Specification of basal region identity after asymmetric zygotic division requires mitogen-activated protein kinase 6 in rice

The formation of body axes, such as the apical-basal axis, is a key step in early embryogenesis in multicellular organisms. The shoot and root apical meristems in flowering plants are formed according to the positional information based on the apical-basal axis (Jürgens, 2001). In Arabidopsis thaliana, mutations affecting apical-basal axis formation often result in severe defects in embryogenesis (Mayer et al., 1991).

The zygote divides asymmetrically, producing a small apical cell with a dense cytosol and a large basal cell with well-developed vacuoles in Arabidopsis (Jürgens, 2001). The apical cell generates the pro-embryo, and the basal cell divides to produce the extra-embryonic suspensor and hypophysis. Mutations affecting asymmetric zygotic cell division affect specification of the basal or apical region (Bayer et al., 2009; Jeong et al., 2011; Lukowitz et al., 2004; Yu et al., 2016). This division is regulated by a mitogen-activated protein kinase (MPK) cascade that includes YODA (YDA) MPK kinase (MPKKK) in Arabidopsis embryo (Lukowitz et al., 2004). The loss-of-function yda mutants lack basal cell identity and the suspensor, whereas constitutively active YDA exaggerates suspensor development. YDA also regulates asymmetric cell division in stomatal development (Bergmann et al., 2004). MPK3 and MPK6 function downstream of YDA (Wang et al., 2007). Similar to the yda mutant, the loss-of-function mpk3 mpk6 double mutant undergoes zygotic cell division with two similar daughter cells and loses the basal domain (Wang et al., 2007). It has also been shown that MPK6 and its upstream MKK4/5 in Arabidopsis are involved in embryo development (Bush and Krysan, 2007; López-Bucio et al., 2014; Zhang et al., 2017). The YDA-MPK3/6 pathway is activated by the interleukin 1 receptor-associated kinase (IRAK)/Pelle-like kinase gene SHORT SUSPENSOR (SSP) transcribed in sperm cells and by the peptide named EMBRYO SURROUNDING FACTOR 1 (ESF1), which accumulates in the central cell before fertilization (Bayer et al., 2009; Costa et al., 2014). Thus, the YDA-MAPK3/6 pathway, which is activated by stimuli from the parental origin regulates asymmetric cell division and results in the specification of the basal region. Thus, in Arabidopsis, asymmetric zygotic division is directly linked to apical-basal patterning.

WUSCHEL-RELATED HOMEOBOX (WOX) transcription factors are early markers for the apical or basal cell fate in Arabidopsis. The expression of WOX8 and WOX9 is regulated by WRKY2, the activity of which is regulated by YDA pathway-mediated phosphorylation in the basal cell and specifies its fate (Haecker et al., 2004; Ueda et al., 2011, 2017). WOX8 and WOX9 also non-autonomously promote the expression of WOX2 in the apical cell (Breuninger et al., 2008; Haecker et al., 2004).

In monocots, the mechanism of polarity establishment and organogenesis in early embryogenesis is less studied than in Arabidopsis, but orthologues of many genes involved in early embryogenesis in Arabidopsis are found in monocots (Itoh et al., 2016; Zhao et al., 2017). The expression of ZmWOX9A and ZmWOX9B, which are WOX8 and WOX9 counterparts in maize, is detectable shortly after fertilization, but that of ZmWOX2, an orthologue of Arabidopsis WOX2, is detected much later in maize embryogenesis (Chen et al., 2017; Nardmann et al., 2007). Thus, the involvement of WOX genes in apical and basal patterning is not the same in Arabidopsis and maize.

In this study, we have analyzed the globular embryo 4 (gle4) rice mutant and propose a model for apical-basal patterning in early embryogenesis in rice. This mutant produces a round-shaped embryo without organ differentiation (Hong et al., 1995; Kamiya et al., 2003). We found that GLE4 encodes OsMPK6, the Arabidopsis orthologue of which regulates asymmetric cell divisions in the zygote and meristemoid mother cell (Wang et al., 2007). Three groups have reported mutations in this gene in rice (Liu et al., 2016; Minkenberg et al., 2017; Yi et al., 2016); however, these studies did not analyze asymmetric zygotic division or apical-basal patterning.

We examined the phenotypes of gle4/osmpk6 with a focus on zygotic asymmetric division and apical-basal patterning. We found that GLE4/OsMPK6 is essential for cellular differentiation rather than for asymmetric zygotic division and determines the basal region at the early stage of rice embryogenesis. Our findings provide a new insight into cellular differentiation and asymmetric division through the function of MPK signaling, and raise the possibility that the asymmetric zygotic division and cellular differentiation that lead to apical-basal polarity can be separately regulated.

The original gle4 lines, gle4-1 and gle4-2, were described as globular shaped embryo mutants and were isolated from a collection of organogenesis-defective mutant rice stock (Hong et al., 1995); the other tree gle4 lines, gle4-3, gle4-4 and gle4-5, were obtained by screening mutants with a similar phenotype to gle4 from chemical- and transposon-induced mutant populations. All five lines showed segregation of mutant embryos in a manner of single recessive inheritance (Table S1). All five lines are heterozygotes and we have never recovered homozygous plants, suggesting embryo lethality of these mutations. The embryo phenotype in all five mutant alleles is stable and similar (Fig. S1). The gle4 embryos were globular, but were not globular arrested, because they grew beyond the globular stage while keeping the shape globular (Fig. 1B,C). Until 3 days after pollination (DAP), the gle4 embryos were morphologically indistinguishable from wild-type embryos (data not shown, see Fig. 3 upper and lower panels). At 4 to 6 DAP, wild-type embryos started to form the coleoptile, shoot apical meristem and root apical meristem (Fig. 1D-F), whereas gle4 failed to form discrete organs or tissues except for epithelium-like epidermis and a mass of small cells that resembled vascular tissues at the center (Fig. 1G-I). This is consistent with a previous report showing the expression of several protodermal genes such as ROC1 and RAmy1A, and a vasculature-specific gene, OsPNH, in gle4 (Kamiya et al., 2003). The morphological phenotypes are similar to previously reported osmpk6 mutants in rice, although the expression pattern of protodermal genes is different (Yi et al., 2016; see later). At 9 DAP, organogenesis in wild-type embryos was completed, but most of the organs were missing in gle4 embryos (Fig. 1A-C).

In the wild type, the suspensor at the base of the embryo connects it with mother tissue (Fig. S2) and is composed of a population of round-shaped cells resembling a cluster of grapes; no such cell population was found at the base of mutant embryos (Fig. S2). Overall, the gle4 mutant phenotype suggests that GLE4 regulates organogenesis during embryogenesis.

We cloned GLE4 by transposon tagging using gle4-3, which was derived from a TOS17 transposon mutagenesis library (Figs S3 and S4). We found that GLE4 encoded OsMPK6 (locus ID Os06g0154500), a member of the MPK family and an orthologue of Arabidopsis MPK6. We conducted molecular complementation using gle4-3 mutant homozygous calli. As described by Kamiya et al. (2003), it is possible to induce and culture gle4 mutant homozygous calli on culture medium, although mutant calli do not regenerate normal shoots but produce many small leaf like organs on the regeneration medium (Kamiya et al., 2003). We used this character to assess molecular complementation. We infected gle4-3 homozygous mutant calli with agrobacterium harboring the complementation vector carrying OsMPK6 cDNA under the control of ubiquitous promoter, and placed on the regeneration medium to see whether the normal shoots regenerate from mutant calli. As shown in Fig. S5, the normal shoots and plants were regenerated and recovered from gle4-3 mutant homozygous calli by the introduction of the complementation vector, whereas the introduction of kinase inactive mutant version of OsMPK6 cDNAs did not regenerate normal shoots. Thus, molecular complementation analysis supported the cloning of GLE4. gle4-2 and gle4-3 carry a splice site mutation and Tos17 transposon insertion, respectively, within the conserved kinase catalytic domain, and gle4-4 and gle4-5 carry nonsense mutations within the conserved kinase catalytic domain (Fig. S4). gle4-1 carries missense mutation (G213E) in the conserved kinase catalytic domain. The glycine residue at 213 amino acid residue of OsMPK6 is highly conserved residue among MPK family in plants (Fig. S4). Based on the similarity of the phenotype and the mutant lesions of five alleles, we consider that those are all null alleles. This is consistent with previous reports, which showed that the loss-of-function mutations in OsMPK6 result in similar embryo lethality (Yi et al., 2016; Minkenberg et al., 2017). Yi et al. (2016) reported two insertion alleles of OsMPK6, osmpk6-1 and osmpk6-2. Although, they argue that there are some difference in the expression of epidermal marker genes such as ROC1 or RAmy1A between osmpk6 and gle4 described by Kamiya et al. (2003), this could be merely due to the difference in the timing of embryo RNA sampling.

Because the embryos of Arabidopsis MPK pathway mutants (such as yda or mapk3 mapk6 double mutant) fail in the asymmetric zygotic division and often have no discrete organs (Lukowitz et al., 2004; Wang et al., 2007), we investigated the cell division patterns of the rice zygote and early embryo of the wild type and gle4.

Fig. S6 shows the course wild-type embryogenesis in rice at early stages. In the wild-type embryos, zygotes divided into small apical and large basal cells (Fig. 2A), although the elongation of the zygote was not as evident as in Arabidopsis (Lukowitz et al., 2004). Apical and basal cells divided to form the four-cell stage embryo (Fig. 2B). The angle between the planes of this second round of cell division appeared to be random (Fig. 2E), which result in various arrangements of the four embryonic cells. The cell divisions therefore seem to occur unpredictably, which is commonly observed in monocots and dicots, although not in Arabidopsis (Wardlow, 1955).

To compare the pattern of zygotic and early embryo cell divisions in gle4 mutants, we analyzed gle4 segregating ovaries by self-pollination of heterozygous gle4 plants; we expected to recover approximately 25% of mutant embryos (Table S1). In gle4 segregating ovaries, all of the 17 two-cell embryos showed smaller apical and larger basal cells similar to those of the wild type, suggesting no obvious differences in zygotic division between the wild type and gle4 (Fig. 2C,D). In all of the 35 gle4 segregating ovaries, the four-cell-stage embryos had a small apical side and large basal side; the cellular arrangement was indistinguishable between gle4 and the wild type (Fig. 2B,F-H). To confirm the latter observation, we assessed the cell arrangement by mapping the positions of the nuclei (Fig. 2F-H). By setting virtual planes between the nuclei of the two daughter cells of the apical or basal cells, we measured the angles of the intersections of the two planes, compared their distributions between the wild type and the mutant segregating population (Fig. S7), and found no difference (Fig. 2F-H). This suggests that OsMPK6 is dispensable for cell division patterning at an early stage of embryogenesis in rice.

Asymmetric zygotic division in both the wild-type and gle4 embryos suggested that apical-basal polarity was normal in the mutant embryos. To confirm this, we tested marker gene expression in the wild type and mutant segregating populations. We have previously developed a set of marker genes that indicate regional differentiation at early stages of rice embryogenesis (Itoh et al., 2016). We used one apical marker and four basal markers for in situ hybridization analysis at 3 DAP. Unexpectedly, the apical marker was expressed in the entire gle4 embryo, whereas the expression of basal marker genes was mostly abolished (Fig. 3). Thus, in the gle4 embryo, the identity of the basal region is lost, and the cells acquire apical character.

We further confirmed this by transcriptome analysis of RNA recovered by laser microdissection (LMD) from 3 DAP embryo of wild type and gle4-1. Briefly, we extracted total RNA from embryos recovered by LMD then genotyped each embryo by DNA sequencing after cDNA synthesis, because, at 3 DAP there is no morphological difference and we could not distinguish wild-type and gle4-1 embryos. Mixtures of RNA from three mutant or wild-type embryos were then subjected to transcriptome analysis with triplicates for both (Fig. S8). A summary of RNA-sequencing experiment is shown in Table S2. We detected expression of 20,737 and 20,834 genes in wild-type and mutant embryo, respectively. Among them, 725 genes and 604 genes were differentially expressed in gle4-1 at high and low levels, respectively at FDR<0.001 (Figs S9 and S10). A list of differentially expressed genes is shown in Tables S3 and S4. We compared these differentially expressed genes (DEGs) with the list of genes with localized expression in 3 DAP embryo at the apical and basal sides that were identified in our previous work (Itoh et al., 2016) (Fig. 4A-C). We discovered that the majority of genes with low expression in gle4-1 were genes expressed at the basal side in wild type (Fig. 4A,B), and those with high expression in gle4-1 were genes expressed at apical side in the wild type (Fig. 4A,C). The transcriptome analysis also suggests that, in the gle4 embryo, the identity of the basal region is lost and the cells acquire apical character.

Next, we compared DEGs with the list of genes expressed at shoot, scutellum, epiblast and radicle at 7 DAP of wild type (Itoh et al., 2016) (Fig. 4D-F). The majority of genes with high expression in gle4-1 were scutellum-specific genes at 7 DAP. This suggests that, at 3 DAP in the wild type, the apical side is destined to become scutellum and the basal side mainly constitutes the remaining parts such as shoot, epiblast and radicle. We further investigated the expression of rice OsWOX2 and OsWOX8/9, whose Arabidopsis orthologues are expressed in apical (WOX2) and basal (WOX8/9) cells (Zhang et al., 2010). In the transcriptome data, OsWOX2, OsWOX8/9A and OsWOX8/9B expression was detected in the wild type and the expression levels of both genes were decreased or not detected in gle4-1, although the expression level of OsWOX2 and OsWOX8/9B are very low in the wild type at this stage (Fig. 4G). Because WOX8/9 expression in Arabidopsis resides in the basal region, the reduction of OsWOX8/9 in gle4-1 is expected. Interestingly, OsWOX2 expression is also decreased in gle4-1. Thus, the expression pattern of OsWOX2 at early embryogenesis is not the same as WOX2 in Arabidopsis.

To test whether OsMPK6 is involved in the establishment and/or maintenance of basal region identity, we examined the expression of one apical and one basal marker at 1 and 2 DAP. In the wild type, the basal marker was expressed in the entire embryo at 1 DAP (Fig. 5A) and its expression became localized at the embryo base at 2 DAP (Fig. 5E); the apical marker was undetectable at 1 DAP but started to be expressed at 2 DAP (Fig. 5C,G). In gle4 segregating embryos, the expression of the apical marker was not detected at 1 DAP (Fig. 5D), but it appeared in the entire embryo at 2 DAP (Fig. 5H); the basal marker was not detected at 1 or 2 DAP (Fig. 5B,F). Overall, these data indicate that the entire rice embryo has basal identity at 1 DAP, the apical region is established at 2 DAP and GLE4/MPK6 is indispensable for the establishment and/or maintenance of the basal region (Fig. 6).

The molecular genetic analysis of the model plant A. thaliana has deepened our knowledge of the embryogenesis of flowering plants. The simple and stereotyped pattern of cell divisions in Arabidopsis embryos at early stages facilitates the tracing and prediction of cell divisions until a fairly late stage (e.g. heart stage). Most textbooks introduce the Arabidopsis embryo as a model for embryogenesis in flowering plants, but many flowering plants undergo embryogenesis with more unpredictable cell divisions after zygotic division (Wardlow, 1955). The stereotyped cell divisions imply that cellular lineage is important for regional differentiation in embryogenesis, but the basic body pattern of the plant embryo is maintained even in mutants with disturbed cell division patterns in early embryogenesis (Torres-Ruiz and Jürgens, 1994; Spinner et al., 2013). This suggests that regional differentiation in the embryo does not depend on cellular lineage information.

Using rice gle4 mutants, we showed that GLE4 encodes OsMPK6, an orthologue of Arabidopsis MPK6 and that gle4 mutants undergo asymmetric zygotic division similar to that of the wild type. We also found that the osmpk3 gle4/ osmpk6 double mutant rice embryo is indistinguishable from that of the gle4/osmpk6 single mutant (Fig. S11). In addition, our previous analysis of gene expression profiles in early stages of rice embryogenesis revealed that OsMPK3 expression is very low (Itoh et al., 2016; RICEXPRO database, ricexpro.dna.affrc.go.jp). These suggest that the MPK cascade that includes OsMPK3 and OsMPK6 is dispensable for asymmetric zygotic division in rice.

In zygotic division in rice, small apical and large basal cells are formed (Fig. 2A,C,D); the cells then divide to form a four-cell embryo. We showed that the arrangement of the daughter cells of the apical and basal cells is arbitrary both in the wild type and gle4. This is consistent with the observations that many monocot and dicot species undergo non-stereotyped cell divisions after asymmetric zygotic division (Wardlow, 1955). Because the first sign of morphological abnormality in gle4 is observed in the 4 DAP embryo, which consists of more than 1000 cells, we speculate that GLE4 acts in regional specification at earlier stages and/or it functions in organogenesis. Using a set of region-specific markers in the early rice embryo, we showed that GLE4 is pivotal for the establishment or maintenance of the basal region, which occurs by 1 DAP (Figs 3-5).

Mutations in the Arabidopsis YDA (MPKKK)-MPK3/6 cascade lead to a defect in asymmetric zygotic division and result in two daughter cells of similar size (Lukowitz et al., 2004). This is similar to the gle4 phenotype because in both cases the basal part is absent, suggesting that the fundamental function of the MPK3/6 pathway in early embryogenesis is the establishment and/or maintenance of the basal region. In Arabidopsis, the establishment/maintenance of the basal region directed by the MPK3/6 pathway occurs at the asymmetric zygotic division, whereas in rice it occurs after this division. In rice, asymmetric zygotic division and MPK3/6-dependent basal specification can be uncoupled and are controlled by different systems. We cannot exclude the possibility that yet unknown genes are involved specifically in the apical or basal cell specification after zygotic division both in the wild type and gle4, i.e. the presence of another layer of apical-basal polarity establishment independent of the MPK3/6 pathway.

Analysis using our marker gene sets suggested that, in early rice embryo, the basal identity may appear first by all cells by 1 DAP, and then the apical marker starts to be expressed. The transcriptome analysis of 3 DAP embryo also revealed that the apical region mainly contributes to the scutellum and the rest of the embryonic organs such as the shoot, epiblast and radicle are from the basal region. This is the reason why OsWOX2 and many other SAM-specific genes, such as OSH1, were downregulated in gle4 (Fig. 4G) (Kamiya et al., 2003; Sato et al., 1996). Interestingly, the loss of basal identity in gle4 leads to expansion of apical character in mutant embryo, suggesting there is a competitive regulation and interaction between apical and basal regions (Fig. 6). Similar interactions between embryonic domains have been reported in Arabidopsis (Breuninger et al., 2008; Smith and Long, 2010). Even though the origin of the shoot apical meristem is different between rice and Arabidopsis, the interaction between apical and basal regions seems conserved.

In the asymmetric zygotic division in Arabidopsis, the YDA-MPK3/6 cascade is activated by a paternal gene product named Short Suspensor (SSP), a cytoplasmic receptor kinase (Bayer, et al., 2009). In the asymmetric division of meristemoid mother cells, this MPK pathway regulates BASL, which shows polarized localization in cells (Dong et al., 2009). No clear orthologue of SSP or BASL is encoded in the rice genome, suggesting that, although this MPK pathway, including MPK3/6, is conserved, its input and output have diverged among flowering plants. This is consistent with the fact that MPK3/6 signaling is activated by changing biotic stresses such as pathogen infections. The signals upstream and downstream of the rice GLE4/OsMPK6 pathway involved in regional specification in the early rice embryo remain to be elucidated.

The gle4-1, gle4-2, gle4-4 and gle4-5 seeds were obtained from chemically mutagenized populations, and gle4-3 seeds were from a Tos17-mutagenized population (gle4-3). To observe cell division patterns at 18 and 24 h after pollination (HAP), gle4-1^+/− and Taichung 65 seedlings were grown in a controlled growth chamber as described by Ohnishi et al. (2011). To obtain samples for in situ hybridization and histological observations of 4-9 DAP embryos, gle4-1^+/− and gle4-3^+/− seedlings were grown in a paddy field.

The heterozygous lines of gle4-3 were grown in paddy field and 2-3 g of leaves were harvested from each plant for genomic DNA extraction (Agrawal et al., 2001). We harvested seeds from each plant after self-pollination, and determined the genotype of each plant by counting the segregation of mutant embryo as wild type (WT) or heterozygous (H). Based on this genotype, we used the linkage with the TOS17 insertion described by Agrawal et al. (2001). The flanking sequence of TOS17 insertion was recovered using a thermal asymmetric interlaced PCR method followed by DNA sequencing (Miyao et al., 2003).

For molecular complementation test, a 1.7 kb cDNA encoding OsMPK6 (Os06g0154500) was inserted downstream of the 35S promoter in the binary vector pB1101-Hm. The procedure to introduce the kinase inactivation mutations to OsMPK6 is described in Kishi-Kaboshi et al. (2010). The resulting plasmid was introduced into Agrobacterium tumefaciens EHA101, which was then used to transform mutant calli as described by Hiei et al. (1994).

Pistils containing embryos at 4-9 DAP were cut in half and fixed in 37% formaldehyde:glacial acetic acid:70% ethanol (1:1:18) under vacuum three times for 15 min each and then overnight. Fixed embryos were hydrated through a graded ethanol series with a microwave processor (250 W, 1.5 min each step) on ice. Solution was replaced with 1×PBS, and the embryos were stained with 50 μg/ml propidium iodide containing 100 μg/ml RNase in 1×PBS overnight in the dark at 37°C. Embryos were dehydrated through an ethanol series and cleared by methyl salicylate with a microwave processor (300 W, 1.5 min each step) on ice. Propidium iodide fluorescence emission at 570-670 nm (excitation at 559 nm) was captured using confocal microscopes (FV1000 and FV3000; Olympus).

To prepare samples for scanning electron microscopy, 7 DAP embryos were fixed in 4% paraformaldehyde and 0.25% glutaraldehyde in 0.05 M sodium phosphate buffer (pH 7.2) under vacuum at room temperature for 1 h and then overnight. The samples were dehydrated through a graded ethanol series and t-butanol series, and t-butanol was sublimated under vacuum at a sub-zero temperature. Samples were coated with platinum ion and observed under a scanning electron microscope (S-3000N; Hitachi).

To observe cell division patterns during early embryogenesis, ovules in pistils isolated at 18 and 24 HAP were stained by the pseudo-Schiff method (Vollbrecht and Hake, 1995). Acriflavine fluorescence emission at 530-630 nm (excitation a 515 nm) was captured using the FV1000 microscope.

Using coordinates of the nuclei in confocal images of embryos at the two- or four-cell stage, we plotted nucleus positions and constructed a 3D model (see Fig. S7). At the two-cell stage, the nuclei of the apical and basal cells were connected with a line. The distance from the top of the apical cell or the bottom of basal cell to the plane perpendicular to this line was measured. For the four-cell stage, nuclei of each apical and each basal cell were connected by lines (A_apical and A_basal). Planes perpendicular to these lines were named plane 1 and plane 2, respectively. Plane 3 was perpendicular to line B, which connected the midpoints of the A_apical and A_basal lines. The angles between planes 1 and 3, and planes 2 and 3, and between the intersections of planes 1 and 3 (line C) and planes 2 and 3 (line D) were measured and are presented in Fig. 2F-H, respectively. The 3D-model construction and the measurements were performed using a FV10-ASW 4.2 viewer (Olympus) and Metasequoia (Tetraface) software.

Sections (8 μm) were cut with a rotary microtome. In situ hybridization was performed as described previously (Kouchi and Hata, 1993). To produce DIG-labeled OsMPK6 sense and antisense probes, a 900 bp DNA fragment was amplified by PCR from a cDNA clone provided by NIAS Japan (AK111942) and used as a template for in vitro transcription with a Maxi Script in vitro transcription kit (Thermo Fisher Scientific). Probes for transcripts specific for the apical or basal region were prepared as described previously (Itoh et al., 2016). Hybridization was conducted at 52°C overnight.

Ovaries of gle4-1 heterozygous plants at 3 DAP were fixed in 75% ethanol:25% acetic acid. After dehydration in a graded ethanol series, the tissues were embedded in paraffin and sectioned at f 16 µm. Complete serial sections of embryo regions were placed onto PEN membrane glass slides (Leica) as described previously (Takahashi et al., 2010). After drying at 42°C, slides were immersed in 100% Histoclear II (National Diagnostics) for 10 min twice to remove paraffin, followed by air-drying at 4°C. Three or four sections covering the entire one embryo were collected into a tube from sections using a Leica LMD6000 Laser Microdissection system.

Total RNAs were extracted from the LMD-isolated tissues using a PicoPure RNA isoloation kit (Life Technologies) according to the manufacturer's instructions. The qualities and concentrations of RNAs were measured using a 2100 Bioanalyzer (Agilent Technologies) with RNA6000 Pico kit (Agilent Technologies).

After dissection of 131 embryos and extraction of RNAs, we genotyped each embryo as follows. From the embryo RNAs, cDNAs were synthesized using One Step SYBR PrimeScript RT-PCR Kit II (TAKARA) and then subjected to PCR amplification and DNA sequencing using primers (F4: GATACTGATCTGCATCAAATTA, R1: CAGCCCACAGACCACACATC) that amplify around the region of the gle4-1 mutation site to determine the genotype of each embryo.

Based on the results of genotyping mentioned above and RNA quality check, three mutants and three wild-type embryo RNAs with higher RNA quality were mixed into a single tube in triplicate. Thus, we prepared three wild-type and gle4-1 samples using nine embryos for each. These RNA samples were converted to cDNA using the SMARTer Ultra Low Input RNA Kit for Sequencing-v4 (Clontech) according to the manufacturer's instructions. After converting cDNA, sequencing libraries were constructed using Nextera XT DNA Library Preparation Kit (Illumina). Sequencing was carried out with the Illumina HiSeq 2500 platform and the resulting reads were mapped to the reference genome of Oryza sativa (IRGSP 1.0) using RAP-DB annotations with STAR (Dobin et al., 2013). The ‘quantMode’ argument of STAR was used for reads counting. Transcript expression was evaluated using the EdgeR package in R (Robinson et al., 2010), and transcript abundance was estimated by counts per million mapped fragments (CPM). Differentially expressed genes were selected using FDR with P<0.001 and fold change (FC)>2.

A method from Mikami et al. (2015) was used. The target sequence of OsMPK3 was selected as 5′-AGCTTACGTTCTCGTGGTCG-3′ (antisense strand) by CRISPR-P v1.0 (cbi.hzau.edu.cn/cgi-bin/CRISPR). Selection criteria were as follows: (i) cleavage site in an exon; (ii) aggregate scoring for all possible gRNAs (guide RNAs) S_guide>95; and (iii) close to the translation start site.

Seeds of a heterozygous plant with the o-45 allele [OsMPK6 (+/−)] were used for callus induction as in (Toki et al., 2016). Several calli were genotyped on the basis of the EcoRV digestion pattern of PCR products amplified with the following dCAPS primers: o-45_F, 5′-GCCACATGGACCACGAGGAT-3′; o-45_R, 5′-CAAAATCTGCCTAAAAATCGAG-3′. PCR was conducted following the manufacturer's instructions using GoTaq Colorless Master Mix (Promega). As homozygous OsMPK6 (−/) calli were not obtained, heterozygous OsMPK6 (+/−) calli were used to generate transgenic plants. To create an osmpk3 single mutant and a osapk3 osmpk6 double mutant, OsMPK6 (+/−) calli were transformed with the OsMPK3 CRISPR vector as described previously (Hiei et al., 1994). Each T1 seed of OsMPK3 CRISPR/OsMPK6 (+/−) T0 plants was divided into the embryo and endosperm. Endosperm genomic DNA was used for genotyping. The OsMPK3 genotypes of the T1 seeds were determined by sequencing the PCR products with the following primer set: MPK3_genotyping_F1, 5′-AGCGTAGTGGTTGACTGGTTG-3′; MPK3_genotyping_R1, 5′-CACAGTTCACTCACCTGGCAG-3′. PCR was conducted following the manufacturer's instructions with GoTaq Green Master Mix (Promega). The OsMPK6 genotypes of the T1 seeds were determined as above. After genotyping, embryos of the wild-type, osmpk3 single mutant, osapk6 single mutant and osmpk3 osmpk6 double mutant were stained with Propidium Iodide, cleared with methyl salicylate and observed under a confocal laser scanning microscope (FV1200; Olympus).

We thank Dr Masaki Endo for providing vectors for CRISPR/cas9 experiments.