), which is known to be critical for active DNA demethylation in the central cell and thereby for gene imprinting and endosperm development13, a developmental function has not been established for ROS1. In this study, we found that ROS1 loss-of-function mutants have a defect in epidermal cell patterning that is strikingly similar to the EPF2 loss-of-function phenotype. We further show that the promoter region of the EPF2 gene in ros1 mutants is hypermethylated, which leads to a dramatic decrease in its mRNA level. Our findings provide the first evidence that active DNA demethylation initiated by ROS1 plays an important role in controlling the dispersed stem cell population, the stomatal lineage cells, in plant development.
ros1 mutant epidermis has more stomatal stem cells
The loss-of-function mutants, ros1-4 and rdd (ros1-3 dml2-1 dml3-1), do not exhibit apparent developmental phenotypes at the whole-plant level11. However, microscopic observation reveals that ros1-4 has clusters of small cells in the leaf epidermis, as exhibited in the epf2-1 mutant (Fig. 1a–d and Supplementary Fig. 1). The ‘small-cell-cluster’ phenotype is more severe in the rdd triple mutant than in ros1-4 (Fig. 1c,d and Supplementary Fig. 1); however, the ros1 mutation seems to be the major contributor to the phenotype in the rdd mutant. The number of small cells is more than three times greater in ros1-4, rdd and rdd-2 (ros1-4 dml2-2 dml3-2) than in the Col wild type, although the numbers of stomata are not different in the 3-day-old seedlings (Fig. 1e). Similar to ros1-4, the ros1-3 and ros1-6 mutants also exhibit a ‘small-cell-cluster’ phenotype (Supplementary Fig. 2). Furthermore, F1 progenies from ros1-4 × ros1-3 but not ros1-4 × Col or ros1-3 × Col display the mutant epidermal patterning phenotype (Supplementary Fig. 3), and the phenotype in ros1-4 is largely rescued by expression of the wild-type ROS1 gene (Fig. 2a–c). Previous research showed that the clustered small cells in epf2 are stomatal lineage cells, which express the SPCH gene6. Because of the similarity in phenotypes, we suspected that the clustered, small cells in the ros1-4 epidermis could be stomatal lineage cells. We crossed the ros1-4 mutant with the stomatal cell fate-marker lines SPCHpro:nucGFP and MUTEpro:nucGFP. SPCH is required for the initiation of asymmetric cell division in stomatal development, and SPCHpro:nucGFP expression is mainly found in the early stomatal lineage cells3 (Fig. 3a). As expected, all of the small-cell-clusters in ros1-4 have SPCHpro:nucGFP expression (Fig. 3b,c), demonstrating an enlarged population of stomata precursors in ros1-4. In contrast, MUTE expression is required for termination of asymmetric cell division and it promotes the transition to guard mother cells (GMCs)4, and the expression of MUTEpro:nucGFP is restricted to the meristemoids in the wild-type Col (Fig. 3d). In ros1-4, only a portion of the small-cell-clusters has MUTE expression (Fig. 3e,f). The behaviours of the SPCH and MUTE markers are consistent with those found in epf2 (ref. 6). We also examined the cotyledons of 10-day-old epf2-1, ros1-4, rdd and rdd-2 plants and found that these mutants have a similar increase in the numbers of stomata compared with the wild-type Col (Supplementary Fig. 4). Together, these results suggest that some of the cells in the small-cell-clusters are arrested before the GMC stage in epf2-1, ros1-4, rdd and rdd-2 mutants. Therefore, similar to epf2 mutations ros1 mainly affects the specification of stomatal lineage cells but does not dramatically change the differentiation of these cells into GMC.
ros1 mutations cause silencing of EPF2
Given the similarity in phenotypes of the ros1 and epf2 mutants, we next analysed the EPF2 expression level in ros1-4, rdd and rdd-2 mutants at the early seedling stage using quantitative RT–PCR. The results showed that EPF2 expression is dramatically reduced in the ros1-4 and rdd mutants relative to Col (Fig. 4a). As DNA methylation in promoter regions usually decreases gene transcription, we suspected that the reduced expression of EPF2 in ros1-4, rdd and rdd-2 might result from a hypermethylation of the EPF2 promoter DNA. We searched for the methylation status of the EPF2 promoter from the DNA methylomes of ros1-4 and rdd mutants10. The region upstream of the EPF2 promoter has an 835-bp hAT-like transposable element (TE) insertion with high methylation in Col, ros1-4 and rdd. We found increased DNA methylation levels in ros1-4 and rdd mutants relative to the wild-type Col in the EPF2 promoter immediately downstream of the TE insertion (Fig. 4b). Individual locus bisulfite sequencing confirmed that the EPF2 promoter is hypermethylated at all cytosine sequence contexts (CG, CHG and CHH, where H is C, A or T) in ros1-4, rdd and rdd-2 (Fig. 4c). These results strongly suggested that DNA hypermethylation at the EPF2 promoter leads to the decreases in EPF2 expression that could be responsible for the small-cell-cluster phenotype in ros1-4 and rdd mutants.
We also performed quantitative RT-PCR analysis for epf2-1, ros1-4, rdd and rdd-2 to measure changes in gene expression associated with stomatal development. The results indicated that, in epf2, the genes involved in stomatal development, such as SPCH, MUTE, TMM, SCRM/ICE1 and ERL1, are upregulated in expression relative to Col (Fig. 4a and Supplementary Fig. 5). Consistent with the phenotypic similarity between epf2-1 and ros1-4 as revealed by microscopic observations, similar upregulation of these genes was also observed in ros1-4, rdd and rdd-2 mutants (Fig. 4a and Supplementary Fig. 5).
The epidermal phenotype of ros1 is dependent on SPCH
SPCH loss-of-function mutants lack stomata and have only jigsaw-puzzle-shaped pavement cells in their leaf epidermis3,4 (Fig. 3h), which is nearly identical to those in EPF2-overexpression lines5,6. Previous genetic and biochemical studies showed a tight connection between EPF2 and SPCH14. The perception of the EPF2 peptide by ERECTA is thought to activate a mitogen-activated protein kinase cascade that phosphorylates SPCH to downregulate its function15. Thus, EPF2 negatively regulates SPCH function at the early stage of stomatal development. If ROS1 regulates EPF2 expression and if the strong reduction in EPF2 gene expression is responsible for the small-cell-cluster phenotype in ros1-4, we would expect the loss-of-function spch-3 mutation to be epistatic to ros1-4. We crossed ros1-4 with spch-3 and generated ros1-4, spch-3 (−/−, +/−) plants. Of the progenies of selfed ros1-4, spch-3 (−/−, +/−) plants, 24.3% were of the ros1-4, spch-3 (−/−, −/−) genotype and showed the spch-3 phenotype (Fig. 3g–i and Table 1). The genetic analysis indicates that ROS1 acts upstream of SPCH in stomatal development.
EPF2 expression rescues the epidermal phenotype in rdd
We next tested whether expression of EPF2 could rescue the small-cell-cluster phenotype in the rdd mutant. We generated rdd plants carrying the dexamethasone (DEX)-inducible EPF2 construct (DEXpEPF2). The seedlings of DEXpEPF2 were germinated and grown on medium containing 2.5 or 25 μM DEX. The control plants grown on medium without DEX showed the rdd mutant phenotype (Fig. 3j). However, DEX-inducible ectopic expression of the EPF2 gene rescued the rdd mutant defect in epidermal cell patterning and small cell number (Fig. 3j–m). With a high concentration of DEX, part of the epidermis produced stomata-free or low-density area as previously observed when EPF2 was overexpressed5,6 (Fig. 3l). These results clearly show that the reduced expression of EPF2 was responsible for the small-cell-cluster phenotype in ros1 and rdd mutants.
EPF2 silencing and epidermal defect in ros1 depend on RdDM
Previous work showed that the RD29Apro:luciferase transgene silencing triggered by ros1-1 can be suppressed by mutations in components of the RNA-directed DNA methylation (RdDM) pathway such as NUCLEAR RNA POLYMELASE D1 (NRPD1) and HISTONE DEACETHYLASE6 (HDA6)16. We constructed the double mutants nrpd1-3 ros1-4 and hda6 ros1-4 and used them to test whether the small-cell-cluster phenotype and EPF2 silencing and promoter DNA hypermethylation in ros1-4 are caused by RdDM. The nrpd1-3 and hda6 single mutants did not have significant epidermal defects (Fig. 5a,c). However, these mutations strongly suppressed the small-cell-cluster phenotype in the ros1-4 epidermis, since the epidermal patterning of nrpd1-3 ros1-4 and hda6 ros1-4 is almost identical to that of the nrpd1-3 or hda6 single mutants (Fig. 5b,d,g). The NRPD1 gene encodes the largest subunit of Pol IV, which is thought to act at an early step of the RdDM pathway leading to the biogenesis of 24-nt small interfering RNAs (siRNAs) that direct the de novo DNA methyltransferase DRM2 to methylate DNA at RdDM target loci (Supplementary Fig. 6a)17,18. Using small RNA northern blot analysis, we detected 24-nt siRNAs in ros1-4 as well as in Col with a probe from the EPF2 promoter (Fig. 5j and Supplementary Fig. 7). The nrpd1 mutation blocked the accumulation of the siRNAs in Col and in the ros1-4 background (Fig. 5j). These results strongly suggest that the siRNAs are the initial trigger of EPF2 promoter DNA hypermethylation in ros1-4. Furthermore, we crossed the DNA methyltransferase triple mutant drm1-2 drm2-2 cmt3-11 (ddc) and ros1-4 and constructed the drm1-2 drm2-2 cmt3-11 ros1-4 quadruple mutant and the drm1-2 cmt3-11 ros1-4 triple mutant. The epidermal patterns of drm1-2 drm2-2 cmt3-11 triple mutant were relatively normal (Fig. 5e,g). The drm1-2 cmt3-11 ros1-4 triple mutant still exhibited the small-cell-cluster phenotype; however, the phenotype was not observed in the drm1-2 drm2-2 cmt3-11 ros1-4 quadruple mutant (Fig. 5f,g, Supplementary Fig. 6b,c). The results indicated that the drm2-2 mutation was responsible for the suppression of the small-cell-cluster phenotype in the drm1-2 drm2-2 cmt3-11 ros1-4 quadruple mutant, which is consistent with DRM2 being the de novo methyltransferase in the RdDM pathway17,18. Consistent with microscopic observations, quantitative RT-PCR and bisulfite sequencing analyses showed that EPF2 expression levels and DNA methylation status at all cytosine contexts in the EPF2 promoter region in nrpd1-3 ros1-4, ros1-4 hda6 and drm1-2 drm2-2 cmt3-11ros1-4 mutants were restored to almost the same level as in nrpd1 and hda6 single mutants or in the drm1-2 drm2-2 cmt3-11 triple mutant (Fig. 5h,i). Taken together, these results strongly suggest that siRNA-triggered de novo DNA methylation through the RdDM pathway is responsible for the DNA hypermethylation of the EPF2 promoter region and therefore for the downregulation of EPF2 gene expression in the ros1-4 mutant.
In this study, we have provided strong evidence that ROS1-initiated active DNA demethylation is necessary for the expression of the EPF2 gene. Owing to the insertion of a hAT-like TE in the upstream region of EPF2 promoter, DNA methylation tends to spread from the TE into EPF2 promoter to cause silencing of EPF2. ROS1 ensures EPF2 expression by erasing DNA methylation to control a proper population of stomatal lineage cells (Fig. 6a,b). The DNA methylation of the EPF2 promoter and silencing of EPF2 are dependent on RdDM. Our results suggest that ROS1 opposes RdDM action to ensure EPF2 expression (Fig. 6c). It was reported recently that low relative humidity triggers the methylation of SPCH and FAMA via RdDM19,20. It would be interesting to determine whether ROS1 may play a role in opposing the RdDM action on SPCH and FAMA under low humidity. It would also be interesting to investigate whether ROS1 function may be regulated by environmental and/or developmental cues. DME, another member of the ROS1 family of 5-methylcytosine DNA glycosylases/lyases13, is regulated by developmental cues since it is preferentially expressed in the central cell of the female gametophyte and in the progenitor of the endosperm21. The DME-mediated demethylation is necessary for genomic imprinting of the maternal allele of MEDEA and other genes in the endosperm21,22,23. DME is also important for the genomic demethylation in vegetative nuclei of developing pollens24,25,26. DME-mediated demethylation affects genes targeted by RdDM27,28. Initiation and maintenance of stem cell populations are important in all higher eukaryotes. Our study demonstrates that ROS1-initiated active DNA demethylation ensures the expression of EPF2, and thereby is important for the size of the stomatal stem cell population in the leaf epidermis. Future research will determine whether active DNA demethylation may also be important for other stem cells in plants.
The following mutants and transgenic lines were described previously: SPCHpro:nucGFP3, MUTEpro:nucGFP29, drm1-2drm2-2cmt3-11 (ddc)30, epf2-1 (ref. 6) (SALK_102777), spch-3 (ref. 3) (SAIL_36_B06), nrpd1-3 (SALK_128428) rdd11. rdd-2 is a triple mutant of ros1-4 (SALK_ 045303), dml2-2 (SALK_113573) and dml3-2 (SALK_056440). Arabidopsis T-DNA insertion lines were obtained from the Arabidopsis Biological Resource Center (http://www.arabidopsis.org). The hda6 mutant was kindly provided by Dr Zhizhong Gong (China Agricultural University).
Construction of plasmid DNA
One hundred nanograms of total DNA were analysed by sodium bisulfite genomic sequencing using the BisulFlash DNA Modification Kit (Epigentek; http://www.epigentek.com/catalog/index.php) according to the manufacturer’s protocol. A 1-μl aliquot of bisulfite-treated DNA was used for each PCR reaction. PCR was performed in a total volume of 20 μl using ExTaq (Takara; http://www.takarabiousa.com/). Sequenced fragments were amplified with primers specific for each region (Supplementary Table 1). Amplified PCR products were subcloned into the pCRII-TOPO vector (Invitrogen; http://www.invitrogen.com/site/us/en/home.html) following the supplier’s instructions. For each region, more than 20 independent clones were sequenced from each sample.
Expanding cotyledons were stained with FM1-43 (Invitrogen; T35356), and fluorescence was detected with an Olympus BX53 fluorescence microscope with a GFP filter. A Leica SP5 confocal microscope was used for observation of GFP and FM4-64 (Invitrogen; F34653). Imaging conditions were excitation at 488 nm and a collecting bandwidth at 500–540 nm for GFP, and excitation at 561 nm and a collecting bandwidth at 600–645 nm for FM4-64.
Small RNA northern blotting
Total RNA was extracted from 2 g of 10-day-old seedlings, and the small RNA was enriched by PEG precipitation. A 30-μg quantity of small RNA per lane was run on a 15% acrylamide gel and then transferred to Hybond-N+ membranes. Hybridization followed the manufacturer’s instructions.
DNA methylome data
How to cite this article: Yamamuro, C. et al. Overproduction of stomatal lineage cells in Arabidopsis mutants defective in active DNA demethylation. Nat. Commun. 5:4062 doi: 10.1038/ncomms5062 (2014).
Gene Expression Omnibus
The work was supported by funding from the Chinese Academy of Sciences and National Institutes of Health grants R01 GM070795 and R01 GM059138.
Supplementary Figures 1-7 and Supplementary Table 1