**1. Introduction**

Rice (*Oryza sativa* L.) is a staple food for more than half the world's population; improving yield and grain quality is therefore of grea<sup>t</sup> importance. Grain yield can be increased considerably through the use of hybrid rice strains, but the grain quality of these varieties does not fully satisfy customer demands [1]. As such, grain quality improvement is the current focus of many rice geneticists worldwide. Furthermore, cultivars with a range of di fferent grain qualities are desirable for specific production or medicinal purposes.

Starch, which is the major storage carbohydrate in rice grains, constitutes approximately 90% of a rice grain, of which approximately 18% is amylose and 82% is amylopectin. The ratio of amylose to amylopectin plays an important role in rice grain structure, appearance, and eating quality. Amylose content is the most important determining factor for almost all of the physicochemical properties of rice starch, including its gelling, pasting, firmness, turbidity, freeze–thaw stability, syneresis, and retrogradation properties [2,3]. Amylopectin content governs the formation of crystalline granules and paste thickening [4]. Much research in recent decades has focused on understanding the genetics and biochemistry of starch biosynthesis, which is a complex process through which glucose generated by photosynthesis is converted to starch via several intermediaries. AGPase catalyzes the first committed step of starch biosynthesis and, as such, plays an important role in regulating starch synthesis in rice endosperms. Rice grains from plants carrying mutations in the AGPase gene exhibit shrunken endosperms and reduced starch content as a result of significant reductions in starch synthesis [5,6]. By contrast, the overexpression of AGPase genes increases seed weight and starch content [7]. Both amylose and amylopectin utilize ADPG as the activated glucosyl donor, and subsequent synthesis steps use different enzymes for amylose and amylopectin production. Amylose is synthesized by *GBSSI* (granule-bound starch synthase), which is encoded by *Wx* (waxy). Loss-of-function mutations at the *GBSSI*/*Wx* locus result in the elimination of amylose synthesis, leading to a lack of long-chain amylopectin, as well as a complete lack of amylose [8,9].

Starch synthesis is also transcriptionally regulated by transcription factors that presumably function within biosynthetic networks [1]. These include alkaline leucine zipper transcription factor OsbZIP58, which regulates *OsAGPL3*, *Wx*, *OsSSIIa*, *SBE1*, *OsBEIIb*, and *ISA2* transcription through binding to their promoters [10]. A lack of *ISA* in the rice sugary mutant impacts the expression of several other genes related to starch synthesis [11]. The high-resistant starch (RS) mutant, which is defective in SSIIIa, affects the activity of PPDK (pyruvate phosphate dikinase) and AGPase and leads to increases in lipid and amylose synthesis [12]. The white-core endosperm mutant, which is defective in *OsPK2*, affects the expression of genes involved in glycolysis/gluconeogenesis, pyruvate/phosphoenolpyruvate metabolism, fatty acid metabolism, and starch synthesis [13].

Among endosperm mutants, only those with a floury and white core contain round and loosely packed starch granules that grind easily [14]. To date, seven floury endosperm genes (*FLO1–7*) have been identified in rice [15–21], six of which have been cloned and their functions characterized. Three of these genes are not directly involved in starch synthesis and are instead involved in protein–protein interactions that regulate starch-synthesis-related genes. *Flo2* encodes a tetratricopeptide repeat motif that affects *Wx* expression, Flo6 encodes a C-terminal carbohydrate-binding module 48 domain that regulates isoamylase1, and Flo7 harbors an N-terminal transit peptide that does not directly interact with starch synthesis genes [19–21]. The three remaining floury endosperm genes participate in starch synthesis pathways; Flo4 encodes a PPDK, Flo5 encodes a rice-soluble starch synthase gene (*OsSSIIIa*), and *osagpl2–3* (also named *Flo6*) encodes AGPase [15,18,22]. In rice starch metabolism pathways, amylopectin-synthesizing enzymes, including starch synthases (*SSIIIa*, *SSIVb*, and *SSIIa*), starch-branching enzymes (*SBEI* and *SBEIIb*), and *PUL*, physically interact with each other to form multienzyme complexes [23,24]. The multienzyme complexes also contain other enzymes, such as PPDK and AGPase, that control the partitioning of ADP-Glc into starch and lipid [12,24,25]. This suggests that the proteins encoded by the three floury endosperm starch synthesis genes could interact with one another. The PPDK encoded by *Flo4* could promote AGPase activity in the complex by directly supplying PPi (pyrophosphate) for the conversion of ADP-Glc to Glc-1-P, which would enhance lipid biosynthesis. *SSIIIa* could inhibit the activity of PPDK/AGPase [25], and this interaction could be disrupted in the *SSIIIa* mutant (G to A in an exon), releasing PPDK and AGPase and increasing lipid synthesis. Simultaneously, loss of *SSIIIa* activity could divert the carbon flow needed for amylose synthesis by *GBSSI* (encoded by *Wx*), and as a result, the balance of lipid and amylose, which are components of RS, could be disrupted [12,25]. For AGPase, an SNP in the open reading frame (ORF) region of *osagpl2–3* may impair the activity of the functional domain that interacts with ADP-Glc. The effect of this mutation is particularly apparent at the endosperm development stage and results in the mutant floury endosperm phenotype [22]. Nevertheless, direct evidence of PPDK's function in the multienzyme complex remains unknown.

Rice has three PPDK-encoding genes organized at two loci: PPDKA, which encodes a cytosolic isoform (*OsPPDKA*), and PPDKB, which has two promoter sites and produces cytosolic and chloroplastic enzymes (cyPPDKB and chPPDKB, respectively) [8,15,26]. Of these, cyPPDKB is critical during the filling stage of rice grain development. During this stage, cyPPDKB is produced in abundance and acts to provide carbon skeletons for amino acid and lipid synthesis through the reversible interconversion of pyruvate and Pi to phosphoenolpyruvate and PPi. As seed development progresses from the early stage to the final storage product accumulation stage, the level and activity of cyPPDKB decrease rapidly in response to the combined posttranslational mechanisms of threonyl phosphorylation and protein degradation [27]. Cytosolic PPDKB is also involved in the glycolytic and gluconeogenic pathways and is impacted by anoxia [28–31]. To determine PPDK function in developing seeds, experiments to transgenically eliminate the PPDK gene from rice endosperms, to release PPDK from the phosphate-su fficient condition, or to upregulate PPDK expression during the middle and late rice grain-filling stages would be logical. However, the first report of a T-DNA insertional knockout mutant of the rice PPDK gene (*flo4-1*, *flo4-2* and *flo4-3*) indicated that rice with inactivated PPDK produce unexpected opaque seeds with a high lipid content [15].

The Namil(SA)-flo1 rice mutant was developed via sodium azide mutagenesis of *Oryza sativa* ssp. japonica cv. Namil [32]. By screening the mutant stock of Namil, two allelic mutants exhibiting floury endosperms were isolated and named "Namil(SA)-flo1" and "Namil(SA)-flo2" (Suweon 542). Suweon 542 exhibited a milky-white opaque endosperm, except for a thin peripheral area. Physicochemical analysis of the Suweon 542 endosperm revealed a loosely packed structure of irregular and globular-shaped starch granules, a low protein content and grain weight, and high amylose content compared with its wild-type progenitor, Namil. During dry milling, Suweon 542 had significantly lower grain hardness, finer particle size, more loosely packed starch granules, and lower starch damage than Namil and other rice cultivars [33]. Genetic analysis of the floury endosperm characteristics of Suweon 542 revealed that the location of the target gene was in the 19.33–19.73 Mbp region on chromosome 5 between markers RM18624 and RM18639, and map-based cloning revealed a G →A SNP in exon 8 of cy*OsPPDK* (*flo4-4*), which was responsible for the floury endosperm of Suweon 542 [33,34]. Moreover, the floury endosperm of Namil(SA)-flo1 controlled by one recessive gene and the locus of Namil(SA)-flo1 was localized to the 17.7–20.7 Mbp region of chromosome 5 [35]. However, the grain characteristics and molecular mechanisms regulating the floury endosperm in Namil(SA)-flo1 remain unknown. Fine mapping and molecular cloning of floury endosperm-related genes are needed to verify and elucidate the genetic framework of floury endosperm development in rice.

In this study, the agronomic traits and grain physicochemical properties, including suitability for dry milling of Namil(SA)-flo1, were investigated. Map-based cloning revealed that a C → T SNP in exon 2 of cy*OsPPDK* results in a missense mutation from Ser to Phe at amino acid position 101. Co-segregation analysis and qRT-polymerase chain reaction (PCR) confirmed that *flo4-5* was responsible for the development of the floury endosperm during the grain filling stage and indicated the involvement of cy*OsPPDK* in grain quality and seed number control.
