**4. Discussion**

Cereal crops accumulate high levels of starch in the seed endosperm as an energy reserve. Endosperm-defective mutants provide valuable genetic material for elucidating the gene networks that regulate starch synthesis and amyloplast development during grain filling. In this study, we isolated and characterized the Namsil(SA)-flo1 mutant, which exhibited an almost entirely milky-white opaque kernel, except for a thin peripheral area of the grain (Figure 1a). Compared to the wild type, the Namsil(SA)-flo1 mutant exhibited slightly higher amylose levels, lower protein levels, reduced kernel weight, and increased total lipid content (Table 2). No difference in seed size was observed (Figure 1c). Electron microscope visualization of the transverse sections of the Namsil(SA)-flo1 and wild-type mature endosperms showed that the Namsil(SA)-flo1 mutant was loosely packed with irregular and round-shaped compound starch granules, while the wild type contained densely packed polyhedral starch granules (Figure 1b). However, the physicochemical characteristics of Namsil(SA)-flo1 were different from those reported for *flo4-1*, *flo4-2,* and *flo4-3*, which were generated via T-DNA insertion into the cy*OsPPDK* gene [15,33,34]. We speculate that the cy*OsPPDK* gene may have diverse roles in the rice seed filling stage. Fine mapping, sequencing, and co-segregation analysis revealed that a C to T SNP in exon 2 of cy*OsPPDK* (the novel recessive floury gene was named *flo4-5*) was responsible for the Namsil(SA)-flo1 mutant phenotype (Figures 2 and 3). Based on our limited expression analysis method, this is not clearly explained; however, we hypothesize that the transcript from the mutated sequence compared to that of the normal sequence may not produce normal PPDK to increase the transcription level in the mutant, in order to compensate the normal transcript concentration in the cell. This hypothesis will be proven in a further study. Along with an increase in the transcripts of PPDK/AGPase, the expression of changes for starch synthesis genes (Figure 5) suggests that PPDK has a role in compound starch granule formation and starch synthesis during rice endosperm development.

PPDK catalyzes the freely reversible conversion of pyruvate, ATP, and Pi into phosphoenolpyruvate, AMP, and PPi, and is involved in diverse functions in various plant tissues. PPDK is ubiquitous in all rice tissues, but is generally of low abundance [43]. In developing rice kernels, PPDK is abundantly expressed during the syncytial/cellularization stage, with markedly reduced expression seen during the middle and late rice grain filling stages [15,27]. The presence of PPDK in high abundance in developing seeds suggests the importance of the enzyme; however, little is known regarding the molecular properties of rice PPDK.

The interconversion of phosphoenolpyruvate and pyruvate by PPDK leads to difficulties in evaluating PPDK function in seed development [43]. Previous studies have suggested several hypotheses for endosperm PPDK function, including (1) provision of pyruvate for lipid synthesis, (2) gluconeogenesis and provision of hexose for starch biosynthesis, (3) control of metabolic fluxes through its contribution to PPi homoeostasis, and (4) regulating glycolytic flux and energy charge [24,27,28,30, 44,45]. In *Zea mays,* the PPi generated by PPDK can be channeled directly to AGPase within the protein complex, driving the plastidial AGPase reaction in the direction of ADPGlc breakdown to Glc-1-P, which can in turn support amino acid and lipid biosynthesis [24]. However, a loss-of-function mutation in the rice cy*OsPPDK* gene resulted in a lower amylose content, reduced kernel weight, increased total lipid content, and no change in storage protein or amino acid contents [15]. This suggests that PPDK may have several diverse roles in seed development, depending, in part, on the direction of the cycle [43].

In amylopectin biosynthesis, *SS* connects glucose in an α-1,4 glycosidic bond formation, and then *BE* (branching enzyme) connects a short chain to the C-6 hydroxyl group via hydrolysis of internal α-1,4 glycosidic bonds and formation of α-1,6 branches. *DBE* (starch-debranching enzyme: Isoamylase (*ISA*) and pullulanase (*PUL*)) cleaves aberrant branches through catalytic hydrolysis of α-1,6 glycosidic bonds. A lack of *SS* during rice endosperm amylopectin formation results in a decrease in polymerized chain formation and an increase in amylose characteristics [12,46–49]. Mutant *ae*, which is defective in the gene encoding *BE*, exhibits high proportional levels of amylose as a result of reductions in amylopectin production [50,51]. A loss-of-function mutation of the gene encoding *DBE* results in a sugary phenotype in which endosperm starch is completely replaced by phytoglycogen [52,53]. Mutant *sug-h*, which is defective in *OsISA1* and *OsBEIIa*, exhibits a severely disrupted amylopectin structure [54].

A recent study revealed that rice amylopectin-synthesizing enzymes such as *SSI, SSIIIa, BEI, BEIIa, BEIIb, ISA1,* and *PUL* are physically associated with each other and form active protein complexes with various partnership patterns [12,23]. Further research in maize and rice indicates that PPDK and plastid AGPase may also participate in these enzyme complexes [12,24]. In these complexes, PPDK catalyzes the freely reversible conversion of pyruvate, ATP, and Pi into phosphoenolpyruvate, AMP, and PPi, but in photosynthesis, PPDK generates phosphoenolpyruvate by the hydrolysis of PPi used for CO2 fixation in C4 plants [24]. Plastid AGPase also catalyzes a reversible reaction, but the direction of the reaction only depends on the relative concentrations of PPi and ATP [30]. Therefore, high PPi concentrations should orient the reaction to the direction of ADP-Glc degradation. However, the PPi levels in amyloplasts are very low as a result of high pyrophosphatase (PPase) activity, and the reaction therefore proceeds toward ADP-Glc synthesis. Zhou et al. [12] hypothesized that the starch biosynthetic enzymes in the complex might exert a constraining e ffect on PPDK and AGPase to control the partitioning of ADP-Glc into lipid and starch.

Based on previous research and our findings, we hypothesize that a single amino acid change (S to F) in the PPDK protein in the flo4-5 mutant may disrupt the amylopectin-synthesizing multienzyme complexes and release PPDK to allow an abundance of PPi in amyloplasts. The high concentration of PPi would then promote plastid AGPS2a and AGPS2b activity and direct the reaction toward ADP-Glc degradation, producing more Glc-1-P for lipid biosynthesis. At the same time, PPDK would actively increase transcription of the *SSIIb* and *SSIIc* genes without increasing starch content. This supports the view that the catalytic activity of one or more of the starch synthase isoforms is insu fficient to accommodate the excess ADP-glc present in the amyloplast, but that other constraints within the amyloplast stroma control carbon flux into starch [55]. Further research is required to test this hypothesis.

The number of spikelets per panicle substantially increased (by approximately 32.5%) in the Suweon 542 mutant (*flo4-4*) compared with the wild type [33]. In the flo4-5 mutant, the number of spikelets per panicle (Table 1) increased only slightly. This may have been due to environmental effects on yield, whereby the average high temperature during the first 4 days after pollination must reach ~33 ◦C for yield increase to occur [56]. AGPase, the rate-limiting starch biosynthetic enzyme, is a heterotetramer composed of two identical small and two identical large subunits. The enzyme is allosterically controlled, and the isoform in the cereal endosperm is heat-labile [57]. In maize, the *Shrunken*-2 (*Sh*2) gene encodes the large subunit of endosperm AGPase, and overexpression of the gene enhances heat stability and reduces phosphate inhibition, leading to an increase in maize yield (up to 64%) via an increase in seed number [56]. The small subunit of the maize endosperm AGPase is encoded by the gene *Brittle*-2 (*Bt*2), and overexpression of this enzyme also increases yield by about 35%, albeit by an increase in seed number rather than seed weight. This increase is dependent on the average daily high temperature reaching ~33 ◦C during the first 4 days after pollination [57]. Furthermore, expression of an altered large subunit of endosperm AGPase increases potato tuber yield by 35% [58], wheat yield by 38% [59], and rice yield by 23% [60]. Notably, the yield increases in maize, rice, and wheat are due to increases in seed number rather than increases in individual seed weight. In rice, increases in seed number are a consequence of an enhanced probability of seed development rather than an increase in the number of ovaries. In wild-type rice, only about half of the ovaries develop into fully mature kernels. The remaining kernels cease development, and their contents disintegrate [56,57]. Overexpression of the small subunit of the maize endosperm AGPase (ZmBt1) in rice enhances ADP-Glc synthesis and imports into amyloplasts, but does not lead to further enhancement in seed weight, even under elevated CO2 [55]. These observations sugges<sup>t</sup> that the PPDK mutant might enhance the expression of *AGPS2a* and *AGPS2b*, as with AGPase transgene expression, leading to increases in the expression of *AGPS2a*, *AGPS2b*, *SSIIb*, and *SSIIc,* and an increase in seed

number rather than seed weight. Our results sugges<sup>t</sup> that enhanced synthesis of ADP-Glc in the rice seed endosperm may increase seed number. Future research regarding the metabolic mechanisms underlying the increase in seed number will be useful for breeding programs aimed at improving rice yield.
