Variations of endoreduplication and its potential contribution to endosperm development in rice (Oryza sativa L.)

Endoreduplication is the phenomenon by which cells increase their ploidy. Endoreduplication is initiated by the transition from the mitotic cell cycle to the endocycle, in which DNA replication occurs without a subsequent chromosome separation and cytokinesis, and is enhanced by endocycle reiteration. This process appears to play an important role in endosperm development, but the characteristics of endoreduplication in the endosperm of rice (Oryza sativa) remain unclear. To elucidate the features and variations of endoreduplication in rice endosperm, endoreduplication progression in the developing endosperm was compared among 10 cultivars based on flow cytometry and fluorescence microscopy. The flow cytometric analysis detected significant differences among 10 cultivars in the following three parameters: mean ploidy of all nuclei, the proportion of nuclei ≥6C (%E, an estimate of the initiation of the endocycle), and the mean ploidy of nuclei ≥6C (E6P, an estimate of the reiteration of the endocycle). However, no significant correlation between %E and E6P was observed, suggesting that the initiation and reiteration of the endocycle are independently regulated. Fluorescence microscopy revealed that the ploidy of the nuclei was higher in the intermediate region than in the central and peripheral regions of the endosperm. Cells with a higher ploidy were larger in the developing endosperm. Furthermore, the mean ploidy in the developing endosperm was significantly correlated with the mean cell size in the mature endosperm. These results indicate that endoreduplication progression in the endosperm differed significantly among the 10 rice cultivars and such differences may influence endosperm cell size. Abbreviations: Ak: Akitakomachi; DAP: days after pollination; DAPI: 4′,6-diamidino-2-phenylindole; E6P: mean ploidy of nuclei ≥ 6C; Ha: Habataki; Ho: Hokuriku193; IR: IR64; Ka: Kasalath; Ki: Kinuhikari; Ko: Koshihikari; Ni: Nipponbare; PEG: polyethylene glycol; Sa: Sasanishiki; Ta: Takanari; %E: proportion of nuclei ≥ 6C. ARTICLE HISTORY Received 15 October 2018 Revised 25 December 2018 Accepted 8 January 2019


Introduction
Endoreduplication, which is also called endoreplication, is the phenomenon by which cells increase their ploidy (Breuer et al., 2010;Dante et al., 2014;De Veylder et al., 2011;Sugimoto-Shirasu & Roberts, 2003). During endoreduplication, the chromatids are duplicated exponentially, while the number of chromosomes remains unchanged. Endoreduplication is initiated by the transition from the mitotic cell cycle to a modified cell cycle called the endocycle, in which DNA replication occurs without a subsequent chromosome separation and cytokinesis. Ploidy increases are due to the reiteration of the endocycle.
Although the exact role of endoreduplication remains unknown, it is believed to contribute to cell enlargement, increase gene expression levels, accelerate growth, and increase DNA storage (Barow, 2006;Chevalier et al., 2011;Kowles, 2009;Nguyen et al., 2007;Sabelli & Larkins, 2009;Sugimoto-Shirasu & Roberts, 2003). The prevailing hypothesis concerning the role of endoreduplication is that it influences cell enlargement (Chevalier et al., 2014;Orr-Weaver, 2015;Sugimoto-Shirasu & Roberts, 2003). The transition from cell proliferation to cell expansion is often accompanied by the transition from the mitotic cycle to the endocycle. Furthermore, many studies have observed a strong correlation between ploidy and cell size. However, a causal relationship between endoreduplication and cell size has been disputed because some mutants do not exhibit a concurrent decrease in endoreduplication and cell size (Leiva-Neto et al., 2004;Vilhar et al., 2002). Considered together, these findings indicate that endoreduplication may determine the maximum capacity for cell growth instead of strictly regulating cell growth (Breuer et al., 2010).
Endoreduplication occurs in many cell types, especially large, metabolically active, or highly specialized cells (De Veylder et al., 2011). Examples include endosperm cells in Poaceae species (Sabelli, 2012;Sabelli & Larkins, 2009). A primary endosperm cell is formed with a 3C nucleus (1C represents the ploidy of a nonreplicated haploid genome) because of the fusion of two polar nuclei in the embryo sac and one sperm nucleus of the pollen grain, but the ploidy of endosperm cells gradually increases during endosperm development via endoreduplication (Sabelli & Larkins, 2009). In Poaceae species, endoreduplication follows several endosperm development phases, such as syncytium formation, cellularization, and mitosis (Sabelli, 2012;Sabelli & Larkins, 2009). Inhibited endoreduplication reportedly impairs endosperm development (Barrôco et al., 2006). Additionally, endoreduplication may influence the yield and/or quality of cereal grains. For example, polyploidy frequency as well as the number of cells per endosperm are correlated with seed weight in wheat (Brunori et al., 1993). The mean ploidy of maize endosperm cells tends to be higher in popcorn than in dent corn (Dilkes et al., 2002). Brachypodium distachyon endosperm, in which endoreduplication is less extensive than that in barley, is composed of small cells and contains abundant (1,3;1,4)-β-glucan (Trafford et al., 2013). Therefore, clarifying the features and contribution of endoreduplication in the endosperm of Poaceae species may lead to improvements in cereal grain yield and/or quality.
The intensive investigation of endoreduplication progression in maize endosperm cells has clarified the associated spatiotemporal patterns (Nguyen et al., 2007;Sabelli, 2012;Sabelli & Larkins, 2009). At approximately 8 days after pollination (DAP), maize endosperm cells asynchronously switch from the mitotic cell cycle to the endocycle. The endoreduplication of cells in the central region precedes that of cells in the peripheral region. Consequently, there is a ploidy gradient, with a higher ploidy in the inner region of the maize endosperm. A similar gradient is observed for maize endosperm cell size, and there is a strong correlation between the ploidy and size of endosperm cells (Kowles, 2009). However, endoreduplication progression differs among maize cultivars, and this difference is derived from both the initiation and reiteration of the endocycle (Dilkes et al., 2002).
In contrast to maize, studies of endoreduplication in the developing endosperm of rice are limited. Feulgen microspectrophotometry has been used to reveal the temporal progression of endoreduplication in the rice endosperm (Kono et al., 1979;Ramachandran & Raghavan, 1989). The mean DNA content per nucleus remains low until 4 days after anthesis, and rapidly increases from 4 to 8 days after anthesis (Ramachandran & Raghavan, 1989). Endoreduplication progression in the endosperm varies among the spikelets developing at different panicle positions. The endosperm cells from apical spikelets have more endoreduplicating nuclei than those from basal spikelets (Panda et al., 2015(Panda et al., , 2018. The cell cycle regulatory genes, KRP1 and CCS52A, are important for the transition from the mitotic cell cycle to the endocycle, and changes to the expression of these genes hinder endosperm development (Barrôco et al., 2006;Su'udi et al., 2012). Endoreduplication in rice endosperm cells is regulated by phytohormones such as ethylene and cytokinin. The application of 1methylcyclopropene (an ethylene action inhibitor) and 6benzylaminopurine (a cytokinin) increases the ploidy of endosperm cells in the dense-panicle rice cultivar 'OR-1918' (Panda et al., 2016(Panda et al., , 2018. However, to the best of my knowledge, endoreduplication progression has not been compared among rice cultivars. Information regarding the diversity of endoreduplication in rice endosperm among cultivars may be useful for developing methods that apply endoreduplication in breeding programs to improve crops. Additionally, there is a lack of 'ploidy maps' revealing the distribution of rice endosperm cells in different ploidy classes. Ploidy maps have been generated for maize, sorghum, and teosinte (Dermastia et al., 2009;Kladnik et al., 2006;Vilhar et al., 2002), and may provide meaningful insights into how the endocycle mechanisms are coordinated with the growth of individual cells in a specific tissue (De Veylder et al., 2011).
The objective of this study was to clarify the features and variations of endoreduplication in rice endosperm. To examine endoreduplication in detail, the three parameters mean ploidy, proportion of nuclei ≥6C (%E), and mean ploidy of nuclei ≥6C (E6P) were evaluated as described by Dilkes et al. (2002) on the basis of flow cytometric data. The mean ploidy of all nuclei is frequently used as an index of endoreduplication, while the parameter %E is used to estimate initiation of the endocycle. For example, a high %E, which indicates a low proportion of 3C nuclei, suggests that mitotic cell division has decreased and that the recruitment of nuclei to the endocycle has increased. The parameter E6P is used to estimate reiteration of the endocycle. For example, a high E6P implies that reiteration of the endocycle has been promoted. The relationship among these three parameters is as follows: mean ploidy = 3 × (1 − %E/100) + E6P × %E/100. To clarify the spatial progression of endoreduplication in rice endosperm, ploidy maps were generated by fluorescence microscopy. In the present study, five japonica and five indica cultivars were used to evaluate the variation in endoreduplication progression among rice cultivars. The cultivars used are popular in Japan, and chromosome segment substitution lines with some pairs of these cultivars have recently been developed (Ando et al., 2008;Takai et al., 2014). Furthermore, the relationship between endoreduplication and cell size was investigated to clarify whether endoreduplication contributes to the control of cell size in rice endosperm.

Plant materials for examining the spatiotemporal progression of endoreduplication
In 2014, Oryza sativa cv. 'Kinuhikari' (a japonica cultivar) plants were grown under natural conditions in a paddy field at the Western Region Agricultural Research Center, National Agriculture and Food Research Organization (Fukuyama city, Hiroshima, Japan, 34°29′ N, 133°23′ E). Caryopses were germinated in a nursery bed on 9 May, and seedlings were transplanted to the paddy field at a hill spacing of 15 cm × 30 cm, with two seedlings per hill, on 29 May. Chemical fertilizer (N:P 2 O 5 :K 2 O = 14%:14%:14%) was applied as a basal dressing at a rate of 28.5 g m −2 , and ammonium sulfate (N = 21%) was applied at a rate of 9.5 g m −2 at 20 days before heading. Subsequent experiments involved the superior caryopses located at the fourth and fifth nodes from the apex of the upper two primary branches on the panicle. The pollination dates of these caryopses were recorded, and the caryopses were harvested from five individual plants at each sampling time point. For flow cytometry, two caryopses per plant were dehulled with forceps, immediately placed in a microcentrifuge tube, and stored at −80°C. Because rice leaves reportedly contain only 2C nuclei (Barow & Meister, 2003;Martínez et al., 1994), flag leaves were sampled at heading and stored at −80°C to use as a reference material for flow cytometry. For fluorescence microscopy, caryopses at 5 and 9 DAP were fixed in Farmer's solution (95% ethanol:glacial acetic acid, 3:1 [v/v]) containing 1 mM MgCl 2 for 1 h, and then stored in 70% ethanol containing 1 mM MgCl 2 at −4°C.
To determine dry weight, 10 caryopses were dried at 60°C for 72 h.
Plant materials for comparing endoreduplication among 10 rice cultivars The three rows of individual cultivars were replicated four times and arranged in a randomized complete block design. The distance between rows was 30 cm, and that between hills was 15 cm. Two seedlings were transplanted per hill. Chemical fertilizer (N:P 2 O 5 :K 2 O = 14%:14%:14%) was applied as a basal dressing at a rate of 28.5 g m −2 , and ammonium sulfate (N = 21%) was applied at a rate of 9.5 g m −2 at 1 month after transplantation. Pollination dates were recorded for the superior caryopses located at the fourth, fifth, and sixth nodes from the apex of the upper two primary branches on the panicle. The caryopses were periodically harvested from several plants in the central row and were preserved as described above. To examine endoreduplication of the endosperm in each cultivar at a similar developmental stage, the caryopses harvested when the dry weight was about 25% of that at maturity were used in subsequent experiments (Supplement 1).

Flow cytometry
Samples were prepared for flow cytometric analysis based on a modified version of the procedures described by Galbraith et al. (1983). Using a disposable scalpel, samples were chopped for 2.5 min in 1.2 mL chopping buffer. The homogenate was filtered through a 70-µm nylon mesh filter, and nuclei were stained with 30 µM propidium iodide. The stained nuclei were analyzed by flow cytometry with a Guava® easyCyte 6-2 L benchtop flow cytometer (Merck Millipore, Billerica, MA, USA). Ten thousand events were counted for the caryopses, and 5,000 events were counted for flag leaves.

Fluorescence microscopy
Samples were prepared for fluorescence microscope analysis using a modified version of the procedures described by Kobayashi et al. (2013). Fixed caryopses were embedded with polyethylene glycol (PEG), and median transverse sections (12 µm thick) were sliced with a microtome. After removing PEG from the sections with phosphate buffer, the sections were mounted in VECTASHIELD® Mounting Medium with DAPI (4′,6-diamidino-2-phenylindole) (Vector Laboratories, Inc., Burlingame, CA, USA). The sections were observed with the IX71 fluorescence microscope (Olympus, Tokyo, Japan) equipped with a WU filter set (excitation wavelength 330-385 nm, emission wavelength 420 nm). Images were captured with a VB-7000 CCD camera (Keyence, Osaka, Japan). To obtain high-resolution images, several endosperm portions were photographed under high magnification (20×) and combined.
To analyze the ploidy in endosperm sections, images were examined based on the modified procedures for trichome nuclei (Kobayashi et al., 2012;Maes et al., 2008). Images captured with a CCD camera were converted to grayscale using Photoshop Elements 12 (Adobe Systems, San Jose, CA, USA), and analyzed using ImageJ 1.44p (National Institutes of Health, USA; http://imagej.nih.gov/ ij/). The area of nuclei that had distinct contours was manually circumscribed, and their integrated density (the product of area and mean gray value) was measured as an index of fluorescence intensity. The integrated density value was log transformed, and histograms of log-integrated density were drawn to evaluate the ploidy of each nucleus. Transverse sectional areas of 30 cells in the respective ploidy classes were then measured using ImageJ 1.44p. In some cultivars, the section contained fewer than 30 cells with 24C nuclei. In such instances, the data for all cells with 24C nuclei were included.

Measurement of cell size in mature caryopses
In 2015, mature caryopses were fixed and embedded with PEG as described above. Median and longitudinal transverse sections were sliced with a microtome. After removing PEG from the sections with water, sections were stained with 0.05% toluidine blue solution (Wako Pure Chemical Industries, Osaka, Japan) for 5 min. After rinsing with water for 30 s, images of the sections were captured with a YDU-2 digital stereomicroscope (Yashima Optical Co., Tokyo, Japan). To obtain highresolution images, several endosperm portions were photographed under high magnification (250×) and combined.
To analyze the cell size in median transverse sections, images were examined using modified procedures described by Morita et al. (2005). Images of the endosperm with an aleurone layer were prepared by tracing cell contours on a Cintiq 13HD graphic tablet (Wacom Co., Saitama, Japan). The area of each cell was analyzed using ImageJ 1.44p, and the mean cell area was determined. To calculate mean cell height, the length of 50-70 cell layers in longitudinal sections was measured. Mean cell volume was calculated by multiplying the mean transverse sectional area of cells by the mean cell height.

Statistical analysis
Statistical analysis was carried out using JMP 11 software (SAS Institute Japan Inc., Tokyo, Japan). Data were analyzed by one-or two-way analysis of variance (ANOVA), and then means were compared using a Tukey-Kramer multiple comparison test (P < 0.05).

Spatiotemporal endoreduplication progression in rice endosperm
The temporal progress of endoreduplication in the endosperm of rice 'Kinuhikari', a japonica cultivar, was examined using flow cytometry. Flow cytometric analysis detected four or five peaks in fluorescence intensity of the nuclei in developing caryopses (Figure 1(a-c)), whereas a single peak was detected in flag leaves (Figure 1(d)). The peak with the lowest fluorescence intensity in the caryopses coincided with the peak in the flag leaves. Rice leaves reportedly contain only 2C nuclei (Barow & Meister, 2003;Martínez et al., 1994); therefore, the peak with the lowest fluorescence intensity in the caryopses was judged to be 2C, which is derived from the embryo and the pericarp. Using this 2C peak as a reference, the other peaks in the caryopses were judged to be 3C, 6C, 12C, and 24C, which were derived from the endosperm (Figure 1(a-c)).
An analysis of the flow cytometry data enabled the determination of the composition of the nuclei in different ploidy classes in rice endosperm ( Figure 2). In this study, the dry weight of a developing caryopsis relative to that of the mature caryopsis was used as a reference for caryopsis development (Figure 2(a)). The 12C nuclei, which were exclusively formed via endoreduplication, were detected as early as 5 DAP (Figure 2(b)), when the relative weight of the caryopsis was about 5% (Figure 2(a)). The proportion of nuclei with an elevated ploidy increased until 8 DAP ( Figure  2(b)), when the relative weight of the caryopsis was about 20% (Figure 2(a)). Thereafter, the proportions of nuclei with an increased ploidy was relatively unchanged until 10 DAP (Figure 2(b)), when the relative weight of the caryopsis was about 30% (Figure 2(a)). On the basis of the composition of nuclei in different ploidy classes, the mean ploidy of all nuclei in the endosperm, which is frequently used as an index of endoreduplication, was calculated. The mean ploidy was 4.8C at 5 DAP, and increased significantly to 6.9C at 8 DAP (Figure 2(c)). From 8 to 10 DAP, the mean ploidy did not change significantly. In the present study, endoreduplication in rice endosperm older than 10 DAP could not be measured because the flow cytometer was frequently clogged with developed starch granules.
To clarify the spatial progression of endoreduplication in rice endosperm, median transverse sections of the caryopses stained with DAPI were examined by fluorescence microscopy (Figure 3). Only small nuclei were observed in the endosperm at 5 DAP (Figure 3(a)), whereas many large nuclei were observed at 9 DAP (Figure 3(c)). Image analysis of DAPI fluorescence from endosperm nuclei revealed several peaks in the histograms of integrated density (Figure 3(e,f)). Flow cytometric analysis clarified that the lowest ploidy of rice endosperm nuclei at these stages was 3C (Figure 1). Therefore, the peak with the lowest integrated density was judged to be 3C, and the other peaks were determined as 6C, 12C, and 24C (Figure 3(e,f)). Based on the histograms, the ploidy of each nucleus in the section was determined and superimposed on the images to construct ploidy maps (Figure 3(b,d)). At 5 DAP, most of the nuclei in rice endosperm cells were 3C and 6C. However, some nuclei in the intermediate region between the central and peripheral regions were 12C (Figure 3(b)). At 9 DAP, the ploidy varied greatly among the nuclei in the different endosperm regions. The ploidy was higher in the intermediate region than in the peripheral and central regions (Figure 3(d)).

Comparison of endoreduplication among 10 rice cultivars
The temporal progression of endoreduplication based on the relative dry weight of developing caryopses in the indica cultivar 'Kasalath' (Supplement 2) exhibited a similar pattern to that in the japonica cultivar 'Kinuhikari' (Figure 2). Therefore, endoreduplication progression in the endosperm of 10 rice cultivars was compared at the developmental stage when the dry weight of the developing caryopsis was about 25% of that at maturity (Supplement 1). Significant differences in the mean ploidy, %E, and E6P were observed among the 10 cultivars in 2015 (Figure 4). Similar results were obtained in 2014 (Supplement 3), and highly significant correlations between years were observed ( Figure 5). The effects of years on the mean ploidy and E6P were significant, while the interaction of cultivars and years was not significant for each parameter (Supplement 4). The mean ploidy differed among cultivars, with the lowest and highest mean ploidies observed for Ak and Ta, respectively (Figures 4(a) and 5(a)). The following two groups of cultivars were detected based on %E: Ak, Sa, Ko, Ki, and Ka formed a low %E group, whereas Ho, Ha, Ni, IR, and Ta belonged to a high %E group (Figures 4(b) and 5(b)). Differences in E6P were observed among cultivars; E6P was lowest in Ak and highest in Ka and Ta (Figures 4(c) and 5(c)). No significant correlation between %E and E6P was observed in either year ( Figure 6).
To compare the spatial progression of endoreduplication in the developing endosperm among cultivars, ploidy maps were constructed at the stage when the dry weight of the developing caryopsis was about 25% of that at maturity (Figures 7 and 8). A similar distribution pattern of nuclei with different ploidies was observed for each cultivar. The ploidy was higher in the intermediate region than in the peripheral and central regions of median transverse sections of the caryopsis. However, the endosperm of cultivars with a higher mean ploidy or E6P, as indicated by flow cytometry, contained more 24C nuclei.
Similar to the ploidy distribution, cells in the intermediate region were larger than those in the central and peripheral regions in all examined cultivars ( Figures  7 and 8). To assess the effect of endoreduplication on rice endosperm development, the transverse sectional area of endosperm cells in the respective ploidy classes was measured. Cells with a higher ploidy were significantly larger in the developing endosperm (Table 1, Figure 9 see the vertical axis). Given that a significant interaction between ploidies and cultivars was detected (Table 1), the transverse sectional area of endosperm cells within the same ploidy classes was compared among the cultivars (Figure 9, Supplement 5). Significant differences in cell area among the cultivars were observed within the 3C and 6C classes (Figure 9, Supplement 5). The area of cells with 3C nuclei was negatively correlated with %E and E6P, while the area of cells with 6C nuclei was negatively correlated with E6P ( Figure 10, Supplement 6). Within the 12C or 24C classes, no significant differences in the transverse sectional area of endosperm cells were observed among the examined cultivars (Figure 9, Supplement 5).  To evaluate the effects of endoreduplication in the developing endosperm on the mature caryopsis, the dry weight and cell size of the mature caryopsis in 2015 were examined (Table 2, Supplements 7 and 8). One-way ANOVA detected significant differences in caryopsis weight, the mean transverse sectional area, and the mean volume of endosperm cells among the 10 rice cultivars (Table 2). Although no significant correlation between the mean    ploidy in the developing endosperm and the weight of the mature caryopsis was observed (Figure 11(a)), a significant correlation was detected between the mean ploidy in the developing endosperm and the mean transverse sectional area or volume of endosperm cells in the mature caryopsis (Figure 11(b,c)).

Spatiotemporal progression of endoreduplication in rice endosperm
Endoreduplication was detected as early as 5 DAP in rice endosperm (Figures 1(a), 2(b) and 3(b)), although the mean ploidy of all nuclei was low at 5 DAP (Figure 2 (c)). The mean ploidy of endosperm cells increased until 8 DAP, and thereafter remained relatively constant (Figure 2(c)). This temporal progression is consistent with the results for rice endosperm examined using Feulgen microspectrophotometry (Ramachandran & Raghavan, 1989). In rice endosperm, mitotic cell division  begins at 4 DAP and ceases at 8 or 9 DAP (Hoshikawa, 1967). These observations indicate that endoreduplication overlaps temporally with mitosis in rice endosperm. A similar overlap of mitosis and endoreduplication occurs in maize, in which endosperm cells gradually and asynchronously switch from the mitotic cycle to the endocycle (Sabelli & Larkins, 2009). Thus, cells in the mitotic cycle and in the endocycle coexist in the developing endosperm of Poaceae species, and the degree of endoreduplication differs among nuclei. Ploidy maps clearly revealed the distribution of nuclei with varying degrees of endoreduplication in the rice endosperm (Figure 3). The ploidy was higher in the intermediate region than in the central and peripheral regions of the rice endosperm (Figure 3 (d)), with similar spatial patterns observed in all 10 analyzed rice cultivars (Figures 7 and 8). Interestingly, this spatial progression of endoreduplication differed from that seen in previous reports of maize, sorghum, and teosinte; the ploidy is reportedly higher in the inner region of the endosperm (Dermastia et al., 2009;Kladnik et al., 2006;Vilhar et al., 2002). In maize endosperm, the ploidy gradient is due to the spatiotemporal pattern of the mitosis/endoreduplication switch. This switch from the mitotic cycle to the endocycle starts with cells in the central region and spreads outward (Nguyen et al., 2007;Sabelli & Larkins, 2009). Cells in the central region of rice endosperm are also formed earlier than those in the outer region (Hoshikawa, 1967). Therefore, cells in the central region of the rice endosperm should cease mitosis earlier than those in the outer region. However, 12C nuclei were not detected in cells in the central region of the rice endosperm at 9 DAP, when the nuclei in the intermediate region were ≥12C (Figure 3(d)). Furthermore, the cells in the central region of rice endosperm reportedly commence the process of programmed cell death around 10 DAP (Kobayashi et al., 2013). These results suggest that nuclei in the central region retain a low ploidy throughout rice endosperm development.

Comparison of endoreduplication among 10 rice cultivars
Endoreduplication progression in endosperm cells varied among the 10 examined rice cultivars (Figure 4, Supplement 3). For example, the mean ploidy, which is a typical parameter representing endoreduplication progression in whole endosperm cells, differed significantly among the cultivars, with the highest and lowest mean ploidies observed in Ta and Ak, respectively. The rank order of mean ploidy among the cultivars was similar in 2014 and 2015 (Figure 4, Supplement 3), and highly significant correlations between years were observed ( Figure  5). These results imply that endoreduplication progression is genetically regulated in rice endosperm.
To examine endoreduplication progression in rice endosperm in detail, %E and E6P were calculated. The parameter %E is applied to estimate the initiation of the endocycle, while E6P is used to estimate the reiteration of the endocycle. As reported for maize (Dilkes et al., 2002), no significant correlation between %E and E6P was observed in rice endosperm ( Figure 6). For example, on the basis of %E, Ka belonged to the lowest group, whereas on the basis of E6P, it belonged to the highest group (Figure 4(b,c)). Additionally, ploidy maps revealed that Ka had more 24C nuclei than Ni, Ha, and IR, which were similar to Ka regarding mean ploidy (Figures 7 and 8). These results suggest that the initiation and reiteration of the endocycle are independently regulated in rice endosperm cells. This independence probably enables cells to cease mitosis early and to remain within a low ploidy class (≤6C), as observed in the cells located in the central region of rice endosperm (Figures 3, 7 and 8). Previous studies characterized some regulatory mechanisms underlying the initiation of the endocycle in rice endosperm. For example, the CDK inhibitor Orysa;KRP1 is important for the exit from the mitotic cell cycle (Barrôco et al., 2006). Meanwhile, OsCCS52A, which is an activator of the anaphase-promoting complex that initiates cyclin degradation, also mediates the exit from the mitotic cell cycle (Su'udi et al., 2012). The application of 1methylcyclopropene increased the frequency of nuclei ≥6C by promoting the production of CDKAs and CYCD2:2, while inhibiting the production of CYCB2:2 (Panda et al., 2016). In contrast to the emerging understanding of endocycle initiation, reiteration of the endocycle is less well understood in rice endosperm. In Arabidopsis thaliana, some mechanisms for terminating the reiteration of the endocycle have been identified (Breuer et al., 2010). For example, a transcriptional regulator, GT-2-LIKE1 (GTL1), is produced only during the post-branching stages of leaf trichomes, and GTL1 loss-offunction mutations lead to an additional round of the endocycle. Future rice endosperm studies should focus on the reiteration of the endocycle.
In the developing rice endosperm, proportional relationships between ploidy and cell size were observed. Cells with a higher ploidy had a larger transverse sectional area (Table 1, Figure 9 see the vertical axis). Similar relationships have been reported in the developing endosperm of maize and sorghum (Kladnik et al., 2006;Vilhar et al., 2002). Furthermore, Panda et al. (2018) reported that a 6-benzylaminopurine treatment simultaneously increases the ploidy and size of rice endosperm cells. Accordingly, endoreduplication appears to affect rice endosperm cell size. In addition to the variation in cell size reflective of ploidy variation, significant differences in cell size were observed among cultivars within the same ploidy classes, such as 3C and 6C (Figure 9(a,b)). The transverse sectional area of endosperm cells with 3C nuclei was negatively correlated with %E and E6P, and that of cells with 6C nuclei was negatively correlated with E6P ( Figure 10). Endosperm cells in the cultivars with low %E or E6P probably remained within the same ploidy class for a relatively long period. Therefore, a prolonged duration within the same ploidy class might promote cell expansion. Variations in cell size associated with ploidy were greater than those associated with cultivars ( Figure 9  (a,b)). For example, the Ak cells with 3C nuclei, which were the largest cells with 3C nuclei among the 10 rice cultivars studied (648 µm 2 ), were smaller in area than the Ka cells with 6C nuclei, which were the smallest cells with 6C nuclei (1017 µm 2 ). These results suggest that endoreduplication considerably influences rice endosperm cell size. Furthermore, the effects of endoreduplication on cell size in the developing endosperm appeared to continue in the mature endosperm, because the mean ploidy in the developing endosperm was significantly correlated with the mean mature endosperm cell size (Figure 11(b,c)).
Although endoreduplication appeared to play an important role in determining rice endosperm cell size, there was no significant correlation between the mean ploidy in the developing endosperm and the weight of the mature caryopsis (Figure 11(a)). Because organ size is determined not only by cell size, but also by cell number (Orr-Weaver, 2015), the variation in cell size associated with endoreduplication is probably cancelled by the diversity in cell number in the examined rice cultivars. However, the genetic variation related to the endoreduplication of endosperm cells might be useful for improving caryopsis weight. For example, new rice genotypes with heavy caryopses may be developed by crossing genotypes with high ploidies and cell numbers per endosperm, as proposed for wheat (Brunori et al., 1993).
The enlargement of endosperm cells associated with endoreduplication may affect the quality of rice as food. The surface area of a sphere (4πr 2 ) generally does not keep pace with increases to the volume (4/3πr 3 ). Therefore, for endosperms with an identical volume, the endosperm composed of large cells will contain less cell wall material than the endosperm containing small cells. Endosperm cell walls are considered to affect the texture of cooked rice (Shibuya & Iwasaki, 1984). These results imply that endoreduplication in endosperm cells affects the texture of cooked rice via changes in the amount of cell wall material. Thus, rice quality may be improved by modifying the endoreduplication in endosperm cells.
In conclusion, this study clarified the characteristics of endoreduplication in the developing rice endosperm and their variations among rice cultivars. Rice endosperm exhibited a typical spatial progression of endoreduplication, with the ploidy of nuclei higher in the intermediate region than in the central and peripheral regions of the endosperm. Endoreduplication progression differed significantly among 10 rice cultivars. This diversity reflects the variation in the initiation and reiteration of the endocycle, which are two processes that are independently regulated in rice endosperm. There were proportional relationships between ploidy and cell size in the developing rice endosperm. Furthermore, the mean ploidy in the developing endosperm was significantly correlated with mature endosperm cell size. These results suggest that endoreduplication influences rice endosperm cell size. Therefore, the genetic variation mediating the endoreduplication of endosperm cells may be applicable for improving rice yield and/or quality via changes in cell size.