Conserved functional control, but distinct regulation, of cell proliferation in rice and Arabidopsis leaves revealed by comparative analysis of GRF-INTERACTING FACTOR 1 orthologs

Leaves are major above-ground parts in most plant species. Leaf size and shape is species specific, but leaf primordium development begins as a lateral protrusion of the shoot apical meristem (SAM) in all plant species. The variation in leaf size and shape is determined by several developmental processes driven by common and diverged genetic programs after leaf initiation. Genetic variation and conservation of these programs during leaf development is a major topic in plant developmental biology (Tsukaya, 2014, 2017).

Regulation of cell proliferation is an important determinant of leaf size and shape (Ichihashi and Tsukaya, 2015; Tsukaya, 2017). Thus, many genes associated with leaf cell proliferation have been analyzed (Gonzalez et al., 2012; Nelissen et al., 2016). Members of the TEOSINTE-BRANCHED1/CYCLOIDEA/PCF (TCP) and NGATHA (NGA) gene families, which encode plant-specific transcriptional factors, are involved in several aspects of leaf development by repressing cell proliferation in the marginal meristems of leaves of the snap dragon, Arabidopsis, tomato, creeping bentgrass and rice (Ori et al., 2007; Efroni et al., 2008; Hervé et al., 2009; Kieffer et al., 2011; Yang et al., 2013; Zhou et al., 2013; Alvarez et al., 2016). KLUH (KLU) and PLASTOCHRON1 (PLA1) are other type of regulators of cell proliferation during leaf development in Arabidopsis and rice, respectively (Miyoshi et al., 2004; Anastasiou et al., 2007; Mimura and Itoh, 2014). KLU/PLA1 encodes a member of the cytochrome P450 monooxygenase subgroup, CYP78A, although its substrate is unknown. Because loss-of-function mutants of KLU and PLA1 exhibited smaller leaves with fewer cells, KLU/PLA1 was proposed to be involved in production of an unknown signaling molecule that positively regulates cell proliferation in leaves (Anastasiou et al., 2007). Overexpression of KLU, PLA1 and ZmPLA1, which are maize PLA1 orthologs, resulted in larger leaves and reproductive organs, indicating that KLU/PLA1 positively regulates cell proliferation (Hibara et al., 2016; Sun et al., 2017).

GROWTH-REGULATING FACTOR (GRF) genes are plant-specific transcriptional regulators that play important roles in the regulation of cell proliferation (Kim and Tsukaya, 2015). The Arabidopsis genome harbors nine GRF genes, most of which are expressed in organs and tissues with high growth activity, and positively regulate cell proliferation (Kim et al., 2003; Horiguchi et al., 2005; Kim and Lee, 2006). In addition, seven out of the nine GRF genes are targeted by the microRNA miR396; thus, the site of GRF action is dependent on the miR396 accumulation pattern (Jones-Rhoades and Bartel, 2004; Liu et al., 2009; Rodriguez et al., 2010; Wang et al., 2011). In monocot species such as rice and maize, most GRF genes positively regulate cell proliferation during leaf development by a mechanism similar to that in Arabidopsis (van der Knaap et al., 2000; Choi et al., 2004; Li et al., 2010; Nelissen et al., 2015), although maize ZmGRF10 reportedly negatively regulates cell proliferation (Wu et al., 2014).

GRF-INTERACTING FACTOR (GIF) genes also control cell proliferation during leaf development (Kim and Tsukaya, 2015). GIF was first identified as a protein partner that physically interacts with GRFs. GIFs function as transcriptional co-activators for GRFs and other transcription factors by forming complexes with SWI2/SNF2 chromatin remodeling-related proteins (Vercruyssen et al., 2014). Among the three GIF genes in the Arabidopsis genome, GIF1, also known as ANGSTIFOLIA3 (AN3), positively regulates cell division in young leaf primordia (Kim and Kende, 2004; Horiguchi et al., 2005). Importantly, in Arabidopsis leaf primordia, AN3/AtGIF1 mRNA is detected only in mesophyll cells of the basal part of leaf primordia (Kawade et al., 2013) and AN3/AtGIF1 protein moves between cells between epidermis and parenchymatous cells (Kawade et al., 2013), which is required for proper leaf organogenesis. Recently, it has been shown that the AN3/AtGIF1 protein is required to move among cells lying along the longitudinal axes of leaf primordia to ensure appropriate positioning of the leaf plate meristem (Kawade et al., 2017). Although GIF1, GIF2 and GIF3 redundantly regulate various aspects of plant development (Kim and Kende, 2004; Lee et al., 2009, 2014), a single loss-of-function mutation in AN3/GIF1 resulted in smaller and narrower leaves with reduced cell proliferation (Horiguchi et al., 2005; Vercruyssen et al., 2014). In grass species, the functions of GIF1 orthologs are largely conserved, as are the proteins with which these factors interact (Nelissen et al., 2015). OsGIFs physically interact with OsGRF4/GLW2, OsGRF6 and OsGRF10, and activate their transcription (Liu et al., 2014; Li et al., 2016).

Compensation is a phenomenon closely related to cell proliferation in leaves (Tsukaya, 2002; Hisanaga et al., 2015). It is post-mitotic cell expansion in leaves that occurs when the cell number is decreased by mutations or overexpression of cell proliferation-related genes. The an3 mutant shows typical compensation syndrome, exhibiting smaller leaves with a reduced number of enlarged cells. Cell enlargement is not a direct effect of loss of AN3 function, because overexpression of AN3 does not affect cell size (Kim and Kende, 2004; Horiguchi et al., 2005; Lee et al., 2009). Similarly, overexpression of a cyclin-dependent kinase inhibitor gene, KRP2, in Arabidopsis resulted in suppressed cell proliferation and enlarged cells in leaves (Ferjani et al., 2013a,b). In addition, it has been revealed, using kinetic and genetic analysis of cell size enlargement, that the mechanisms that trigger compensation in an3 mutant and KRP2-overexpressing plants are different (Kawade et al., 2010; Ferjani et al., 2013b; Hisanaga et al., 2015). In rice, overexpression of a cyclin-dependent kinase inhibitor homolog, OsKRP1, induced cell enlargement by compensating for the reduced cell number. Thus, compensation occurs in diverse plant species (Barrôco et al., 2006; Horiguchi and Tsukaya, 2011).

Several pathways and mechanisms involved in leaf development are conserved among plant species. However, knowledge of how common genetic factors exert their functions in different developmental backgrounds is limited. In this study, we evaluated the phenotypic effects of loss of function and overexpression of rice MAKIBA3 (MKB3), an ortholog of Arabidopsis AN3. Expression analysis and use of a protein movement assay revealed the functional conservation, but divergent regulation, of orthologous genes in the two species.

mkb3 was identified as a recessive mutant that showed an abnormal leaf morphology. As mkb3 roots exhibited no obvious abnormality (Fig. S1), we investigated the gross leaf phenotype of mkb3. Because mkb3 showed shortened and narrowed leaves (Fig. 1A), we measured several parameters in the fifth leaf of mkb3 and the wild type. The length and width of the leaf blade were reduced by 72% and 75% compared with the wild type, respectively (Fig. 1F,G). In contrast, cross-sections revealed that the fifth leaf of the mkb3 leaf blade is rolled adaxially and was 156% thicker than that of the wild type (Fig. 1B,C,H).

To understand how the reduction in leaf size is related to the cell size of mkb3 leaves, we evaluated the size of epidermal cells along the proximal-distal, medial-lateral and adaxial-abaxial axes of the leaf (Table 1). The length of adaxial epidermal cells of the leaf sheath along the proximal-distal axis in mkb3 plants was significantly greater than in the wild type. The width of epidermal cells of the abaxial leaf blade along the medial-lateral axis in the mkb3 mutant was slightly larger, and the thickness of bulliform cells, which are specialized epidermal cells on the adaxial surface of the leaf blade, was significantly greater in mkb3 plants than in the wild type (Fig. 1D,E and Table 1). Accordingly, cells were generally enlarged along all axes of the leaf, despite the reduced leaf size along the proximal-distal and medial-lateral axes. Therefore, the number of cells along the proximal-distal and medial-lateral axes in leaves of mkb3 plants were calculated to be lower than that in wild type, but those along the adaxial-abaxial axis are unaffected.

Cell enlargement is often accompanied by increased nuclear ploidy. In some plant species, endoreduplication, which is DNA replication without mitosis, leads to increased leaf ploidy often correlated with epidermal cell size (Katagiri et al., 2016). Thus, the ploidy level of wild-type and mkb3 leaves was analyzed, although endoreduplication does not normally occur in rice leaves. The ploidy level of mkb3 leaves was normal, i.e. only the 2C peak was observed in both wild-type and mkb3 leaves (Fig. S2). Therefore, cell enlargement in the mkb3 mutant is not caused by endoreduplication.

We also investigated the vascular differentiation of wild-type and mkb3 plants using the fifth leaf blade (Table 1). Although the morphology of the vascular bundles was not changed (Fig. 1D,E), the number of small vascular bundles in the mkb3 mutant was significantly reduced. However, the intervals between vascular bundles were slightly increased (Fig. 1B,C). This indicates that the mkb3 mutation leads to a reduced number of vascular bundles, reflecting the reduced leaf width (Russell and Evert, 1985; Dannenhoffer et al., 1990), but the intervals between the vascular bundles were not influenced by the reduction in the leaf size.

mkb3 plants produced a panicle significantly reduced in length (Fig. 2A,B). In the wild type, the internode length decreases gradually from top to bottom. The mkb3 mutant also showed this tendency, but the top two internodes were greatly reduced in length (Fig. 2A,B). The length of the primary rachis branches was also reduced, but their number was not affected, indicating that MKB3 is involved in the elongation of internodes and rachis branches during panicle development (Fig. 2C-E).

mkb3 spikelets also exhibited morphological abnormalities. The shape of the lemma and the palea was distorted, and the width of the palea was significantly reduced (Fig. 2F,G). Although the number and shape of floral organs (lodicule, stamen and pistil) were not affected (Fig. 2H,I), differentiation of pollen in mkb3 anthers was incomplete (Fig. S3). In addition, abnormalities in integument elongation and ovule formation were observed in some mkb3 pistils (Fig. 2J,K). Therefore, MKB3 is also required for the differentiation of some floral tissues. These abnormalities may be the cause of the sterility of the mkb3 mutant.

We next identified MKB3 by a map-based cloning strategy. Rough and fine mapping using mkb3 homozygous plants derived from an F₂ population crossed between mkb3 heterozygotes and cv. Kasalath (wild type) predicted that the mutation was located between two markers on chromosome 3 (Fig. 3A). Among the predicted genes in two BAC contigs, we found a base substitution from G to A at the 3′ splicing site of the third intron of Os03g0733600 (Fig. 3A). Os03g0733600 was predicted to be an SSXT family protein with homology to the human transcription co-activator synovial sarcoma translocation protein.

To confirm that this mutation in Os03g0733600 was responsible for the phenotype of mkb3, a 8251 bp fragment of Os03g0733600, including the 3308 bp putative promoter and 1898 bp terminator regions, was introduced into mkb3 calli. The regenerated plants harboring the Os03g0733600 fragment showed a normal phenotype, whereas plants regenerated with empty vector had narrow, rolled leaf blades, identical to the mkb3 mutant (Fig. 3C,D). These data confirmed that Os03g0733600 is MKB3.

Phylogenetic analysis in several plant species revealed that MKB3 is an ortholog of AN3/GIF1, which functions as a transcriptional co-activator and positively regulates cell proliferation by interacting with GRF transcriptional regulators in Arabidopsis (Fig. 3E and Fig. S4), although the extent of amino acid identity and similarity between MKB3 and AN3 were not high: 50% and 57%, respectively. MKB3 protein has a conserved SNH domain, which is required for the interaction with GRF, at the N terminus (Fig. 3B) (Kim and Tsukaya, 2015). Accordingly, MKB3 is predicted to regulate cell proliferation by interacting with GRFs. Indeed, direct interaction between OsGIF1 (a synonym of MKB3) and GRFs has been reported (Duan et al., 2015). The mkb3 mutation generates a premature stop codon between the SNH domain of the N terminus and the glutamine- and glycine-rich domain of the C terminus of the MKB3 protein, which may indicate that the C termini of GIF1 homologs play important roles (Fig. 3A and Fig. S4).

To explore whether MKB3 was functionally equivalent to Arabidopsis AN3, we performed an interspecific complementation test. Introduction of MKB3 cDNA fused with GFP under the control of the AN3 promoter into the an3-4 strain almost rescued the leaf phenotypes (Fig. 3F,G,J and Fig. S5), as did the introduction of AN3-GFP (Fig. 3F-H and Fig. S5), indicating that AN3 can be substituted by MKB3 and that the protein functions are conserved between rice and Arabidopsis.

To assess the MKB3 expression pattern during the plant life cycle, we searched the rice gene expression database RiceXPro (ricexpro.dna.affrc.go.jp/). The results showed that MKB3 was highly expressed in tissues of young inflorescences, moderately in those of pistils, ovaries and early embryos, and at a low level in mature leaf blades and roots. Thus, based on the expression profile of MKB3 and the phenotype of the mkb3 mutant, it was predicted that MKB3 is predominantly expressed in young tissues with active cell proliferation (Fig. 4A), as has been shown for AN3/AtGIF1 (Horiguchi et al., 2005; Vercruyssen et al., 2014).

Next, we investigated the spatial expression pattern of MKB3 by in situ hybridization. In a longitudinal section of a shoot apex, MKB3 mRNA was detected mainly in the basal part of young leaf primordia (P1 to P4), and a strong signal was observed in the leaf margins of P2 and P3 primordia. In addition, mRNA was present at the abaxial boundary between a P3 leaf and stem, but not in the SAM (Fig. 4B). In cross-sections, MKB3 was expressed in the P1, abaxial epidermis and the marginal domain of P2 and P3, but not in the SAM, presumptive vascular bundles or adaxial domain of the inner tissue of P2 and P3 primordia. Expression decreased gradually from the abaxial to the adaxial side of P2 and P3 primordia, although weak expression was detected in the adaxial epidermis of the marginal region (Fig. 4C). In the distal region of the shoot apex, expression was evident in both the adaxial and abaxial epidermis of P2 primordia (a future part of the leaf blade), indicating that the expression pattern differed between the leaf blade and the sheath (Fig. 4D). During reproductive development, a ring-like MKB3 expression pattern was observed at the nodes of the spikelets and internodes of rachis (Fig. 4E). At the floral organ differentiation stage, MKB3 expression was mainly detected on the surface of floral organ primordia (Fig. 4F).

These results indicate that MKB3 is predominantly expressed in the outer layers of younger tissues with high cell proliferation activity. To confirm this, we explored the cell proliferation pattern around the shoot apex using histone H4 expression as a marker of cell division (Fig. 4G). Histone H4-expressing cells were detected in P1, P2 and P3 leaf primordia. Such a cell division pattern was comparable with that associated with MKB3 expression (Fig. 4C). However, cell proliferation was also observed in regions not expressing MKB3, such as the adaxial epidermis of the P3 leaf primordium.

To further assess the function of MKB3, we analyzed MKB3-overexpressing plants by introducing a construct harboring MKB3 cDNA fused to the rice ACTIN promoter. Three independent T₁ lines harboring the transgene were obtained (Fig. 5 and Fig. S6), and T₂ plants derived from one of the T₁ line were used (Fig. 5). These plants had normal morphology, but produced longer and wider leaves than control plants, associated with high-level MKB3 expression (MKB3ox) (Fig. 5A,B,L). The length and width of the fifth leaf blade of MKB3ox plants were 36% and 16%, respectively, greater than those of control plants (Fig. 5D,E). However, cross-sections of MKB3ox and control leaves revealed that the thickness of the fifth leaf blade (Fig. 5C,F), and the size of epidermal cells along the proximal-distal, adaxial-abaxial and medial-lateral axes, were not significantly different (Table 2). Therefore, it was calculated that the larger leaf size of MKB3ox was due to the increased number of cells, and overexpression of MKB3 did not affect cell size. In contrast, the number of small vascular bundles and the intervals between the vascular bundles of MKB3ox plants were increased by 17% and 8%, respectively, although the number of large vascular bundles was not altered (Fig. 5A,B; Table 2).

During reproductive development, MKB3ox plants produce longer panicles with longer primary rachis branches compared with control plants, but the number of primary rachis branches was not altered (Fig. 5G-I). MKB3ox plants were fertile and set seeds with significantly larger dimensions along all axes (Fig. 5J,K), as also reported elsewhere (Duan et al., 2015; Li et al., 2016; He et al., 2017). This indicates that overexpression of MKB3 promotes growth of internodes and seeds, but does not exert harmful effects on tissue differentiation of reproductive organs, in contrast to loss of function of MKB3.

Phenotypic analysis of MKB3-overexpressing plants suggested that MKB3 is a positive regulator of cell proliferation. In addition, MKB3 is not involved in cell size control, because the cell size of MKB3ox plants was not changed. Therefore, cell enlargement in the mkb3 mutant was a secondary effect of reduced cell proliferation, rather than the direct cause of MKB3 dysfunction.

AN3, the Arabidopsis ortholog of MKB3, positively regulates cell proliferation in leaves. Although AN3 mRNA does not accumulate in epidermal cells, AN3 is required not only by mesophyll cells but also by epidermal cells (Kawade et al., 2013). This is achieved by movement of the AN3 protein from mesophyll cells to epidermal cells. Inter-cell-layer movement of the AN3 protein is essential for coordination of proliferation in mesophyll and epidermal cells, and for regulation of leaf size (Kawade et al., 2013).

MKB3 was strongly expressed in the epidermis of young leaf blades, but weakly or not expressed in the abaxial side of the inner tissues of the P3 leaf sheath, vascular bundles and the SAM (Fig. 4B-D). Thus, the expression patterns of rice MKB3 and Arabidopsis AN3 differ markedly.

To determine whether protein movement and its functional importance are conserved between rice and Arabidopsis, we performed transgenic analysis using a strategy similar to that of Kawade et al. (2013). We prepared three transgenes: GFP as a nonfunctional control, a chimeric gene comprising MKB3 cDNA fused with GFP (MKB3-GFP) and MKB3 cDNA fused with three copies of GFP (MKB3-3×GFP) as an MKB3 protein mobility control. It is known that the AN3-3×GFP protein product is functional but unable to move between cells in Arabidopsis leaves (Kawade et al., 2010, 2013). These constructs were introduced into wild-type and mkb3 calli under the control of the native MKB3 promoter, and five independent T₁ transgenic plants regenerated from wild-type and mkb3 calli of each line were observed (Fig. 6A).

In the wild-type background, pMKB3::GFP (WT/pMKB3::GFP) and WT/pMKB3::MKB3-3×GFP transgenic plants showed a wild-type phenotype, but WT/pMKB3::MKB3-GFP could not be obtained for an unknown reason (Fig. 6B). In the mkb3 background, pMKB3::MKB3-GFP transgenic plants (mkb3/pMKB3::MKB3-GFP) showed a phenotype almost identical to the wild-type control, indicating that pMKB3::MKB3-GFP is functional and can complement mkb3 defects (Fig. 6B). However, mkb3/pMKB3::MKB3-3×GFP plants produced narrowed and shortened leaves, similar to the mkb3/pMKB3::GFP control (Fig. 6B). This suggests that introduction of pMKB3::MKB3-3×GFP cannot rescue the mkb3 phenotype and that MKB3 movement is required for normal leaf development. To explore whether the chimeric MKB3-3×GFP protein was functional, as is true of AN3-3×GFP, we introduced the AN3-3×GFP and MKB3-3×GFP genes under the control of the AN3 promoter into Arabidopsis an3-4 (Fig. 3F,G,I,K and Fig. S5). We found that the extent of phenotypic rescue of an3-4 leaves by pAN3::MKB3-3×GFP was similar to that afforded by pAN3::AN3-3×GFP; thus, both AN3-3×GFP and MKB3-3×GFP mediated partial recovery of leaf size, cell number and cell size (Fig. S5). This strengthened our suggestion that MKB3-3×GFP was equivalent to AN3-3×GFP in terms of both function and mobility.

Next, we evaluated the protein accumulation pattern in the shoot apex of transgenic plants using GFP to fluorescently detect (Fig. 6C-E) and immunolocalize GFP-tagged proteins (Fig. 6F-H). GFP signals in WT/pMKB3::GFP and mkb3/pMKB3::GFP plants were observed mainly in young leaf primordia, but the tissue specificity was low (Fig. S7), possibly because of GFP protein movement. GFP protein is known to move between cells through plasmodesmata (Kim et al., 2005). GFP signals in pMKB3::MKB3-3×GFP plants, which produced an immobile version of MKB3, were predominantly observed in the epidermis and marginal region of leaves, but not in the inner tissue of leaves or the SAM, in both the wild-type and mkb3 backgrounds (Fig. 6C-E,G). This protein accumulation pattern is generally consistent with that of MKB3 mRNA determined by in situ hybridization (Fig. 4B-D), although MKB3-3×GFP protein accumulation in the epidermis of young leaf primordia and leaf blades was more evident than MKB3 mRNA (Figs 4C,D, 6F,G). In contrast, the GFP accumulation pattern in mkb3/pMKB3::MKB3-GFP was markedly different from that in WT/pMKB3::MKB3-3×GFP and mkb3/pMKB3::MKB3-3×GFP. GFP signals were detected throughout leaf primordia; i.e. not only in the epidermis but also in the inner tissue, where MKB3 mRNA was not detected (Fig. 6E,H). In addition, MKB3-GFP protein was detected in the basal part of the SAM, where MKB3 mRNA was not observed. These results suggest that MKB3-GFP protein moves from an MKB3-expressing domain – the epidermis of the leaf – to a non-expressing domain: the inner tissues of the leaf primordia, vascular bundles and the SAM.

Phylogenetic analysis indicates that MKB3 is an ortholog of Arabidopsis AN3/GIF1. In addition, complementation of an3 phenotype by MKB3 indicates that MKB3 protein is functionally equivalent to AN3/GIF1. AN3/GIF1 is not only a central regulator of leaf cell proliferation but also involved in biologically interesting phenomena, such as compensation triggered by mutation of an3 and inter-cell-layer communication by movement of AN3 (Horiguchi et al., 2005; Kawade et al., 2013, 2017). Thus, mkb3 could facilitate comparative studies of two evolutionarily diverged species: rice and Arabidopsis. Our results indicate that MKB3 and AN3 have conserved functions in most aspects, but they are regulated differently in the two species. The similarities and differences between MKB3 and AN3 are discussed below.

Although mkb3 was identified as a mutant with abnormal leaf morphology, our phenotypic analysis revealed that the mkb3 mutant exhibits abnormalities not only in leaf morphology and size but also in internode and rachis elongation, spikelet and floral morphology, and fertility, indicating that MKB3 pleiotropically affects plant development during the life cycle in rice. In leaf development, the length and width of the mutant leaf blades were significantly reduced despite the increased size of the cells. This indicates that MKB3 positively regulates cell proliferation in leaves. This effect of MKB3 on cell proliferation was supported by the phenotype of MKB3-overexpressing plants, which showed increased leaf size with unchanged cell size. In reproductive development, mkb3 internodes, panicles and rachis were shorter than those of the wild type. Although the cellular responses of these organs and tissues to mkb3 mutation remain undetermined, the phenotypes possibly suggest that MKB3 also enhances cell proliferation in stem-like organs during the reproductive phase. This is consistent with the expression of MKB3 in the basal parts of those organs. In contrast, the number of primary rachis branches was not affected in both mkb3 mutants and MKB3-overexpressing plants. Accordingly, MKB3 is not involved in branch meristem activity. The reduction of the width of lemma and palea in the mkb3 spikelets suggests that the growth of these leaf-like organs in the spikelet is under the control of MKB3. In addition, the outer integument was incompletely elongated in the mkb3 pistil, likely also due to a defect in cell proliferation.

Taken together, the role of MKB3 and AN3 as positive regulators of cell proliferation in leaves is conserved between rice and Arabidopsis. In contrast, abnormalities in stem elongation and floral development were observed in mkb3 but not in gif mutants, suggesting that MKB3 is required for the development of these organs in rice. However, a gif1 gif2 gif3 triple mutant of Arabidopsis showed reduced stem elongation and abnormal reproductive organ development (Lee et al., 2009). Regarding integument development, a single an3 mutant showed a shortened outer integument, similar to that in mkb3 (Lee et al., 2014). Accordingly, although functional redundancy among the paralogs of Arabidopsis GIF genes may mask the defects in some plant parts in a gif1/an3 mutant background, the function of AN3 and MKB3 in plant development is largely conserved between rice and Arabidopsis. Recently, a gif1 mutant in maize was reported (Zhang et al., 2018). One of the conspicuous phenotypes was loss of determinacy in axillary meristems, which was observed in neither Arabidopsis an3 nor rice mkb3. Accordingly, the indeterminacy in meristems may be a specific phenomenon of gif1 mutant in maize.

Cell size in mkb3 leaves was increased along the three axes, whereas that in leaves of MKB3-overexpressing plants was not changed. This indicated that enlargement of cells in mkb3 leaves is not a direct effect of mkb3, but an indirect effect of decreased cell proliferation or cell number. This is thought to be a typical example of compensation, which is observed in many Arabidopsis mutants and transgenic plants, such as an3, fugu1-5, erecta, and KRP2-overexpressing plants (Horiguchi and Tsukaya, 2011; Hisanaga et al., 2015). Although compensation in rice was also reported in OsKRP1-overexpressing plants (Barrôco et al., 2006), no compensation triggered by a defect in orthologous genes in different species has been reported. In our study, mkb3, the counterpart of Arabidopsis an3, showed clear compensation, indicating that compensation and its underlying mechanisms are conserved in eudicots and monocots. In addition, the ploidy level in mkb3 leaves was not altered, although endoreduplication does not normally occur in rice leaves. Thus, cell enlargement in mkb3 is not caused by ectopic activation of the endoreduplication pathway.

The vascular bundle arrangement in mkb3 is possibly a compensation-related phenotype. mkb3 leaf blades have significantly fewer vascular bundles, although the intervals between vascular bundles were slightly increased. The increased intervals of vascular bundles in mkb3 may be due to the enlargement of cells between vascular bundles by compensation. This explanation is supported by molecular evidence that the position of vascular bundles is determined at the early stage of leaf development before cell expansion starts (Nishimura et al., 2002; Itoh et al., 2005). Thus, compensation affects not only cell and leaf size but also, indirectly, the arrangement of vascular bundles.

The most conspicuous difference between MKB3 and AN3 is their expression patterns. MKB3 is expressed in younger leaf primordia with active cell proliferation. However, the spatial expression pattern in P2 and P3 primordia is unique in that MKB3 is predominantly expressed in the epidermal cells of the leaf blade, but not in the mesophyll cells of the adaxial side of the leaf sheath. In contrast, AN3 mRNA accumulates in mesophyll cells, but not in epidermal cells (Kawade et al., 2013).

In Arabidopsis, AN3 protein moves from mesophyll cells to epidermal cells, which facilitates coordination of cell proliferation activity between the different cell layers (Kawade et al., 2013). In rice, MKB3-GFP protein was present not only in epidermal cells but also in inner mesophyll cells where MKB3 mRNA was not detected, indicating that MKB3 moves from the epidermal cells to the mesophyll cells of leaf primordia. In addition, introduction of immobile MKB3-3×GFP did not rescue the phenotype of mkb3. Thus, MKB3 protein movement is essential for a normal cell proliferation pattern in rice leaf, similar to AN3 in Arabidopsis.

It is interesting that the direction of movement differs between MKB3 and AN3; i.e. MKB3 moves from epidermal cells to mesophyll cells of leaf blades, and AN3 from mesophyll cells to epidermal cells. In addition, MKB3 moves from the abaxial region to the adaxial regions of leaf sheath primordia. A possible explanation for the difference is that the direction of protein movement is associated with the difference in cell proliferation pattern between rice and Arabidopsis. During early development of rice leaf, leaf primordia grow in a conical shape to form two structurally different parts: the leaf blades and leaf sheaths (Itoh et al., 2005). Leaf blades are flattened structures, whereas leaf sheaths are crescent-shaped in cross-section. To form a leaf sheath, a decreasing gradient of cell proliferation from the outer (abaxial) region to the inner (adaxial) region of the leaf primordia is essential. Therefore, high cell proliferation activity is more important in the abaxial domain than in the adaxial domain of the leaf sheath. This is also true for organs with a cylindrical or columnar structure, such as internodes and rachis. The MKB3 expression pattern and MKB3 protein movement may be involved in generating the cell proliferation gradient. That is, MKB3 proteins produced on the abaxial side of the leaf sheath primordia move toward the adaxial side and are diluted, creating a gradient of MKB3 protein amount in the adaxial-abaxial direction. In contrast, coordination in the adaxial-abaxial direction is not necessary in Arabidopsis leaves and rice leaf blades, because leaf proliferation activity is similar between the adaxial and abaxial sides of leaf primordia. This notion is supported by the expression pattern of PLA1, which is also involved in leaf cell proliferation (Miyoshi et al., 2004; Mimura and Itoh, 2014). Although both rice PLA1 and its Arabidopsis ortholog KLU regulate leaf size by controlling cell proliferation, their expression patterns are different. Rice PLA1 is expressed mainly on the abaxial side of younger leaf sheaths, similar to MKB3 (Miyoshi et al., 2004). In contrast, Arabidopsis KLU does not show a polarized expression pattern in the adaxial-abaxial direction in leaf (Zondlo and Irish, 1999). These polarized expressions of MKB3 and PLA1 suggest a requirement for a mechanism of coordinating the cell proliferation gradient during normal leaf development in rice. MKB3 protein movement might be involved in formation of specialized structures such as leaf sheaths, although the complete reversal of the direction of protein movement between the leaf lamina of Arabidopsis and the leaf blades of rice remains intriguing.

In summary, rice MKB3 positively regulates cell proliferation in leaves, and mutant leaves display clear compensation, indicating a conserved function and effect on leaf development of MKB3 and Arabidopsis AN3. In addition, protein movement of MKB3 is essential for normal development of rice leaves. However, the direction of MKB3 movement is different from that of AN3, suggesting that developmental diversity is mediated by expression pattern and/or protein movement, but not by protein function.

mkb3 was identified as a recessive mutant showing adaxially rolled leaves, and was derived from an M₂ population of rice (Oryza sativa L.) cv. Taichung-65 (T-65) mutagenized with N-methyl-N-nitrosourea. Mutant and wild-type plants were grown in pots or in a paddy field. Transgenic plants were grown in a biohazard-secure greenhouse at 30°C during the day and 25°C at night.

A heterozygous MKB3/mkb3 plant was crossed with cv. Kasalath (spp. indica), and mutant plants in the F₂ population that exhibited the rolled phenotype were used for mapping. With the aid of cleaved amplified polymorphic sequences and sequence-tagged site markers, the MKB3 locus was roughly mapped onto the short arm of chromosome 3. Using 224 mutant plants of the F₂ generation, the MKB3 locus was limited to a region covering two bacterial artificial chromosome (BAC) contigs (OSJNBa0079G12 and OSJNBa0027J18). As we found a homolog of Arabidopsis AN3 in this region, we compared the nucleotide sequences of the gene between the wild-type and mkb3 genes.

The MKB3 cDNA nucleotide sequence and MKB3 amino acid sequence were obtained from GenBank (AK058575 and BR001474). Multiple sequence alignments were performed with the aid of ClustalX. A phylogenetic tree was constructed employing the neighbor-joining method. A Marchantia polymorpha sequence served as an outgroup.

Tissues were fixed in 4% (w/v) paraformaldehyde in 0.1 M sodium phosphate buffer, dehydrated in a graded ethanol series, infused with HistoClear (National Diagnostics) and embedded in Paraplast Plus (McCormick Scientific). Microtome sections (8 µm) were placed on glass slides (Matsunami Glass). For histological observations, sections were stained with Delafield's Hematoxylin and observed under a light microscope. For in situ hybridization, digoxigenin-labeled antisense and sense RNA probes for AKB3 and histone H4 were prepared. Because the sense probes did not yield specific signals, only antisense probe data are presented here. In situ hybridization and immunological detection using alkaline phosphatase were performed using the methods of Kouchi and Hata (Kouchi and Hata, 1993). For immunolocalization of GFP protein, shoot apices of three T₂ transgenic lines: WT/pMKB3::GFP, WT/pMKB3::MKB3-3×GFP and mkb3/pMKB3::MKB3-GFP were sampled 2 weeks after regeneration. The methods were basically those of Smith et al. (1992) with minor modifications. Tissues were fixed, dehydrated and embedded as described above. Dewaxed, rehydrated paraffin sections were treated for 10 min with proteinase K (100 μg/ml in PBS). Slides were incubated in PBS with 1 mg/ml BSA for 30 min, and then with alkaline phosphatase-conjugated anti-GFP antibody (abcam_ab6661) diluted 1:1000 in PBS with 1 mg/ml BSA for 2 h. The remaining steps were identical to those used for in situ hybridization.

To measure leaf cell sizes in the central-marginal and adaxial-abaxial directions, cross-sections of the fifth leaf blades of wild-type and pACT::MKB3 plants were prepared. The epidermal cell widths were calculated by dividing the width of the region on the abaxial side of the epidermis in which cells were enumerated by the number of cells. Bulliform cell thickness was measured directly on cross-sections. To measure leaf cell sizes in the distal and proximal directions, the cells of the adaxial surface of the leaf sheath of the fifth leaf were directly observed by fluorescence microscopy. To measure the numbers of vascular bundles and the intervals between them, cross-sections of the fifth leaf blade of wild-type and pACT::MKB3 plants were evaluated.

For intraspecific complementation testing, an 8251 bp genomic fragment of MKB3, including the 3308 bp putative promoter and 1898 bp terminator regions was cloned into the pPZP2H-lac binary vector (Fuse et al., 2001). The cloned vector and empty control vector were transformed into mkb3-homozygous calli via Agrobacterium-mediated transformation (Hiei et al., 1994). For interspecific complementation testing, we used the pAN3::AN3-GFP and pAN3::AN3-3xGFP transgenic plants described elsewhere (Kawade et al., 2010). pAN3::MKB3-GFP and pAN3::MKB3-3xGFP transgenic plants were established using binary vectors, R4 pGWB504-pAN3::OsAN3 and R4 pGWB501-pAtAN3:OsAN3-3xGFP, respectively, employing the floral dip method (Clough and Bent, 1998). The vectors were constructed via LR reactions using the Multisite gateway system (Life Technologies) to yield pENTR/D-TOPO-MKB3 (containing MKB3 cDNA without a stop codon), pDONR P4-P1R-pAN3 (containing about 2.0 kb upstream of the AN3 gene) (Kawade et al., 2013), and binary vectors containing GFP and 3×GFP (Nakagawa et al., 2007). To generate MKB3-overexpressing plants, MKB3 cDNA was inserted into a binary vector containing the rice ACTIN promoter and the NOS terminator (pACT::MKB3) (Kamiya et al., 2003). For phenotypic analysis of pACT::MKB3, we used plants harboring pACT::MKB3 that segregated from the T₂ generation. High-level expression of MKB3 was confirmed by RT-PCR using RNA samples from leaf tissues of T₂ plants. To evaluate protein movement, pMKB3::GFP, pMKB3::MKB3-GFP and pMKB3::MKB3-3×GFP constructs were prepared. In pMKB3::GFP, the GFP-encoding sequence was inserted into the MKB3 fragment within the vector used for complementation testing, and the MKB3-encoding sequence was removed. Similarly, a chimeric gene composed of MKB3 cDNA fused with GFP (MKB3-GFP) and MKB3 cDNA fused with three copies of GFP (MKB3-3×GFP) were introduced between the MKB3 promoter and the terminator region of the vector. The three constructs were transformed into wild-type and mkb3 homozygous calli via Agrobacterium-mediated transformation (Hiei et al., 1994). All T₁ transgenic plants were evaluated 1 month after regeneration; GFP images around the shoot apices were obtained using a fluorescent microscope (Eclipse-Ti, Nikon) equipped with a confocal laser-scanning system (C1-Si, Nikon). A 488 nm diode laser was used to excite GFP. Emission signals were detected with the aid of a 515/30 nm filter.

We thank R. Soga and H. Teshima (Institute for Sustainable Agro-ecosystem Services, The University of Tokyo) for their assistance in cultivating rice plants. We also thank Tomoyuki Takano and Ayaka Kinoshita of the University of Tokyo, who helped with vector constructions.