The Arabidopsis Exine Formation Defect (EFD) gene is required for primexine patterning and is critical for pollen fertility


  • Jun Hu,

    1. Department of Cell and Development Biology, College of Life Science, State Key Laboratory of Plant Hybrid Rice, Wuhan University, Wuhan, China
    Search for more papers by this author
    • These authors contributed equally to this work.
  • Zhaodan Wang,

    1. Department of Cell and Development Biology, College of Life Science, State Key Laboratory of Plant Hybrid Rice, Wuhan University, Wuhan, China
    Search for more papers by this author
    • These authors contributed equally to this work.
  • Liyao Zhang,

    1. Department of Cell and Development Biology, College of Life Science, State Key Laboratory of Plant Hybrid Rice, Wuhan University, Wuhan, China
    Search for more papers by this author
  • Meng-xiang Sun

    Corresponding author
    1. Department of Cell and Development Biology, College of Life Science, State Key Laboratory of Plant Hybrid Rice, Wuhan University, Wuhan, China
    • Author for correspondence:

      Meng-xiang Sun

      Tel: +86 027 68756170


    Search for more papers by this author


  • Exine, the outermost layer of a pollen grain, has important roles in protecting microspore cytoplasm and determining species-specific interactions between pollen and stigma. The molecular mechanism underlying pollen exine formation, however, remains largely unknown.
  • Here, we report the characterization of an Arabidopsis male-sterile mutant, efd, which exhibits male sterility in first-forming flowers. The Exine Formation Defect (EFD) gene is strongly expressed in microsporocytes, tetrads and the tapetum, and encodes a nuclear-localized de novo DNA methyltransferase.
  • Detailed observations revealed that EFD is involved in both callose wall and primexine formation during microsporogenesis. Microspores in tetrads are not well separated in efd due to an abnormal callose wall. Its plasma membrane undulation appears normal, but primexine patterning is impaired. Primexine matrix establishment and sporopollenin accumulation at specific positions are disturbed, and thus exine formation is totally blocked in efd.
  • We confirmed that EFD is required for pollen exine formation and male fertility via the regulation of callose wall and primexine formation. We also found that positional sporopollenin accumulation is not involved in regulating membrane undulation, but is related to the complete separation of tetrad microspores during primary exine patterning.


The angiosperm pollen wall is composed of two layers: the outer layer exine and the inner layer intine. The exine can be divided into two layers, an inner layer (nexine) and an outer sculpted layer (sexine) that includes the rod-shaped bacula and the tectum (Heslop-Harrison, 1971; Piffanelli et al., 1998). The exine is mainly composed of sporopollenin, which contains derivatives of aliphatic polymers such as fatty acids and phenolic compounds (Ahlers et al., 1999; Bubert et al., 2002; Skirycz et al., 2007). Exine has important roles in providing structural and physical support to the microspore cytoplasm, as well as protection from environmental stresses, and in the species-specific interaction of pollen with the stigma (Zinkl et al., 1999). The intine is the innermost layer of the pollen wall, located between the microspore plasma and the nexine. The intine is a simple layer that consists of cellulose, pectin and proteins. Numerous studies have revealed that the pollen wall is essential for male fertility, and defects in pollen wall development are common causes of male sterility in crops (Dong et al., 2005; Zhang et al., 2007; Dobritsa et al., 2010; Li et al., 2010; Chang et al., 2012).

Previous studies have shown that callose wall formation and primexine patterning are vital for pollen exine formation in Arabidopsis (Paxson-Sowders et al., 2001; Ariizumi et al., 2004; Dong et al., 2005; Nishikawa et al., 2005; Yang et al., 2007; Zhang et al., 2007; Guan et al., 2008; Chang et al., 2012). The callose wall is a protective wall that forms around the microspore mother cell during meiosis, acting as a molecular filter and a mold for primexine patterning (Heslop-Harrison, 1971), and preventing microspore fusion. At the tetrad stage, the primexine matrix, which acts as a scaffold for the initial sporopollenin accumulation, forms outside the undulating plasma membrane (Fitzgerald & Knox, 1995; Paxson-Sowders et al., 1997). The sporopollenin precursors deposit on the primexine, building up the probacula and protectum structures (Blackmore et al., 2007). After the callose wall degenerates and microspores are released from the tetrads, sporopollenin continues to be deposited and the basic exine structure becomes evident (Paxson-Sowders et al., 1997; Piffanelli et al., 1998). Thus, proper callose wall formation and degeneration is important in exine development. Previous studies have shown that callose deposition or dissolution is affected in some mutant tetrads, leading to abnormal exine pattern formation (Table 1). Several other Arabidopsis male-sterile mutants have also been identified as defective in exine formation, providing evidence that primexine formation at the tetrad stage is associated with the undulating surface of the plasma membrane (Table 1).

Table 1. Genes related to callose and primexine formation during Arabidopsis microsporogenesis
Callose CALS5 At2 g13680Dong et al. (2005); Nishikawa et al. (2005)
AtMYB103 At5 g56110Zhang et al. (2007)
Primexine DEX1 At3 g09090Paxson-Sowders et al. (2001)
NEF1 At5 g13390Ariizumi et al. (2004)
RPG1 At5 g40260Guan et al. (2008)
NPU1 At3 g51610Chang et al. (2012)
MS1 At5 g22260Ariizumi et al. (2005); Vizcay-Barrena & Wilson (2006); Ito et al. (2007); Yang et al. (2007)

All of these findings suggest that a complex mechanism underlies microspore exine formation, including callose wall construction and separation of tetrad microspores, plasma membrane integrity and proper undulation, and directional depositing of sporopollenin and anchoring to the membrane. The relationships among these critical developmental events are largely unknown.

In this study, we analyzed the Arabidopsis male-sterile mutant exine formation defect (efd), in which the undulating plasma membrane is well formed but primexine formation is severely impaired and the exine is almost completely absent. The Exine Formation Defect gene (EFD) is believed to be a member of the Dnmt3 family, which localizes to the nucleus. Our study provides evidence that EFD plays an important role in normal callose wall formation and primexine patterning during microsporogenesis in Arabidopsis and is therefore required for pollen wall formation and critical for pollen fertility. We also confirmed that microspore plasma membrane undulation and positional sporopollenin accumulation are not necessarily coupled with each other. The two processes may be regulated by different pathways during exine pattern formation.

Materials and Methods

Plant growth, T-DNA insertion mutant analysis and genetic crosses

Plants were grown in a glasshouse at 22°C with a 16-h light/8-h dark photoperiod. The efd mutant, Arabidopsis thaliana Salk_012057, was obtained from the Arabidopsis Biological Resource Center (ABRC) at Ohio State University, USA. Genomic DNA was extracted and subjected to PCR to identify the insertion using the LP-EFD and RP-EFD primers (5′-AGGCGAGTTCACTTAACTTATTTTG-3′ and 5′-GACTTGCAGTGAAACTTTACCTCAC-3′, respectively) that were recommended by the SIGNaL iSect Tool of the TAIR database ( in combination with the T-DNA specific LBb1 primer (5′-GCGTGGACCGCTTGCTGCAACT-3′). TAIL PCR was performed to determine the insertion site of the T-DNA fragment in the efd mutant.

Reciprocal crosses were performed between the efd mutant and wild-type (WT) Arabidopsis plants by pollinating the stigmas of immature, emasculated flowers of the desired female parent with mature pollen from the desired male parent.

Phenotype characterization and microscopy

Pollen viability was determined by FDA staining using a 5 μg ml−1 FDA solution. Aniline blue staining of pollen tubes was performed as described in Mori et al. (2006). For in vitro pollen germination, pollen grains from freshly anther-dehisced flowers were collected and then cultured in 50 μl of culture medium (described in Fun et al., 2001). To observe anther development, Arabidopsis inflorescence buds were fixed in ethanol : acetic acid (3 : 1, v/v), embedded in Spurr's resin after dehydration, and cut into 0.5-μm sections. To determine pollen morphology, pollen grains of different stages were separated from anthers, stained with DAPI, and observed by microscopy (Olympus, Tokyo, Japan). For callose observation, buds of different stages were fixed and stained with 0.1% aniline blue; the fluorescence of the callose was observed using an Olympus laser scanning confocal microscope.

For examinations using transmission electron microscopy (TEM), Arabidopsis inflorescence buds were fixed in 2.5% glutaraldehyde and then 1% OsO4, and then embedded in Spurr's resin after dehydration. Ultrathin sections were produced as described by Yang et al. (2007). For cryo-SEM observations, fresh stamens at different stages were collected and scanned using Cryo-SEM systems (Quorum PP3000T, East Grinstead, UK; Hitachi S-3400N, Tokyo, Japan). For environmental scanning electron microscopy (ESEM), fresh stamens were adhered to conductive tape and observed on a Quanta 200 microscope (FEI, Hillsboro, OR, USA).

Vector construction and transformation

In order to determine the subcellular location of EFD, onion epidermal cells were bombarded with genes encoding EFD fused with GFP at the C terminal in a pCAMBIA 1302 vector and a control construct with only GFP in the same vector. The native EFD promoter was used to drive the GUS reporter to observe the EFD spatiotemporal expression pattern. The constructs were introduced into Arabidopsis Columbia by Agrobacterium infiltration (Clough & Bent, 1998). GFP expression was detected using a laser scanning confocal microscope (Olympus). Tissues were stained for GUS expression by fixation in 90% acetone for 30 min and washing three times with staining buffer (50 mM phosphate buffer, 2 mM potassium ferricyanide, 2 mM potassium ferrocyanide and 0.2% Triton X-100). Next, the tissues were stained with 2 mM X-Gluc (5-bromo-4-chloro-3-indolyl ß-D-glucuronide cyclohexamine salt) in staining buffer and then destained in 70% ethanol until examination.

In order to complement the male sterility phenotype of efd, a 3078-bp EFD gene fragment with a 735-bp EFD native promoter was amplified by PCR with KOP Plus polymerase (Toyobo Co., Ltd., Osaka, Japan) using a pair of primers named EFD-F-S (5′-NNNNGGTACCACCAACATTCGATATGATCGTAACC-3′) and EFD-F-AS (5′-NNNCTGCAGCACGAAAACATTATATTATGACCGC-3′). The resulting fragment was inserted into pCAMBIA1302. After sequence verification, this construct was introduced into an efd plant by Agrobacterium infiltration (Clough & Bent, 1998). Transformants were obtained by double selection on MS agar plates containing 50 mg ml−1 hygromycin and 50 mg ml−1 cefotaxime sodium.

RNA isolation and RT-PCR analysis

For expression analysis of the efd mutant, total RNA was isolated from inflorescences of Columbia WT and efd using TRIzol reagent (Invitrogen). RT-PCR was used to examine the EFD expression level using the primer pair EFD-S (5′-CTGCCTTATCAGAGAAAGAGGCTGAA-3′) and EFD-AS (5′-TAAGCGAAAGTTTATGCGGAATGTCT-3′).

In order to determine the EFD expression pattern, RNA was isolated from roots, shoots, rosette leaves, inflorescences, and siliques at 1 d after pollination (siliques 1 DAP) and 2 DAP (siliques 2DAP). RT-PCR was performed using the primers EFD-S and EFD-AS.

Structure and phylogenetic analysis of EFD

InterPro Scan analysis was used to analyze conserved domains in EFD. Swiss-PDB Viewer was used to predict the tertiary structure of EFD and compare it with other known DNA methyltransferases. For the phylogenetic analysis, data for 17 DNA methyltransferases were obtained from a BLAST search from the NCBI database. The PHYLogeny Inference Package (PHYLIP, Seattle, WA, USA) was used for phylogenetic analysis.

de novo DNA methylation activity test of EFD in vitro

The full-length EFD cDNA was inserted into pMXB10, and the resulting construct was introduced into the E. coli strain BL21. The BL21 strain containing the pMXB10 EFD plasmid was grown in the presence of 400 μM IPTG for 16 h at 18°C, after which the EFD protein was purified with chitin resin (New England Biolabs, Ipswich, MA, USA).

The DNA methylation activity of the purified EFD was tested as described by Pang et al. (2013). A 612-bp substrate fragment with an AclI recognition site at the 225-bp position was amplified from the Arabidopsis (Columbia) genomic DNA using a primer pair with the sequences 5′-AGGTTTCACTGATTATGAATTTTG-3′ and 5′-GGTTCATGTGTTGTAGGATCA-3′. One microgram of the substrate fragment was incubated with either purified EFD (final concentration, 5 μmol l−1) or M.SssI (New England Biolabs) at 37°C for 3 h. Then, the substrate fragments were purified and digested with AclI for 1 h at 37°C. The digested fragments were separated on a 2% agarose gel. As AclI is sensitive to DNA CpG methylation, a nonmethylated substrate would be digested into 225- and 387-bp fragments, whereas a methylated substrate would not be digested.

Microarray analysis

Whole inflorescences were collected when plants displayed three fertile (WT) or sterile siliques (efd). Total RNA was extracted from c. 100 mg of each sample. The RNA samples were treated with DNase and then purified. Double-stranded cRNA was synthesized using the SuperScript Choice system (Invitrogen) and labeled. The Arabidopsis ATH1 Genome Array (Affymetrix) was screened with the cRNA samples according to the Affymetrix GeneChip Technical Analysis Manual. Two replicates of independently grown material were used. Statistical analysis of microarrays (SAM) was used to analyze the data. A false discovery cutoff of 5% was used for the initial selection of candidate genes, and a secondary filtering of a > log2 fold difference was then applied. The NCBI and TAIR databases were used to annotate genes showing expression changes in efd buds compared with the WT.


Phenotypic identification of the efd mutant

Using suppression subtractive hybridization and mirror orientation selection, we previously analyzed differential gene expression in the egg cell and zygote of Nicotiana tabacum cv SR1 (Ning et al., 2006). Several genes that are expressed in the zygote, but not in the egg cell, were identified. Because the Arabidopsis EFD shares strong identity with one of these genes, a T-DNA-tagged mutant line of the EFD gene was then identified and analyzed. The EFD gene contains three exons and two introns. Thermal asymmetric interlaced (TAIL) polymerase chain reaction (PCR) results showed that the mutant has a T-DNA insertion in the first intron, 266 bp downstream of the ATG initiation codon (Fig. 1a). Reverse transcription (RT)-PCR confirmed that EFD expression is blocked in the efd mutant (Fig. 1b).

Figure 1.

Identification of efd and genetic analysis by reciprocal crosses between Arabidopsis wild-type (WT) and efd plants. (a) Gene structure of EFD and T-DNA insertion. Black boxes, 5′ untranslated region (UTR) and 3′ UTR; gray box, exons encoding amino acids; black lines, introns; triangle, T-DNA insertion site. (b) Expression of EFD in WT, efd and complemented plant. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) gene served as internal reference. (c–e) Comparison of reproductive development of WT (c), efd (d) and complemented (e) plants. (f, g) WT as paternal parent and efd as maternal parent. (f) The appearance of hybrid siliques, (g) shows the ovules inside the siliques. (h–m) efd as paternal parent and WT as maternal parent. (h, j, l) The appearance of hybrid siliques and (i, k, m) the ovules inside the siliques. Bars (h–m), 1 mm.

The efd mutant plants were indistinguishable from WT plants during the vegetative growth phase; however, the efd plant produced very few seeds upon self-pollination and fertility was dramatically reduced as indicated by very short siliques with shrunken ovules (Fig. 1c,d). A complementation assay of EFD expression in the efd mutant was also undertaken by introducing a full-length EFD gene with its native promoter into the efd mutant. As expected, EFD was expressed and fertility was restored in the complementation plants (Fig. 1e, efd-C). Therefore, we confirmed that the loss of EFD function leads to infertility in Arabidopsis.

We compared the seed yield between WT and efd plants by counting the numbers of seeds in the siliques. We defined siliques with > 30 seeds as having normal fertility. The siliques were classified into groups of six according to the order in which they formed. The seed yields and lengths of siliques were then measured (Table 2). In total, 71.87% of the efd siliques were abnormal and only 10.32% of the efd siliques had normal fertility (= 2151), whereas 89.43% of the WT siliques showed normal fertility (= 965). Interestingly, no seeds or only a few seeds were present in the three-first-forming groups of siliques (1st to 18th siliques). In subsequent groups, both the seed yield and silique length gradually increased. The number of seeds as well as silique length were dramatically higher in the fifth group (31st to 36th siliques), which had the appearance of WT siliques and produced an average of 29.75 seeds (= 20); the average seed number in the sixth group (37th to 42nd) was 31.5 ± 11.49 (Table 2). This suggests that the fertility of the efd mutant recovers gradually as the inflorescence develops. Similar fertility recovery processes have been observed in the male-sterile mutants ms2, atgpat1 and rpg1 (Aart et al., 1997; Zheng et al., 2003; Guan et al., 2008). The gal-3 rgl2-1 rga-t2 triple mutant and the gal-3 rgl2-1 gai-t6 rgat2 quadruple mutant also display this phenotype (Yu et al., 2004). However, the mechanism of recovery has not been determined for any of these lines.

Table 2. Fertility and length analysis of Arabidopsis wild-type (WT) plant and efd mutant siliques
Silique groupAverage number of seedsAverage length of siliques (mm)
WT efd WT efd
  1. The siliques were grouped in sixs according to the order in which they formed. Significant difference: *, < 0.05; **, highly significant difference < 0.01.

1–633.29 ± 19.180.00 ± 0.00**10.89 ± 3.963.00 ± 0.00**
7–1250.65 ± 6.070.19 ± 2.10**14.00 ± 0.943.07 ± 0.78**
13–1850.33 ± 5.302.04 ± 7.56**13.06 ± 0.973.81 ± 2.32**
19–2447.71 ± 4.247.50 ± 15.18**11.61 ± 1.975.40 ± 3.81**
25–3046.31 ± 3.5915.84 ± 16.93**11.35 ± 1.597.92 ± 4.05**
31–3644.00 ± 7.8729.75 ± 13.17*12.01 ± 0.8910.96 ± 2.29*
37–4245.00 ± 2.8331.50 ± 11.49*11.00 ± 0.8211.63 ± 1.69

A reciprocal cross between the WT and efd showed that efd plants produced normal-sized siliques when WT was used as the pollen donor (Fig. 1f,g), but the siliques were still infertile when efd pollen grains pollinated the WT stigma. Of all the siliques produced by pollinating WT stigma with efd pollen grains, 20% contained an average of 11.5 normal and 37.2 abnormal young seeds (Fig. 1h,i); 46.67% were elongated but had few or no normal young seeds (Fig. 1j,k), and 33.33% were totally infertile with no post-fertilization development (Fig. 1l,m; Supporting Information Table S1). The female reproductive system was not notably affected in the efd mutant, suggesting that the infertility is caused mainly by defects in the male reproductive system.

Microsporogenesis in the efd mutant is aberrant from the tetrad stage

The dehiscence of efd anthers was found to be defective. At the time that the WT anthers were totally dehisced (Fig. 2a), 32.78% of the efd anthers had not dehisced (Fig. 2b); 48.89% had dehisced but almost no pollen could be observed (Fig. 2c), and 18.33% had dehisced with only a few pollen grains (Fig. 2c; Table S2). Fluorescein diacetate (FDA) staining indicated that 75.4 ± 0.93% of the efd pollen grains were viable (Fig. 2e,f), whereas 98.30 ± 1.60% of the WT pollen grains were viable (Fig. 2g,h); no pollen tubes were observed in self-pollinated efd pistils at 24 h after fertilization (HAF; Fig. 2i).

Figure 2.

Characterization of Arabidopsis efd mutant. (a) Wild-type (WT) anther; (b–d) efd anthers. White arrow in (c) indicates an anther locule that is not well dehisced. (e, f) efd pollen grains; (g, h) WT pollen grains. Pollen viability was revealed by FDA staining. (e, g) Green fluorescence of pollen grains; (f, h) bright field images. The arrows in (f) indicate dead efd pollen grains. (i, j) Pollen tubes that had developed by 24 h after pollination (HAP). (i) No pollen tube can be observed at 24 HAP in efd. (j) Pollen tubes of WT. (k, l) in vitro germination of efd (k) and WT (l) pollen grains collected from 50 flowers and germinated under the same conditions. Bars, 100 μm.

The defect in pollen germination was also confirmed by in vitro pollen germination experiments. Pollen grains from 50 WT and efd flowers each were collected and germinated in vitro for 24 h. Over 80% of the WT pollen grains germinated and produced long pollen tubes (Fig. 2l), but very few efd pollen grains germinated under the same culture conditions (Fig. 2k). These results indicate that although the efd mutant could produce some live pollen grains, its ability to germinate was severely impaired.

By sectioning anthers at different developmental stages, we observed structural differences between WT and the efd mutant. As reported previously, Arabidopsis anther development can be divided into 14 stages according to morphological characteristics (Sanders et al., 1999). At stage 7, both WT and efd anthers were able to generate tetrads surrounded by a callose wall, although the efd callose wall appeared to be thinner (Fig. S1a,b), and the layers of the anther wall were well differentiated. At stage 8, when the callose wall degenerated and single microspores were released in the WT, the efd microspore cytoplasm appeared to be more condensed (Fig. S1c,d), but did not seem to be aborted. The tapetum was fully developed. From stages 9–11, efd pollen already appeared to be empty or were irregular in shape, and some microspores were arrested at the uninucleate stage (Fig. S1f,h,i). Differences in the tapetum could be observed, and the efd tapetum degeneration was notably delayed. At stage 12, tricellular pollens formed in the WT anther and the tapetum was obviously degenerated (Fig. S1k). However, in efd anthers, most of the pollen grains were aborted and only a few were comparable to the WT pollen in size (Fig. S1l); the tapetum remaining around the pollen grain could still be observed. These observations indicate that the cell layers of the anther wall in efd lines were well developed, but the tapetum degeneration had been blocked.

In order to confirm whether the aberrance had already appeared as early as the tetrad stage, we observed the processes of microsporogenesis and male gametophyte development in more detail. At the late tetrad stage in WT, the microsporocytes were well formed (Fig. 3a,b). The tetrads were surrounded by a thick callose wall, and four cells in the tetrad were well wrapped and isolated by the callose wall (Fig. 3c,d). When uninuclear microspores were released from the tetrads, they soon became rounded and their surfaces appeared smooth (Fig. 3e,f). After the first mitosis, WT microspores developed into a bicellular stage (Fig. 3g,h). When flowers opened, the mature tricellular pollen grains were well developed (Fig. 3i,j). In efd plants, the microsporocytes also formed (Fig. 3k,l), but the tetrads had a different appearance. Although they were also surrounded by the callose layer, the microspores in the tetrad were not well separated by the callose wall between them (Fig. 4m,n). After being released from the tetrads, the efd microspore surface had a coarse appearance. The cell wall development was notably disturbed. The microspores were arrested and even collapsed at this stage (Fig. 3o,p). It was also observed that some microspores developed into bicellular (Fig. 3q,r) and tricellular pollen (Fig. 3s,t).

Figure 3.

Morphologies of microsporocytes,tetrads, microspores and pollen grains during Arabidopsis microsporogenesis in wild-type (WT) and efd plants. (a–j) WT material; (k–t) efd mutant material. (a, b) Microsporocytes; (c, d) tetrad (notice that the microspores are well isolated); (e, f) microspores; (g, h) bicellular pollen; (i, j) mature pollen grains; (k, l) microsporocytes; (m, n) tetrad (the microspores are not well isolated); (o, p) microspores; (q, r) bicellular pollen; (s, t) mature pollen grains. The black arrows in (q) and (s) indicate arrested efd microspores. Bars, 10 μm.

Figure 4.

Callose walls of microsporocyte, tetrads and microspores in Arabidopsis wild-type (WT) and efd plants. (a–d) Microsporocytes of WT (a, b) and efd (c, d); (e–h) tetrads of WT (e, f) and efd (g, h); (i–l) microspores of WT (i, j) and efd (k, l). Arrows in (a) and (c) indicate callose generated at the microsporocyte stage. Bars, 10 μm.

In order to characterize the defect in callose wall formation, the microsporocytes, tetrads and microspores released from tetrads were stained with aniline blue, and the fluorescence was observed under the same conditions. The callose produced at the microsporocyte stage showed similar fluorescence intensities in efd and WT anthers (Fig. 4a–d). Later, the callose wall was well established and showed clear fluorescence in WT tetrads (Fig. 4e,f), whereas in efd the callose wall was also observed in tetrads, but the fluorescence of the callose was much weaker than in WT tetrads (Fig. 4g,h), indicating lower callose deposition in efd tetrads. After being released from tetrads, the WT microspores showed strong cell wall staining (Fig. 4i,j), but efd microspores presented a weak signal (Fig. 4k,l). These results indicate that efd microsporogenesis becomes aberrant from the tetrad stage with abnormal callose accumulation, resulting in defective cell wall formation.

Primexine development is affected in efd

Cryo-SEM (Fig. 5) and ESEM (Fig. S2) were used to observe the exine pattern of WT and efd pollen grains. In the WT, mature and well-developed elliptical pollen grains were observed in the freshly dehisced anther (Figs 5a, S2A), and they were also well separated from each other, whereas efd pollen grains were irregular in shape and agglomerated together (Fig. S2B). Some membrane-like structures, which appeared to be remnants of the tapetum, adhered to pollen grains (Fig. S2B). Normally, the mature WT pollen grains have a characteristic ridged surface (Figs 5a, S2C), but the efd pollen grain surface was either smooth (Fig. 5b) or had irregular protrusions (Figs 5c,d, S2D). Obviously, exine development in the mutant was blocked, which may result in the abortion of pollen development.

Figure 5.

Cryo-SEM observations of mature pollen grains of Arabidopsis wild type (WT) and efd plants. (a) Elliptical WT mature pollens with sculptured exine; (b–d) mature efd pollen grains that are irregular in shape. (b) efd pollen grains sticking to each other and to the wall materials from the tapetum. (c) efd pollen with coniform protrusions on the surface; the pollen aborted in later stages. (d) efd pollen with a coarse surface; the pollen also aborted in early stages. Bars, 5 μm.

The detailed process of exine development in the efd mutant was examined by comparing the ultrastructure of tetrads, microspores, bicellular pollen grains, and mature pollen grains of efd and WT plants using TEM. In WT plants, after the completion of meiosis, the primary cell wall material was deposited between the plasma membrane and the callose wall of the microspores, and the primexine then formed between the plasma membrane and the callose wall. Typically, at this stage, the probaculum appeared and was positioned in a regular pattern (Fig. 6a). In the efd mutant, a very thin primexine-like layer that never fully developed was observed between the callose layer and plasma membrane. Notably, no probaculum-like structure could be observed in this primexine-like layer (Fig. 6b), indicating that the primary pattern formation of exine construction was disturbed. After microspores were released from the tetrads, the WT microspores had already formed intine and exine layers, and the exine layer contained both nexine and sexine (Fig. 6c), but the surfaces of the mutant microspores were smooth and the intine layer was thinner than that of the WT. The exine layer contained only nexine, and no baculae or tecta could be observed (Fig. 6d), suggesting that the primary pattern of primexine was necessary for further development of an intact pollen wall. As the first mitosis concluded, the microspores developed into bicellular pollen grains, the intine layer of the WT became thicker, and the exine increased in size with additional polymerization of sporopollenin (Fig. 6e). In the efd mutant, the intine layer also became thicker, but no bacula or tectum was observed (Fig. 6f). As the WT pollen matured, the tapetal cells degenerated and some tapetal material became embedded in the exine or covered the surface to form a pollen coat (Fig. 6g). Debris from tapetal cells could also be found on the surface of mutant pollen, either attached to the surface or remaining in the anther locules (Fig. 6h).

Figure 6.

Ultrastructure of pollen wall development in Arabidopsis wild-type (WT) and efd plants. (a, b) WT (a) and efd (b) tetrads. Only a thin primexine-like layer can be observed between the callose wall and the plasma membrane in the efd tetrad. (c, d) WT (c) and efd (d) microspores. The WT microspore formed a sculpted exine with nexine and sexine, whereas the efd microspore lacked sexine structure and formed no bacula or tectum. (e, f) WT (e) and efd (f) bicellular pollen. The exine structure was increased in size in the WT, whereas the surface of the efd bicellular pollen was still smooth without a sexine structure. (g, h) WT (g) and efd (h) mature pollen grains. The tapetum remnants deposit as tryphine and fill in the cavities in the WT, but tapetum remnants only attached to the surface of the efd pollen grains. (i, j) WT (i) and efd (j) anther with bicellular pollen grains. The tapetum was degenerated in the WT anther, whereas thick tapetum with no sign of degeneration was observed in the efd anther. (k, l) WT (k) and efd (l) anthers with mature pollen grains. The tapetum was completely degenerated in the WT, whereas remnants of the tapetum were still observed in efd. Ba, bacula; Bi, bicellular pollen grain; Cal, callose; I, intine; Msp, microspore; Ne, nexine; P, mature pollen grain; Pc, pollen coat; Pe, primexine; PL, primexine-like layer; RM, remnant of tapetum; T, tapetum; Te, tectum. Bars: (a–h) 1 μm; (i–l) 5 μm.

In summary, cryo-SEM, ESEM and TEM experiments indicated that exine development in the efd mutant is defective and occurs as early as the tetrad stage. Primexine development and sporopollenin deposition were blocked and eventually disrupted exine formation.

EFD encodes a de novo DNA methyltransferase that localizes to the nucleus

EFD encodes a 323-amino-acid protein with a molecular mass of 41 kDa, containing five alpha helices and seven β-strands. S-adenosyl-l-methionine (SAM)-dependent methyltransferase enzymes share little sequence identity, but their structures are conserved (Martin & McMillan, 2002). BLAST searches and InterPro Scan analysis indicated that the EFD protein contains a typical SAM-dependent methyltransferase domain. We also compared the tertiary structure of EFD with that of another known methyltransferase (3EGE), obtained from the Protein Data Bank database. The tertiary structure of this known methyltransferase was determined by X-ray stress analysis (Fig. S3A). The tertiary structure of EFD was predicted using the Swiss-PDB Viewer software (Fig. S3B), and merged with that of 3EGE (Fig. S3C). The merging revealed that EFD has a conserved SAM-dependent methyltransferase conformation, suggesting that EFD encodes a SAM-dependent methyltransferase protein.

Martin & McMillan (2002) showed that the DNA SAM-Mts have a signature motif, E/DXXXGXG, in the SAM-binding N-terminal region. A DVGTGNG motif was found in the N-terminal region of the EFD protein, suggesting that EFD is a DNA methyltransferase. We constructed a translational fusion of EFD with green fluorescent protein (GFP) driven by the EFD native promoter and bombarded it into onion epidermal cells to determine if EFD localizes to the nucleus. GFP alone, with expression driven by the cauliflower mosaic virus 35S promoter, was found to be distributed in both the cytoplasm and nucleus, but the EFD–GFP fusion protein was only detected in the nucleus, consistent with the nuclear localization of DNA methyltransferases (Fig. S3).

DNA methyltransferases in animals are usually divided into three families: Dnmt1, Dnmt2 and Dnmt3. Many DNA methyltransferases have also been found in plants and are also divided into three families according to their substrates: maintenance methyltransferases that belong to the Dnmt1 family, de novo methyltransferases that belong to the Dnmt3 family, and chromomethylases that methylate heteromatin and exist only in plants. We obtained protein sequences of 17 DNA methyltransferases by a BLAST search of the NCBI database, including those from Arabidopsis, maize, carrots, zebrafish and mice. We then performed a phylogenetic analysis to confirm the DNA methyltransferase family to which EFD belongs. The phylogenetic analysis indicated that EFD is related most closely to DanioDnmt3, MusDnmt3A and MusDnmt3B, which are members of the Dnmt3 family and have been shown to perform de novo methylation in zebrafish (DanioDnmt3) and mice (MusDnmt3A and MusDnmt3B) (Fig. S4). This suggests that EFD is likely to be an Arabidopsis de novo DNA methyltransferase.

The EFD was demonstrated to have de novo DNA methylation activity in vitro. A 612-bp substrate fragment amplified by PCR with no CpG methylation was incubated with EFD and then digested by AclI, which is sensitive to CpG methylation. A fraction of the substrate was not digested (Fig. 7), indicating that it was methylated by EFD, thus confirming that EFD is a de novo DNA methyltransferase.

Figure 7.

de novo DNA methyltransferase activity of Arabidopsis EFD. (a) A 612-bp substrate fragment with an AclI site. (b) The substrates were treated with either EFD (lane 2) or CpG Methyltransferase M.SssI (lane 3) and then digested with AclI; untreated substrate digested with AclI (lane 4) and substrate without any treatment (lane 5) are also shown. Arrows indicate methylated substrate fragments treated with either EFD or M. SssI.

EFD is strongly expressed in microsporocytes, tetrads and the tapetum

The spatial and temporal patterns of EFD expression during anther development were determined by different methods. RT-PCR amplification of EFD in different Arabidopsis tissues revealed that EFD is expressed in vegetative tissues such as the root, shoot and rosette leaves. In reproductive tissues, EFD is expressed in inflorescences and in siliques at 2 d after pollination (DAP), but is not or is only slightly expressed in siliques at 3 DAP (Fig. 8a). These results suggest that EFD plays a role before fertilization and in the early embryogenesis stage after fertilization. GUS reporter gene expression driven by the EFD promoter indicated that EFD is expressed in the root, shoot, and leaves of seedlings that had germinated for 7 d on a Murashige & Skoog (MS) plate (Fig. 8b,c). At the reproductive developmental stage, the GUS signal was strongly expressed in the early inflorescence buds, in both stamens and in pistils. The GUS signal was detected only in pistils and not stamens at the bicellular and mature pollen stage (Fig. 8d–f). Sections of stained buds revealed that the GUS signal was expressed in anther layers and microsporocytes at stage 6 of anther development (Fig. 8g). It was later expressed in anther layer cells and tetrads (Fig. 8h). When the microspores were released from the tetrads, the GUS signal was strongly expressed in the tapetum and in early stage microspores (Fig. 8i), but only slightly expressed in late-stage microspores (Fig. 8j). As mitosis began and the tapetum degenerated, a low GUS signal was observed in bicellular (Fig. 8k) and mature pollen grains (Fig. 8l). The expression of GFP-EFD driven by the EFD promoter showed a similar pattern. The GFP signal was ubiquitously expressed in early anthers, including wall layers of the microsporangium and the microsporocyte (Fig. 9a–c) or tetrads (Fig. 9d–f). Later, the signal was detected mainly in the tapetum, but not in the released microspores (Fig. 9g–i), and finally, no GFP signal was observed in the entire anther at the bicellular pollen stage (Fig. 9j–l). The strong expression of EFD in microsporocytes, tetrads and the tapetum is consistent with the role of EFD in exine development.

Figure 8.

Expression pattern of Arabidopsis EFD. (a) EFD expression revealed by RT-PCR. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) gene was served as internal reference. (b) EFD expression revealed by GUS staining in the seedling at 7 d after germination. (c) GUS staining in roots. (d–f) GUS staining in inflorescences. (d) The distribution of the GUS signal in whole inflorescences; 1–3 are buds with mature tricellular pollen, 4–6 are buds with bicellular pollen, and the rest are buds before mitosis. (e) The fifth bud in the inflorescence with a strong GUS signal in the pistil, but no signal in the stamen. (f) Magnification of the young buds displayed in (d), showing a strong GUS signal in both pistil and stamen. (g) Anther section at the microsporocyte stage; strong GUS staining is present in the anther wall layers and the microsporocyte. (h) Anther section at the tetrad stage; GUS staining appears in the tapetum and tetrads. (i) Anther section at the early microspore stage; GUS staining can be observed in tapetum and microspores. (j) Anther section at the late microspore stage, GUS staining appears only in the tapetum. (k) Anther section at the bicellular pollen stage. (l) Anther section at the mature pollen stage. BP, bicellular pollen; M, microsporocyte; MSP, microspore; T, tapetum; TP, tricellular pollen; Trd, tetrad. Bars: (g–l) 20 μm.

Figure 9.

Expression pattern of EFD-GFP fusion protein in transgenic Arabidopsis (Col) plants. (a–d) Anthers at the microsporocyte stage; GFP is expressed in both anther wall layers and in the microsporocyte. (e–h) Anthers at the tetrad stage; GFP is expressed in the anther wall layers and tetrads. (i–l) Anthers at the microspore stage; GFP is mainly expressed in the tapetum, but not in the microspores. (m–p) Anthers at the bicellular pollen stage; no GFP expression is observed in the anther wall layers or the bicellular pollen. (a, e, i, m) GFP signals in anthers; (b, f, j, n) chlorophyll autofluorescence that outlines the anther structures; (c, g, k, o) merged images of GFP and autofluorescence; and (d, h, l, p) magnifications of GFP signals in anthers. Bars: (d, h, l, p) 30 μm; (other panels) 100 μm.

EFD affects the expression of genes related to pollen exine formation and anther development

A comparison of the transcript profile for the whole inflorescence was performed between efd and WT using an Arabidopsis ATH1 Genome Array (Affymetrix, Santa Clara, CA, USA). The analysis revealed that 307 genes were downregulated and 24 genes were upregulated in efd (Table S6). We searched the NCBI database ( and The Arabidopsis Information Resource (TAIR) ( to annotate all of these genes and used the Arabidopsis eFP browser ( to determine their tissue-specific expression patterns. The results showed that 303 of the 307 downregulated genes were expressed specifically in the anther or microspores, and 14 of the 24 upregulated genes were expressed in the anther or microspores, which indicates that a majority of the abnormally expressed genes are involved in stamen development. Of the 331 abnormally expressed genes, 277 were present in pollen transcriptomes, including 271 downregulated and six upregulated genes. In particular, 147 of the 271 downregulated genes were male gametophyte-specific (Table S7) according to previous reports on the pollen transcriptome (Honys & Twell, 2004), which indicates that the altered transcriptome of the efd mutant is concordant with the primexine and exine development defect. We also consulted the Munich Information Center for Protein Sequences (MIPS) to functionally categorize the downregulated or upregulated genes in the efd mutant, with the results indicating that these genes are involved in various biological processes, including metabolism, development, and biogenesis of cellular components, and some also showing binding functions or cofactor requirements (Table 3).

Table 3. Functional categorization of the down- or upregulated genes in Arabidopsis efd buds using the Munich Information Center for Protein Sequences (MIPS) database
Functional categoryDownregulated genes (= 305)Upregulated genes (= 24)
  1. na, not available.

Storage protein1.96%na
Cell cycle and DNA processing2.95%4.16%
Protein synthesis1.31%na
Protein fate (folding, modification, destination)5.57%8.33%
Protein with binding function or cofactor requirement25.90%16.60%
Regulation of metabolism and protein function2.29%na
Cellular transport, transport, transport facilities and transport routes16%13%
Cellular communication/signal transduction mechanism3.27%na
Cell rescue, defense and virulence7.21%16.60%
Interaction with the environment8.52%16.60%
Systemic interaction with the environment3.27%12.50%
Cell fate2.29%na
Development (Systemic)11.40%na
Biogenesis of cellular components11.40%4.16%
Cell type differentiation1.31%na
Subcellular localization40.60%41.60%
Unclassified proteins19.00%20.80%

Excluding proteins of unknown function, the five genes show the greatest downregulation and the five genes show the greatest upregulation in efd buds (Table S3). Among the 10 genes AT4G08670, At2G07560, AT3G01270, AT2G05540, AT2G18550 and AT1G12010 were downregulated in all of ms1, ems1 and spl mutant buds; AT1G12010, AT2G18550, AT2G05540 and AT1G61800 were upregulated in ms1, ems1 and spl mutant buds; AT1G61800 was downregulated in ems1 and upregulated in spl but not changed in ms1; AT4G35010, AT5G4380 and AT5G12010 were not changed in ms1, ems1 and spl mutant buds (Tables S8, S9). MS1 functions in the nucleus of the tapetal cell to regulate primexine formation and pollen wall material secretions from the late tetrad to the newly released microspore stage (Yang et al., 2007). EMS1 has a role at an earlier stage than MS1 and controls the fate of somatic and reproductive cells in the Arabidopsis anther (Zhao et al., 2002; Yang et al., 2007). SPL encodes a transcriptional regulator of sporocyte development in Arabidopsis (Yang et al., 1999). The overlapping of these up- or downregulated genes in ms1, ems1, spl and edf (Tables S8–S10) suggests that they may function through the same downstream pathway. However, as the expression of MS1, EMS1 and SPL did not change in efd and EFD expression did not vary in ms1, ems1,or spl, EFD may regulate pollen wall development in a different way from MS1, EMS1 and SPL.

Some genes known to be involved in Arabidopsis primexine and exine formation were found to be downregulated in the efd mutant, including CALS5, RPG1 and CYP703A2/DEX2 (Table S4). CYP703A2/DEX2 is a key gene in sporopollenin synthesis during microsporogenesis in Arabidopsis (Morant et al., 2007). The expression of CYP703A2/DEX2 was almost undetectable in efd (−11.4 log2 fold change), indicating that sporopollenin synthesis is severely disturbed in efd. Previous studies of RPG1 have confirmed that it has a role in primexine formation and microspore membrane integrity in Arabidopsis (Guan et al., 2008). The lower expression of RPG1 in efd (−5.84 log2 fold change) suggests that RPG1 downregulation may lead to primexine formation defects in efd. CALS5 is vital for callose wall formation during microsporogenesis in Arabidopsis (Dong et al., 2005; Nishikawa et al., 2005), and the downregulation of CALS5 (−3.51 log2 fold change) indicates that callose synthesis and callose wall formation was affected in the efd anther. This is consistent with the phenotypes that were observed in the efd anther.

Sporopollenin, primexine and callose are crucial for exine formation. The downregulation of these genes in the efd mutant indicates that callose and sporopollenin synthesis are affected, disturbing primexine formation and thus leading to the total lack of exine formation in the efd mutant. Sporopollenin precursor secretion and translocation requires ABC transporter proteins, lipid transfer proteins and glycine-rich proteins (GRP; Piffanelli et al., 1998; Yang et al., 2007; McNeil & Smith, 2010). Some genes that encode ABC transporters, lipid transfer proteins and GRPs were also found in the altered transcriptome (Table S5), indicating that the secretion and translocation of sporopollenin precursors may also be affected in efd.


EFD is required for primexine formation in the Arabidopsis microspore

Primexine formation during the tetrad stage is essential for proper exine formation. Previous studies of several mutants defective in primexine formation (ms1, dex1, nef1, rpg1 and npu) have revealed that correct primexine pattern formation is a key factor in exine development. However, details of the process of primexine pattern formation and the molecular mechanisms underlying the process remain largely unknown. In the efd mutant, primexine formation is clearly affected at the tetrad stage. Only a thin primexine-like layer with no probacula-like structure forms between the callose wall and plasma membrane. As the primexine acts as a scaffold, matrix or template for the initial sporopollenin accumulation (Scott et al., 2004; Wilson & Zhang, 2009), the defective primexine formation causes failed exine formation in the efd mutant, which produces pollen almost without exine. EFD is clearly required for primexine formation, and mainly functions in the early stage of primexine initiation.

During the early stages of primexine initiation, the undulating plasma membrane structure in the tetrads might be important for primexine formation, possibly in the process of building up primexine frameworks and producing the spaces for new portions of primexine substances from the microspores (Gabarayeva et al., 2001, 2010; Chang et al., 2012). Among mutants showing defects in primexine initiation-related genes, rpg1 and nef1 show damaged membrane integrity and ms1 and dex1 have an unclear undulating plasma membrane structure, and npu completely lacks plasma membrane undulation; primexine development is affected in all of these mutants. In the efd mutant, an undulating plasma membrane is clearly present but primexine initiation is blocked to some extent. Microarray data indicated that RPG1, which is thought to play a role in maintaining microspore membrane integrity and the timely undulation of microspore membranes (Guan et al., 2008), is notably downregulated in efd (−5.8 log2 fold change), suggesting that EFD is involved in the regulation of plasma membrane integrity and therefore has a role in primexine development.

However, although undulating plasma membrane structures were clearly observed in the efd mutant, primexine formation was found to be severely impaired and no sexine structures form. These data suggest that plasma membrane integrity and undulation are differentially regulated. The membrane undulation may be involved in formation of the specific pattern of the primexine framework, whereas the membrane integrity ensures the primexine initiation and deposition. EFD is likely downstream of NPU, based on the sequence of the phenotypes in both mutants. In the efd mutant, either the secretion of primexine constituents or the directional deposition of primexine material was severely affected; the basic template for exine formation was thus absent, leading to an exine development defect.

Positional sporopollenin deposition may require complete separation of tetrad microspores but does not regulate plasma membrane undulation

Proper callose synthesis and degeneration are essential for exine formation. In the cals5-1 and cals5-2 mutants, callose deposition is almost completely absent, with the four microspores in the tetrad forming a clump and the boundaries being barely recognizable. The baculae and tectum structures are abnormal and the microspores degenerate, resulting in a severe reduction in fertility (Dong et al., 2005). This suggests that the callose layer between microspores in a tetrad is involved in directional sporopollenin accumulation during primary exine patterning. In efd mutant tetrads, similar to cals5 tetrads, the boundaries between the microspores in the tetrad were also ambiguous, but the baculae and tectum structures were entirely absent. These data suggest that the complete separation of tetrad microspores is likely necessary for exine patterning. Accordingly, the CALS5 gene is expressed at a much lower level than in WT (−3.5 log2 fold change), as revealed by the microarray analysis. This absence of callose indicates that EFD is expressed upstream of CALS5 and influences its expression during callose wall formation.

Previous reports also indicated that plasma membrane undulation is a common phenomenon, and it has been proposed to be a critical step in primexine formation (Fitzgerald & Knox, 1995; Paxson-Sowders et al., 1997). There is little direct evidence for a relationship between membrane undulation and primexine formation. DEX1 encodes a membrane-associated protein. The normal rippling of the plasma membrane is absent in dex1 at the tetrad stage, and sporopollenin is deposited randomly (Paxson-Sowders et al., 2001). The loss of NPU, a membrane-localized protein, prevents plasma membrane undulation and sporopollenin deposition, resulting in the complete absence of primexine (Chang et al., 2012). Both mutants exhibit abnormal plasma undulation and the absence of sporopollenin deposition at specific sites. It is not clear if proper plasma undulation and sporopollenin deposition are interrelated, or if undulation is a prerequisite for positional sporopollenin deposition or vice versa. In the present work, we found that in efd microspore plasma undulation is normal, but positional sporopollenin deposition is absent and primexine formation is strongly disturbed. This suggests that although the two events are associated, they are regulated by different pathways. Sporopollenin deposition and its anchoring to the plasma membrane appear to not be involved in regulating plasma undulation. Our data support the idea that plasma undulation occurs independently of and may act as a primary template for positional sporopollenin deposition.

EFD plays a role in exine formation through a novel regulatory pathway as a putative de novo DNA methyltransferase

Methylation of the C5 position of cytosine by DNA methyltransferase is the most common and important modification of DNA in eukaryotes (Martin & McMillan, 2002). Several DNA methyltransferases have been found in plants (Finnegan & Kovac, 2000; Lam et al., 2007; Pavlopoulou & Kossida, 2007; Liu et al., 2010). Research on Arabidopsis MET1 showed that a lack of MET1 leads to a global reduction in cytosine methylation throughout the genome and results in a late-flowering phenotype (Kankel et al., 2003). A study of DMT102 and DMT103 in maize revealed that an active DNA methylation pathway during reproduction is essential for gametophyte development and is likely to have a central role in the differentiation between apomictic and sexual reproduction (Garcia-Aguilar et al., 2010). Conserved plant genes with similarity to mammalian de novo methyltransferases have also been identified in Arabidopsis and maize databases. Members of the Arabidopsis DOMAINS REARRAGED METHYLASE (DRM) family, DRM1 and DRM2, and the maize protein Zmet3 have been identified as de novo DNA methyltransferases (Cao et al., 2000; Zhu et al., 2005; Greenberg et al., 2011).

Structural analysis and in vitro methyltransferase activity assays indicated that EFD is a de novo DNA methyltransferase. The phylogenetic analysis also revealed that the EFD is likely to be a member of the Dnmt3 family, the members of which perform de novo DNA methylation. Methyltransferases of the Dnmt3 family have been identified in many plant and animal species, including mice, humans, zebrafish, Arabidopsis and maize (Okano et al., 1998; Xie et al., 1999; Cao et al., 2000). MusDnmt3a and MusDnmt3b are essential for development in mice, and the inactivation of these two genes blocks de novo methylation in embryo stem cells and early embryos (Okano et al., 1999). The exogenous expression of mouse Dnmt3 induced apoptosis in Xenopus early embryos (Kimura et al., 2002). MusDnmt3a has also been confirmed to play an essential role in paternal and maternal imprinting in mice (Kaneda et al., 2004). In Arabidopsis, the DRM family is known to be important for DNA de novo methylation and gene silencing. The loss of DRM1 function results in slow germination, lower germination and growth rates, curling leaves, abnormal flower organs and late flowering in Arabidopsis (Zhu et al., 2005). DRM2 is a homolog of the mammalian de novo methyltransferase DNMT3 and is known to catalyze de novo methylation together with the involvement of other factors (Greenberg et al., 2011). However, EFD is the first methyltransferase that has been found to be involved in exine formation and male fertility. EFD may catalyze de novo DNA methylation and influence the expression of genes involved in callose synthesis and primexine formation during early microsporogenesis. This may explain why efd-affected downstream genes overlap with ms1-, ems1- or spl-affected genes, but the expression of MS1, EMS1 and SPL does not change in efd and EFD expression does not vary in ms1, ems1 or spl. However, further experimental work is required to investigate the direct target of EFD. This would provide further insights into the molecular mechanisms underlying how DNA methylation works in pollen exine development.


We thank Professors Hong Ma (Fudan University) and Zhongnan Yang (Shanghai University) for their useful discussion and suggestions. This work was supported by the ‘973’ project (2013CB126900; 2013CB945100), Key Project of Chinese Ministry of Education (311026), and the National Natural Science Foundation of China (31070284).