Rice LIKE EARLY STARVATION1 cooperates with FLOURY ENDOSPERM6 to modulate starch biosynthesis and endosperm development

Abstract

In cereal grains, starch is synthesized by the concerted actions of multiple enzymes on the surface of starch granules within the amyloplast. However, little is known about how starch-synthesizing enzymes access starch granules, especially for amylopectin biosynthesis. Here, we show that the rice (Oryza sativa) floury endosperm9 (flo9) mutant is defective in amylopectin biosynthesis, leading to grains exhibiting a floury endosperm with a hollow core. Molecular cloning revealed that FLO9 encodes a plant-specific protein homologous to Arabidopsis (Arabidopsis thaliana) LIKE EARLY STARVATION1 (LESV). Unlike Arabidopsis LESV, which is involved in starch metabolism in leaves, OsLESV is required for starch granule initiation in the endosperm. OsLESV can directly bind to starch by its C-terminal tryptophan (Trp)-rich region. Cellular and biochemical evidence suggests that OsLESV interacts with the starch-binding protein FLO6, and loss-of-function mutations of either gene impair ISOAMYLASE1 (ISA1) targeting to starch granules. Genetically, OsLESV acts synergistically with FLO6 to regulate starch biosynthesis and endosperm development. Together, our results identify OsLESV-FLO6 as a non-enzymatic molecular module responsible for ISA1 localization on starch granules, and present a target gene for use in biotechnology to control starch content and composition in rice endosperm.

The non-enzymatic protein LIKE EARLY STARVATION1 and the starch-binding protein FLOURY ENDOSPERM6 form a functional complex to facilitate ISAMYLASE1 targeting to starch granules in rice endosperm.

IN A NUTSHELL.

Background: Starch accounts for up to 75% of grain weight, and greatly affects cereal crop yield and quality. Significant advances have been made in the functional characterization of starch metabolism enzymes in cereal crops. In addition to starch metabolism enzymes, some non-enzymatic players with starch-binding domains (SBDs) play crucial roles in starch biosynthesis. ISOAMYLASE1 (ISA1), a type of debranching enzyme unable to bind to starch, functions in starch biosynthesis by removing superfluous branch points. However, how ISA1 is targeted to starch granules remains largely unknown.

Question: Comparable phenotypic defects between floury endosperm9 (flo9) and isa1 prompted us to propose that LIKE EARLY STARVATION1 (OsLESV), the target protein responsible for the flo9 phenotype, might regulate rice (Oryza sativa) starch biosynthesis by associating with ISA1. Therefore, we asked whether OsLESV is a non-enzymatic player that facilitates ISA1 targeting to starch.

Findings: We determined that the rice flo9 mutant is defective in starch biosynthesis and endosperm development, similar to the reported isa1 mutants. FLO9 encodes an amyloplast-localized protein homologous to Arabidopsis (Arabidopsis thaliana) LESV. OsLESV is required for compound starch granule initiation in the endosperm. OsLESV directly binds to starch by its C-terminal Trp-rich region. We demonstrated that OsLESV interacts with the reported non-enzymatic player FLO6, and loss-of-function of either gene impairs ISA1 targeting to starch granules. Genetically, OsLESV acts synergistically with FLO6 to modulate starch biosynthesis and endosperm development. Our findings establish a molecular link between non-enzymatic players and ISA1 in rice starch biosynthesis and endosperm development.

Next steps: In addition to facilitating ISA1 localization on starch granules, OsLESV likely exerts other functions in starch biosynthesis and endosperm development. Therefore, it is important to identify other unknown cargos to fully understand the biological significance of OsLESV in regulating starch metabolism in future studies.

Introduction

Starch accounts for up to 75% of grain weight and therefore greatly affects cereal crop yield and quality. Starch is composed of 2 types of glucose polymers, amylose and amylopectin, with the latter constituting 75% to 80% of starch weight (Le Corre et al. 2010; Bao 2019). Amylose is a linear polymer comprising D-glucose molecules connected via α-1,4-linkages, whereas amylopectin is a highly branched, huge polymer formed by regular α-1,6-glycosidic linkages of linear polymers (Jenkins et al. 1993; Thompson 2000).

Starch biosynthesis in the endosperm of cereals occurs in amyloplasts, in which amylose and amylopectin together form insoluble, semi-crystalline particles termed starch grains (SGs). SGs are either compound or single, depending on the species (Matsushima et al. 2010; Myers et al. 2011; Seung et al. 2018; Abt and Zeeman 2020). Amylopectin branches that are arranged in clusters can form stable double helices whose packing leads to the formation of crystalline layers of starch granules, whereas amylose appears to be interspersed among amylopectin clusters within starch granules (Zeeman et al. 2010; Abt and Zeeman 2020; Smith and Zeeman 2020). Although starch in cereal endosperm provides a major calorie source for human food and animal feed, our understanding of the molecular machinery that controls starch biosynthesis in cereal crops remains obscure.

The biosynthesis of amylopectin in starch granules requires the orchestrated activities of at least 3 distinct classes of enzymes (Ball et al. 1996; Nakamura 2002; James et al. 2003; Fujita et al. 2006; Hanashiro et al. 2008; Abt and Zeeman 2020). Starch synthases (SSs) elongate linear glucan chains using ADP-glucose as the glucose donor. Starch branching enzymes (BEs) are responsible for introducing α-1,6 branch points, whereas the 2 types of starch debranching enzymes, isoamylase (ISA) and pullulanase (PUL), are required to remove superfluous or improper branch points to produce crystallization-competent glucans (Kubo et al. 1999). By contrast, the biosynthesis of amylose within the amylopectin matrix only involves granule-bound starch synthase (GBSS). Upon entering the amyloplast or chloroplast, several starch-synthesizing enzymes such as rice (Oryza sativa) starch branching enzyme I (OsBEI), Arabidopsis (Arabidopsis thaliana) starch synthesis III (AtSSIII), and Arabidopsis branching enzyme I (AtBEI) directly associate with starch granules via their own starch-binding domains (Dumez et al. 2006; Valdez et al. 2008; Chaen et al. 2012). Amylopectin-synthesizing enzymes, especially SSs and BEs, form large enzymatically active protein complexes in the amyloplast stroma, thereby facilitating their attachment to starch granules (Liu et al. 2012; Crofts et al. 2015). Notably, ISAs are absent from these complexes (Crofts et al. 2015), and in vitro studies showed that ISA1 essential for amylopectin biosynthesis in rice endosperm cannot bind to starch directly (Peng et al. 2014). Therefore, the molecular mechanism by which ISA1 targets to starch granules remains unclear.

In addition to the starch metabolic enzymes themselves, accumulating evidence suggests that some non-enzymatic players with starch-binding domains, such as carbohydrate-binding module 48 (CBM48), play crucial roles in starch metabolism (Abt and Zeeman 2020; Smith and Zeeman 2020). Rice CBM48 domain-containing FLOURY ENDOSPERM6 (FLO6) is genetically characterized as a non-enzymatic player, which is proposed to function as a starch-binding protein in starch biosynthesis and compound granule formation through a direct interaction with ISA1 (Peng et al. 2014). By contrast, the Arabidopsis FLO6 homolog, PROTEIN TARGETING TO STARCH2 (PTST2) is involved in starch granule initiation in leaves by supplying STARCH SYNTHASE4 (SS4) with preselected malto-oligosaccharides (Seung et al. 2017). Arabidopsis PTST1, which is structurally similar to FLO6, interacts with and targets GBSS to the amorphous regions of starch granules; its loss-of-function substantially influences amylose production in leaves (Seung et al. 2015). Unexpectedly, PTST1 depletion in rice has negligible effects on amylose content in endosperm (Wang et al. 2020). The functional differences in these non-enzymatic proteins suggest that different species and even tissues likely employ distinct sets of non-enzymatic players in orchestrating highly complex starch synthesis. Despite these significant advances, our understanding of the biological roles of these non-enzymatic proteins in starch synthesis remains very limited, especially in cereal crops.

In this study, as part of our continuous effort to understand the molecular machinery responsible for starch biosynthesis in rice endosperm, we isolated and functionally characterized the rice floury endosperm9 (flo9) mutant. The flo9 mutant is defective in amylopectin biosynthesis and partially phenocopies the rice isa1 mutants, which accumulate an assortment of water-soluble-glucans (WSGs) instead of starch in the endosperm (Nakamura et al. 1997; Fujita et al. 2003; Nakagami et al. 2017). FLO9 encodes an amyloplast-localized protein homologous to Arabidopsis LESV, which is involved in starch metabolism in Arabidopsis leaves (Feike et al. 2016). We presented biochemical and confocal microscopy evidence supporting the notion that OsLESV physically interacts with ISA1 and that its loss-of-function significantly perturbs the targeting of ISA1 to starch granules. Furthermore, we establish a mechanistic linkage between OsLESV and FLO6, and propose that they form a functional protein complex to cooperatively modulate starch biosynthesis and endosperm development. Taken together, our findings identify OsLESV as a non-enzymatic regulator that positively modulates amylopectin synthesis in rice endosperm.

Results

The flo9 mutant exhibits a defect in amylopectin accumulation in endosperm

To better understand the molecular mechanisms underlying starch biosynthesis in rice, we conducted a forward genetic screen to identify mutants defective in starch accumulation from an N-methyl-N-nitrosourea-mutagenized mutant library of the elite japonica rice variety W017. One such mutant, flo9, produced floury and shrunken endosperm with dimpled indentations and constrictions at the dorsal and ventral sides, respectively, in sharp contrast to the translucent, plump endosperm of the wild type (WT) (Fig. 1A). Microscopy analysis of transverse sections showed that flo9 endosperm is composed of a translucent periphery, a floury-white intermediate region filled with aberrantly loose SGs, and a hollow core (Fig. 1, B and C). These endosperm defects in flo9 led to a 16.8% loss in 1,000-grain weight compared to the wild type (Supplementary Fig. S1A).

Figure 1.

Phenotypic characterization of the flo9 mutant. A) Comparison of the representative WT and flo9 mature grains. Bars = 2 mm. Arrows and arrowheads indicate dimpled indentations and constrictions at the dorsal and ventral sides, respectively. B) Transverse sections of the representative WT and flo9 mature grains. Bars = 1 mm. Asterisk indicates the hollow core. C) Comparison of SGs observed via scanning electron microscope in magnified regions of WT and flo9 indicated by black squares in B). Bars = 20 μm. D) Iodine-stained transverse sections of the representative WT and flo9 maturing grains at 25 DAF. Bars = 1 mm. Asterisk indicates the hardly-stained core. E) Determination of total starch, amylose, amylopectin, and WSG contents of WT and flo9 mature grains. Values are means ± SD. **P < 0.01 by Student’s t-test (n = 3). F) Differences in CLD patterns of total glucans between WT and flo9. Note that rice powder collected from the periphery, intermediate, and center of grains was separately used to determine the glucan CLD.

Iodine staining revealed 2 different areas of endosperm in flo9: a well-stained area corresponding to the translucent and floury regions, and a barely stained area corresponding to the hollow core (Fig. 1D). Quantification showed that, compared to the wild type, total starch and amylopectin contents were reduced, whereas amylose, protein, and lipid contents were increased (Supplementary Fig. S1, B to G). To further investigate the causal effect of the flo9 mutation on storage substance accumulation, we presented the contents of storage substances per grain given the decreased grain weight. Strikingly, amylose and protein contents were unaffected, whereas total starch and amylopectin contents were reduced by 31.7% and 38.05%, respectively (Fig. 1E; Supplementary Fig. S1H). These results indicate a possible specific defect in amylopectin biosynthesis in flo9. The increase in lipid content may be due to a secondary effect of disrupted starch biosynthesis, as reported previously for many floury endosperm mutants such as flo10, flo15, and floury shrunken endosperm5 (fse5) (Wu et al. 2019; You et al. 2019; Wang et al. 2021).

Given the spatial phenotypic difference of flo9 endosperm (Fig. 1B), we separately examined the chain-length distribution (CLD) of total glucans in peripheral, intermediate, and central regions of flo9 endosperm. As shown in Fig. 1F, the glucan CLD in flo9 peripheral region varied little from that of wild type, while the glucan CLD in flo9 intermediate region exhibited a moderate increase in short chains spanning 6 to 11 degrees of polymerization (DP) and a subtle decrease in intermediate chains compared with those of wild type. Strikingly, we observed a dramatic increase in short chains of 6 < DP < 14 and a relatively gentle variation in intermediate chains in the central part of flo9 endosperm (Fig. 1F).

The absent iodine staining and abnormal glucan CLD in flo9 central regions were most reminiscent of the isa1 mutants, which accumulate WSGs instead of amylopectin (Nakamura et al. 1997; Nakagami et al. 2017). Supporting this notion, the WSG content was considerably elevated in flo9 relative to the wild type (Fig. 1E). Consistent with disrupted starch biosynthesis, both the starch pasting properties and gelatinization properties were also affected in flo9 (Supplementary Table S1; Supplementary Fig. S1, I to K). Together, these results indicate that the flo9 mutation predominantly affects amylopectin biosynthesis in rice endosperm.

The flo9 mutation disturbs compound starch granule initiation

To investigate the cellular basis of the abnormal starch accumulation observed in flo9, we prepared semi-thin and thick sections of developing grains at 6, 9, and 12 days after flowering (DAF). In line with Fig. 1D, SGs in the peripheral and intermediate parts of flo9 endosperm were well-stained with iodine, whereas SGs in the central part of flo9 endosperm, if any, were barely stained (Supplementary Fig. S2, A and B). To further investigate the origin and formation of the hollow core observed in flo9, we stained thick sections of wild type and flo9 developing endosperm using a non-specific β-glucan dye, Calcofluor White. As shown in Supplementary Fig. S2C, similar to wild type, the cell wall structures were readily observed in flo9 central endosperm cells early at 6 DAF. These results suggest that the central cells of flo9 endosperm are formed normally but devoid of typical SGs.

We next compared the morphological differences of SGs in peripheral, intermediate, and central endosperm cells between the wild type and flo9. The sizes of amyloplasts gradually increased from the outside to the inside of endosperm in the wild type, and all the amyloplasts contained polyhedral, sharp-edged, and well-defined granules (Fig. 2, A to C and G to I). At the endosperm periphery, the size and morphology of amyloplasts were largely comparable between the wild type and flo9 (Fig. 2, A, D, G, and J). Compared with the characteristic sharp-edged granules in the intermediate part of wild-type endosperm at 9 DAF (Fig. 2B), amyloplasts in flo9 were filled with numerous scattered, tiny granules at the same stage (Fig. 2E).

Figure 2.

Defects in SG initiation of flo9 grains. A to L) Comparison of SGs from iodine-stained semi-thin sections prepared from developing WT and flo9 endosperm at 9 DAF (A to F) and 12 DAF (G to L). Periphery, intermediate, and center indicate the translucent, floury-white, and hollow regions of flo9 endosperm, respectively, and the corresponding regions of the WT. Asterisks in E) indicate abnormal amyloplasts containing tiny granules. Arrowheads in F) indicate atypical amyloplasts stained weakly with iodine and lacking compound structure. Arrows in F) indicate the amorphous SGs filled with pink-stained phytoglycogen-like substances. Asterisks in K) indicate the amorphous structures with a dark-stained edge in flo9 amyloplasts. Arrows in L) outline the contiguous region that is not stained by iodine. Bars = 10 μm.

During endosperm development, a large, amorphous structure with a dark-stained edge formed inside amyloplasts in the flo9 intermediate regions (Fig. 2K). Notably, we readily observed atypical amyloplasts lacking compound granules in the central part of flo9 early at 9 DAF (Fig. 2F). These abnormal amyloplasts in flo9 were also observed in our previously reported flo6 mutant, suggesting that both proteins most likely play a similar functional role in modulating amyloplast development (Peng et al. 2014). Consistent with the increased levels of WSGs in the flo9 mutant (Fig. 1E), the central part of flo9 endosperm was mainly filled with pink-stained substances (Fig. 2, F and L; Supplementary Fig. S2D), likely corresponding to phytoglycogen, as reported previously for isa1 (Nakamura et al. 1997; Matsushima et al. 2010; Nagamatsu et al. 2022). Taken together, these results suggest that the loss of FLO9 function affects compound starch granule initiation in rice endosperm, especially in the intermediate and central parts.

FLO9 encodes an Arabidopsis LESV homolog OsLESV that is predominantly expressed in endosperm

For genetic analysis, we performed a reciprocal cross between flo9 and the wild type. In the F₂ progeny, wild-type grains and flo9 grains segregated with a ratio of approximately 3:1 (Supplementary Table S2), indicating that the flo9 phenotype is controlled by a single recessive nuclear gene. To isolate the causal gene, we generated an F₂ segregating population by crossing flo9 with the indica variety Nanjing 11. Using 969 individuals with the flo9 phenotype from the F₂ population, we delimited the target locus to a 253-kb genomic region on chromosome 11 containing 31 putative open reading frames (Fig. 3A). Genomic sequencing revealed a C-to-A substitution within the first exon of Os11g0586300 (OsLESV; Feike et al. 2016), which putatively introduces a premature stop codon in place of the wild type Ser-64 residue (Fig. 3B).

Figure 3.

Map-based cloning of the OsLESV gene and expression pattern of the OsLESV protein. A) Fine mapping of the OsLESV locus. The OsLESV locus was located to a 253-kb region between markers J13 and J31 (red vertical lines). Numbers of recombinants and molecular markers are shown. B) Genomic structure and the mutation site of OsLESV. A single nucleotide substitution in the first exon of OsLESV led to a premature stop codon in flo9. Arrowhead indicates the mutation site in flo9. C) Expression of the OsLESV gene driven by its native promoter restored the grain appearance (the upper panel) and SG morphology (the lower panel). L1 to L3 indicate three independent T₄ generation transgenic lines. Bars in the upper panel = 1 mm; bars in the lower panel = 10 μm. D) Determination of total starch and E) amylopectin contents in mature grains of WT, flo9, and complemented transgenic lines. Values are means ± SD. P < 0.05 by Duncan’s multiple range tests (n = 3). F) OsLESV antibodies specifically recognize the endogenous OsLESV protein in total protein extracts of mature grains from WT and complemented transgenic lines but not in flo9. Anti-actin antibody was used as a loading control. G) Protein accumulation profiles of OsLESV in various tissues and different developmental stages of endosperm. Anti-actin antibody was used as a loading control. H) Immunoblot analysis of the spatial distribution of OsLESV within endosperm using total protein extracts from different portions of mature grain and developing endosperm by the OsLESV antibodies. Mature grain-polishing ratio refers to the ratio (w/w) of polished rice to the brown rice. Brown rice was sequentially polished to 90%, 70%, 50%, 30% (w/w) of its original weight (approximately 10 g). Rice powder produced at each polishing step was collected as samples with a 90% to 70%, 70% to 50%, or 50% to 30% of grain-polishing ratio. The gradient decrease of grain-polishing ratio indicated that samples were collected from the exterior to the interior of rice endosperm, thus representing the periphery, intermediate, and center of the endosperm, respectively. Anti-actin antibody was used as a loading control. Three independent experiments were performed. I) Quantification of OsLESV protein in different fractions of H). The intensity of OsLESV was normalized by the loading control of anti-actin antibody using the Image J software. Values are means ± SD (n = 3). The average values were shown on the corresponding columns.

To test whether the mutation in OsLESV is responsible for the flo9 phenotypes, we introduced a construct harboring the full-length coding sequence of OsLESV driven by its native 1,972-bp promoter into homozygous flo9 calli for complementation test. Seeds collected from three T₄ generation transgenic lines showed a complete restoration of the wild-type phenotypes, including grain appearance, compound granule arrangement as well as both starch and amylopectin contents (Fig. 3, C to E). We also generated 3 independent knockout mutants of OsLESV using the CRISPR/Cas9-mediated genome editing technology (Supplementary Fig. S3A). Grains harvested from these mutant lines phenocopied the flo9 mutant (Supplementary Fig. S3, B and C). Furthermore, we raised polyclonal antibodies against OsLESV, which specifically recognized a protein with the expected size in immunoblots of mature grain extracts from the wild type and complemented transgenic lines, but not from the flo9 or knockout mutants (Fig. 3F; Supplementary Fig. S3D), confirming the effectiveness and specificity of the anti-OsLESV antibodies. Together, these results demonstrate that OsLESV is indeed the causal gene responsible for the flo9 phenotypes.

OsLESV is predicted to encode a protein of unknown function composed of 618 amino acids that harbor a tryptophan (Trp)-rich region at its C terminus (Supplementary Fig. S4A). The point mutation in OsLESV of the flo9 mutant introduces a premature stop codon; if the truncated protein accumulates, it would completely lack the conserved Trp-rich region. A BLAST search showed that OsLESV is a single copy gene in the rice genome, but it is highly conserved among starch-synthesizing organisms including land plants and green algae (Supplementary Fig. S4, B and C). Arabidopsis LESV is reported to be involved in starch metabolism in Arabidopsis leaves (Feike et al. 2016). Similarly, the loss of OsLESV function also significantly affected starch turnover in leaves over the day–night cycle, as demonstrated by the disrupted starch and soluble sugar metabolisms (Supplementary Fig. S5).

To investigate the protein accumulation pattern of OsLESV in wild-type plants, we performed an immunoblot analysis of total protein (T) extracts from multiple tissues: roots, stems, leaves, leaf sheaths, panicles, and endosperm at different stages of development. As shown in Fig. 3G, we detected OsLESV protein in all tissues tested, with substantially higher levels in developing endosperm, where it increased during grain filling and peaked at 12 DAF, then subsided.

Because flo9 exhibited the most severe developmental defects in inner endosperm cells, we expected that OsLESV might be abundant in the interior of the grain. To test this notion, we divided developing endosperm into the outer and inner parts, and mature brown grains into the periphery, intermediate, and center, followed by a quantitative immunoblot analysis. OsLESV abundance did show a gradient increase from the exterior to the interior in both developing and mature endosperm (Fig. 3, H and I), which was consistent with the endosperm defects of flo9 (Fig. 1B). Together, these results demonstrate that OsLESV is a plant-specific protein with higher abundance in endosperm, especially in the interior region.

OsLESV associates with SGs by directly binding to starch

The online tool TargetP predicted that OsLESV harbors a 38-amino acid chloroplast transit peptide (cTP) at its N terminus (Emanuelsson et al. 1999; Supplementary Fig. S4A). To confirm the plastid localization of OsLESV, we transiently transfected a Pro35S:OsLESV-green fluorescent protein (GFP) construct into rice protoplasts. The OsLESV-GFP fusion protein exhibited a disc-like localization pattern within chloroplasts (Fig. 4A). To ascertain whether these structures represent SGs, we co-transfected Pro35S:OsLESV-GFP with Pro35S:GBSSII-mCherry, encoding an SG-bound marker protein fused to the fluorescent protein mCherry (Dian et al. 2003), into rice protoplasts. Notably, both fusion proteins co-localized within chloroplasts (Fig. 4B).

Figure 4.

OsLESV binds to starch in vivo and in vitro. A) Representative confocal microscopy images showing that OsLESV-GFP is localized to disc-like structures within chloroplasts of rice protoplasts. Blue signals are autofluorescence from chlorophylls in chloroplasts. The rightmost panel represents the merged image of GFP, mCherry, and chlorophyll fluorescence signals. Bars = 5 μm. B) Representative confocal microscopy images showing that OsLESV-GFP colocalizes with GBSSII-mCherry within chloroplasts of rice protoplasts. Bars = 5 μm. C to E) Different deletions or mutations of OsLESV-GFP fusion vectors were transiently co-expressed with GBSSII-mCherry in rice protoplasts, respectively. Note that OsLESV(ΔCT^346–618)-GFP lacking the C-terminal tryptophan (Trp)-rich region displayed a stromal localization pattern, whereas OsLESV(ΔNT^39–345)-GFP was tightly associated with GBSSII-mCherry within chloroplasts. CT and NT separately indicate the C and N termini of OsLESV. Bars = 5 μm. F) Representative confocal microscopy images showing that OsLESV(ΔNT^39–345)-GFP with mutations of 16 evolutionarily conserved Trp to Ala residues abolished its co-localization with SGs. NT, N terminus of OsLESV; 16W-16A, the mutation of 16 evolutionarily conserved tryptophan (W) residues to alanine (A) in the tryptophan-rich region of OsLESV. Bars = 5 μm. G) Association of OsLESV with starch in developing WT endosperm (9 days after flowering). FLO4 (a cytosolic protein), BEI (a granule-associated protein), and GBSSI (a granule-bound protein) were used as marker proteins for different fractions. The volume of each sample subject to SDS-PAGE was 10 μL. Arrowhead indicates the target band of ISA1. The relative band intensity of proteins was calculated by image J software. Three independent experiments were performed. H to J) Binding of recombinant GST-OsLESV protein or its variants to Sephadex G-10 beads H), amylopectin I), and amylose J) in vitro. CT and NT separately indicate the C and N termini of OsLESV. K) GST-OsLESV-CT recombinant protein but not GST-OsLESV-NT protein specifically bound to amylopectin in vitro. CT and NT separately indicate the C and N termini of OsLESV. L) Mutations of evolutionarily conserved tryptophan Trp residues in the GST-OsLESV-CT recombinant protein abolished its binding to amylopectin in vitro. Equivalent volume (10 μL) of each sample was loaded in (H to L). Asterisk indicates target band of GST-OsLESV-CT(16W-16A) recombinant protein. 16W-16A indicates the mutation of 16 conserved tryptophan (W) residues to alanine (A).

To clarify the domain responsible for OsLESV localization, we generated several fusions between GFP and OsLESV domain truncations. As expected, deletion of the cTP motif (OsLESVΔcTP^1–38-GFP) abolished the chloroplast localization of OsLESV (Fig. 4C). By contrast, the N-terminal 345 amino acid fragment of OsLESV (OsLESVΔCT^346–618-GFP) localized to chloroplasts but exhibited a diffuse signal at the chloroplast periphery, instead of the disc-like signal presented by the full-length protein (Fig. 4D). These results indicate the important role of the C-terminal Trp-rich region for intra-chloroplast distribution of OsLESV-GFP. Moreover, a chimeric OsLESV protein with cTP fused with the C-terminal region (OsLESVΔNT^39–345-GFP) co-localized with GBSSII-mCherry (Fig. 4E), and the mutation of 16 evolutionarily conserved Trp residues to alanine (Ala) (OsLESVΔNT^39–345[16W-16A]-GFP) abolished the localization of OsLESV to SGs (Fig. 4F; Supplementary Fig. S6A). These results indicate that OsLESV is associated with SGs and that its C-terminal Trp-rich region is essential for its sub-plastid localization.

To evaluate the intracellular localization of OsLESV in developing endosperm, we introduced an OsLESV-GFP construct into the flo9 background under the control of the native OsLESV regulatory elements, namely its promoter, intron, and downstream regulatory region (Supplementary Fig. S7A). The appearance of transgenic grains that accumulated the OsLESV-GFP fusion protein was fully restored to that of the wild type (Supplementary Fig. S7, B to D), demonstrating that the OsLESV-GFP fusion protein is biologically functional. Moreover, we also introduced the OsLESV-GFP fusion construct driven by the UBIQUITIN promoter into the flo9 mutant. Transgenic grains with high levels of OsLESV-GFP, as manifested by immunoblotting, also exhibited a wild-type translucent appearance (Supplementary Fig. S8). Furthermore, we performed immunogold electron microscopy of ultra-thin sections prepared from the ProUbi:OsLESV-GFP transgenic endosperm cells in the flo9 background. As shown in Supplementary Fig. S9, gold particles were detected particularly on the starch granules and amyloplast stroma.

To examine whether OsLESV indeed associates with SGs in developing endosperm, we performed a subcellular fractionation assay, followed by immuno-detection with protein-specific antibodies. Accordingly, we separated proteins from developing endosperm into 3 distinct fractions: soluble proteins (S), proteins loosely bound to SGs (LBP), and proteins tightly bound to SGs (TBP). LBP corresponds to proteins that are loosely adsorbed to the granule surface and easily separated by extensive washes or protease digestion, while TBPs are tightly encapsulated into granules and isolated only in boiling SDS buffer (Boren et al. 2004; Grimaud et al. 2008). Immunoblot analysis showed that OsLESV is present in all 3 protein fractions, with relatively higher proportions in the S and LBP than in the TBP fraction (Fig. 4G). Notably, ISA1 accumulated only in the S and LBP fractions (Fig. 4G). As controls, we detected FLO4 (a pyruvate orthophosphate dikinase B [PPDKB]) and GBSSI exclusively in S and TBP, respectively, whereas BEI was present in both the S and LBP fractions (Fig. 4G), which is consistent with previous reports (Utsumi et al. 2011; Hayashi et al. 2018).

To determine whether OsLESV binds to starch directly, we conducted an in vitro glucan-binding assay (Fig. 4; Supplementary Fig. S6B). As control, we used non-starch substance Sephadex G-10 beads to exclude the non-specific binding or protein precipitation (Abt et al. 2020; David et al. 2022), and the results showed that no proteins were detected after the final wash (W) (Fig. 4H). Notably, recombinant glutathione S-transferase (GST)-OsLESV protein, but not free GST protein, was able to bind to both amylopectin and amylose (Fig. 4, I and J).

Truncation analysis showed that the C-terminal Trp-rich region of OsLESV is responsible for its binding to starch (Fig. 4K). To investigate the role of conserved Trp residues in the ability of OsLESV binding to starch, we first evaluated the folding status of GST-OsLESV-CT and GST-OsLESV-CT(16W-16A) recombinant proteins using the circular dichroism (CD) spectroscopy. The results showed that both proteins exhibited similar secondary structure, indicating that the 16 evolutionarily conserved Trp residues mutations did not obviously affect the folding of OsLESV (Supplementary Fig. S6, C to E). Consistent with the effect of mutated Trp residues on the subcellular localization of OsLESV (Fig. 4F), these mutations completely abolished OsLESV binding to starch (Fig. 4L). Together, these results suggest that OsLESV associates with SGs by virtue of its C terminus, similar to Arabidopsis LESV (Liu et al. 2023).

OsLESV is required for the targeting of ISA1 to SGs

These findings that flo9 partially phenocopied isa1 mutants (Figs. 1 and 2) and that OsLESV shared a similar expression pattern with ISA1 in developing endosperm (Supplementary Fig. S10) prompted us to investigate whether OsLESV functions cooperatively with ISA1 in regulating endosperm starch biosynthesis. To this end, we performed yeast (Saccharomyces cerevisiae) two-hybrid (Y2H) assay to assess the possible interaction between OsLESV and ISA1. Strikingly, Y2H assay showed that OsLESV strongly interacts with ISA1 in yeast (Fig. 5A; Supplementary Fig. S11). Domain truncation analysis further showed that the N terminus (OsLESVΔCT^346–618), but not the C terminus (OsLESVΔNT^1–345) of OsLESV specifically binds to ISA1 (Fig. 5A). We verified the interaction between OsLESV and ISA1 in vitro using a pull-down assay and in vivo using a firefly luciferase complementation imaging (LCI) assay in Nicotiana benthamiana leaf epidermal cells (Fig. 5, B and C). Moreover, we confirmed the direct interaction of OsLESV and ISA1 by conducting an in vivo co-immunoprecipitation assay with lysates from transfected rice protoplasts (Fig. 5D). Together, these results suggest that OsLESV specifically interacts with ISA1 via its N-terminal region.

Figure 5.

OsLESV physically interacts with ISA1. A) Y2H assay showing that full-length OsLESV and its N terminus interact with ISA1. AD, activation domain; DDO, control medium (SD/-Trp-Leu); QDO, selective medium (SD/-Trp-Leu-His-Ade). The empty pGADT7 vector was used as a negative control. B) In vitro GST pull-down assay showing that GST-tagged OsLESV but not free GST tag could pull-down His-tagged ISA1. The symbol “+” or “−” indicates the presence or absence of the corresponding protein. C) LCI assay showing that OsLESV can specifically interact with ISA1 in N. benthamiana leaf epidermal cells. ISA3 was used as a negative control. Colored scale bar indicates the luminescence intensity in counts per second (CPS). D) CoIP assay verified the interaction between OsLESV and ISA1 in rice protoplasts. ISA1-Flag was transiently co-expressed in rice protoplasts with OsLESV-GFP or free GFP, respectively. ISA1-Flag could be co-immunoprecipitated by OsLESV-GFP but not free GFP using anti-GFP magnetic beads. The symbol “+” or “−” indicates the presence or absence of the corresponding protein. E) Immunoblot analysis of total ISA1 and its starch association in WT and flo9 developing endosperm (9 days after flowering). Anti-actin antibody was used as a loading control. The arrowhead indicates the ISA1 band. Three independent experiments were performed. F) Quantification of ISA1 protein level in E). The intensity of ISA1 was normalized by the corresponding intensity of anti-actin antibody using the Image J software. nd, no detection. Values are means ± SD. **P < 0.01 by Student’s t-test (n = 3).

We then investigated the effect of OsLESV depletion on subcellular localization of the ISA1-GFP fusion protein (with GFP fused to its C terminus) in rice protoplasts. As shown in Supplementary Fig. S12, when co-expressed with GBSSII-mCherry in wild-type protoplasts, the ISA1-GFP fusion protein partially co-localized with GBSSII-mCherry onto SGs (Supplementary Fig. S12A). Notably, when co-expressed with GBSSII-mCherry in flo9 protoplasts, the co-localization of both fusion proteins was largely compromised (Supplementary Fig. S12B). This altered localization of ISA1-GFP in different background protoplasts was not due to the expression difference of ISA1-GFP, as assessed by quantitative immunoblot analyses (Supplementary Fig. S12, C and D). These results suggested that OsLESV is required for the association of ISA1 with SGs.

Moreover, we validated the impaired targeting of ISA1 to SGs by a subcellular fractionation assay, followed by quantitative immunoblot analysis. Compared to the wild type, we observed a decreased amount of ISA1 in the LBP in flo9 developing endosperm (Fig. 5, E and F). Interestingly, immunoblotting combined with quantitative analysis also showed that ISA1 protein levels were significantly reduced in total protein extracts from flo9 endosperm compared to the wild type (Fig. 5, E and F). Together, these results suggest that OsLESV plays an important role in the delivery of ISA1 to starch granules, and its depletion compromised the abundance of endogenous ISA1 protein.

OsLESV physically interacts with FLO6 through their N termini

We previously demonstrated that FLO6 also interacts with ISA1 and starch (Peng et al. 2014). Subcellular fractionation assay further verified that FLO6 depletion influences the association of ISA1 with starch (Supplementary Fig. S13, A to E). Based on the phenotypic defects in compound granule formation and biochemical evidence, we reasoned that OsLESV might associate with FLO6 to regulate starch biosynthesis, and thus performed Y2H assays to test this hypothesis. Indeed, OsLESV exhibited a strong interaction with FLO6 (Fig. 6A; Supplementary Fig. S13F). Domain truncation analysis revealed that the N termini of both OsLESV and FLO6 are required for their interaction (Fig. 6A; Supplementary Fig. S13F). We further verified the interaction using an in vitro pull-down assay as well as an in vivo LCI assay in N. benthamiana leaf epidermal cells (Fig. 6, B and C). Together, these results suggest that OsLESV physically interacts with FLO6 in vivo.

Figure 6.

OsLESV physically interacts with FLO6 through their N termini. A) Y2H analysis showing that OsLESV and FLO6 interact with each other via their N termini. AD, activation domain; DDO, control medium (SD/-Trp-Leu); QDO, selective medium (SD/-Trp-Leu-His-Ade). B) In vitro GST pull-down assay showing that GST-tagged OsLESV but not free GST can pull-down His-tagged FLO6. The symbols “+” or “−” indicates the presence or absence of the corresponding protein. C) LCI assay showing that OsLESV can interact with ISA1 in N. benthamiana leaf epidermal cells. Chloroplast-localized FLO7 was used as a negative control. Colored scale bar indicates the luminescence intensity in CPS. D) Yeast three-hybrid assay showing that FLO6 can enhance the interaction between OsLESV and ISA1. Note that full-length coding sequence of OsLESV was fused to binding domain (MCS I: multiple cloning site I) and driven by constitutive ADH1 promoter, whereas FLO6 was inserted into MCS II and driven by a Met-responsive promoter. β-galactosidase activity was measured using CPRG as substrate. Values are means ± SD. P < 0.05 by Duncan’s multiple range tests (n = 3). AD, activation domain; DDO, control medium (SD/-Trp-Leu); QDO, selective medium (SD/-Trp-Leu-His-Ade). E) LCI assays showing that the FLO6-GFP fusion protein but not free GFP can promote the interaction of OsLESV with ISA1. ISA3 was used as a negative control. Colored scale bar indicates the luminescence intensity in CPS. F) Effects of OsLESV and FLO6 on ISA1 binding to amylopectin in vitro. Equal amount of each recombinant protein was combined as indicated and co-incubated with amylopectin for 30 min. Amylopectin was pelleted by centrifugation. Proteins in the supernatant (S), the final wash (W), and the pellet (P) were subject to immunoblot analyses using anti-OsLESV, anti-FLO6, and anti-ISA1 antibodies, respectively. Equivalent volume (10 μL) of each sample was loaded. Three independent experiments were performed. The symbol “+” or “−” indicates the presence or absence of the corresponding protein. G) Quantification of the ISA1 protein binding to amylopectin in F). The percentage of starch-binding ISA1 in total ISA1 (S + W + P) was quantified with the Image J software. Values are means ± SD. P < 0.05 by Duncan’s multiple range tests (n = 3). nd, no detection. H) Phenotypic analyses of the flo6 flo9 double mutant. Top panel: Iodine-stained transverse sections of developing grains of W017 (wild type for flo9), flo9, Nipponbare (wild type for flo6), flo6, Kitaake (wild type for isa1), and isa1 at 25 DAF. Bottom panel: The left shows the developing flo6 flo9 caryopsis from 3 to 25 DAF (numbers above denote the DAF). Note that the core region of flo9 as well as most of the isa1 endosperm was not stained by iodine. The right shows a transverse section of the flo6 flo9 grain at 25 DAF. Bars = 2 mm. I) Immunoblot analysis of starch synthesis-related enzymes protein abundance in total protein extracts of developing endosperm of the flo9, flo6, and isa1 single mutant, and their corresponding wild type, as well as the double mutant flo6 flo9 at 25 DAF. Anti-actin antibody was used as a loading control. The arrowhead indicates the band of ISA1.

Given that both OsLESV and FLO6 can interact with ISA1, we hypothesized that OsLESV might act together with FLO6 to facilitate the access of ISA1 to starch granules. Quantitative analysis of β-galactosidase activity in a yeast three-hybrid assay and an in vivo LCI assay revealed that FLO6 expression significantly enhances the strength of the interaction between OsLESV and ISA1, and vice versa (Fig. 6, D and E; Supplementary Fig. S14). Furthermore, we performed an in vitro starch-binding assay to investigate whether OsLESV and FLO6 function together to facilitate the binding of ISA1 to amylopectin. As controls, none of the proteins could bind to the non-starch substance Sephadex G-10 (Supplementary Fig. S15).

Consistent with our previous report (Peng et al. 2014), GST-ISA1 alone was unable to bind to amylopectin (Fig. 6, F and G), whereas upon co-incubation with either GST-OsLESV or GST-FLO6, we observed a considerable amount of GST-ISA1 retained in the pellet fraction. It seems that OsLESV exhibits a significantly stronger ability in mediating ISA1 binding to starch than FLO6 (Fig. 6, F and G), although both fusion protein themselves showed a similar binding ability to starch, as evidenced by the percentage of protein present in the pellet fraction (Fig. 6F; GST-FLO6: 73.37 ± 4.32% versus GST-OsLESV: 71.55 ± 6.23%). Notably, a higher proportion of GST-ISA1 was retained in the pellet fraction in the presence of both GST-OsLESV and GST-FLO6 (Fig. 6, F and G). Collectively, these results suggest that OsLESV physically interacts with FLO6 to facilitate ISA1 targeting to starch.

OsLESV genetically interacts with FLO6 to regulate starch biosynthesis and endosperm development in rice

To examine the genetic interaction between OsLESV and FLO6, we identified 2 new flo6 and isa1 mutants and confirmed their defects in endosperm development by iodine staining during late development (Supplementary Fig. S16, A and B; Fig. 6H). Furthermore, we attempted to cross female flo6 with male flo9, but we were not able to obtain double homozygous mutant plants. We then planted flo9/FLO9 flo6 and flo9 flo6/FLO6 heterozygous plants and identified double homozygous mutant grains by genotyping. Notably, double homozygous flo6 flo9 grains were filled with liquid, and displayed a severely shrunken appearance (Fig. 6H). The double mutant had abolished accumulation of ISA1 protein and obviously reduced levels of PUL and BEI (Fig. 6I). The double mutant displayed a dramatic disruption of starch biosynthesis than either single mutant, as evidenced by the measured starch and WSG contents (Supplementary Fig. S16C).

We noted that the flo6 flo9 double mutant exhibits a more severe phenotypic defect than isa1, suggesting that both proteins exert some other functions apart from affecting ISA1 targeting to starch granules. To support this notion, we knocked out OsLESV or FLO6 in the isa1 background to create the flo9 isa1 and flo6 isa1 double mutant, respectively (Supplementary Fig. S17). It is noted that loss of ISA1 function did not influence the protein abundance of either OsLESV or FLO6 (Supplementary Fig. S17, A to D). As expected, loss-of-function of either gene substantially exaggerated the phenotypic defect of isa1, manifested by the more shrunken grain and obviously decreased starch contents (Supplementary Fig. S17, E and F). A recent study also reported that the barley (Hordeum vulgare) flo6 isa1 double mutant produced grains with more severe phenotypic defects compared with the isa1 mutant (Matsushima et al. 2023). Taken together, several lines of genetic evidence suggest that OsLESV and FLO6 indeed exert other function(s) other than targeting ISA1 to starch granules during endosperm development.

Discussion

FLO9 encodes the Arabidopsis LESV homolog OsLESV that functions in ISA1 binding to starch granules in rice endosperm

Amylopectin accounts for up to 75% to 80% of starch weight; however, the regulatory mechanisms of amylopectin biosynthesis remain obscure. In cereal endosperm, it is thought that amylopectin-synthesizing enzymes such as SSI and BEIIb are delivered to the surfaces of starch granules via their interactions with SSII (Liu et al. 2012). In addition, some starch-synthesizing enzymes/regulators could bind starch granules through their own starch-binding domains, such as the CBM53 domain of AtSS3 and the CBM48 domain of OsBEI (Chaen et al. 2012; Abt et al. 2020). ISA1 is responsible for removing misplaced branches of amylopectin at the outer edges of starch granules, and its functional loss led to the dramatic accumulation of phytoglycogen instead of amylopectin, suggesting its essential roles in amylopectin biosynthesis (Nakamura et al. 1997; Fujita et al. 2003; Nakagami et al. 2017).

Although a putative CBM48 domain was predicted in rice ISA1 (Du et al. 2018), it appears to be longer than the canonical CBMs (147 amino acids [aa] versus 90 to 130 aa), and two conserved Trp residues critical for starch binding are mutated (Supplementary Fig. S18; Christiansen et al. 2009), likely influencing its starch-binding efficiency. Supporting this hypothesis, our biochemical analysis verified that ISA1 cannot bind directly to starch by itself in vitro (Fig. 6F; Peng et al. 2014). In rice endosperm, ISA1 is predominantly present as a homo-oligomer, with a small fraction forming a hetero-oligomer with the catalytically inactive ISA2 (Utsumi et al. 2011). Interestingly, ISA2 has been proposed to assist in ISA1 substrate binding in maize (Zea mays) and potato (Solanum tuberosum) (Hussain et al. 2003; Mehrpouyan et al. 2021). However, the ISA2 loss-of-function mutant exhibited no obvious phenotypic defects in starch biosynthesis of rice endosperm (Utsumi et al. 2011). These findings suggest that plants may have evolved complex mechanisms to facilitate the translocation of ISA1 from the stroma to starch granules, and the molecular machinery by which ISA1 is delivered to starch granules remains to be identified in plants.

Here, we identified flo9 as an endosperm mutant with a major defect in amylopectin biosynthesis (Fig. 1). Notably, apart from its conspicuous opaque phenotype, flo9 produces endosperm with a hollow core (Fig. 1B), which is distinct from other floury endosperm mutants (Bao 2019; Zhao et al. 2022). The defect in iodine staining in the central region (Fig. 1D) together with the amyloplast developmental defects in flo9 (Fig. 2) are most reminiscent of the isa1 and sugary-2 rice mutants (Nakamura et al. 1997; Matsushima et al. 2010; Nakagami et al. 2017; Nagamatsu et al. 2022). Although the causal gene responsible for the sugary-2 phenotype remains unknown, an allelic test showed that sugary-2 is not allelic to isa1 (Nakagami et al. 2017). Our genetic evidence indicates that flo9 is also not a weak isa1 allelic mutant; unexpectedly, this causal gene responsible for flo9 phenotypes encodes a homolog of Arabidopsis LESV (Fig. 3; Feike et al. 2016).

Arabidopsis LESV is highly expressed in leaves, especially during senescence (http://bar.utoronto.ca/), whereas OsLESV showed substantially higher abundance in developing endosperm, particularly early development (Fig. 3G), suggesting that OsLESV plays an important role in starch metabolism in endosperm. In addition, the abundance of OsLESV increased gradually from the exterior to interior of the endosperm (Fig. 3, H and I), coinciding with the most affected parts in flo9 endosperm (Fig. 1B). These results suggest that starch biosynthesis in rice endosperm is spatially orchestrated by factors with distinct sub-endosperm accumulation abundance. Supporting this notion, loss-of-function mutation of FLO7, which is predominantly expressed in grain periphery, causes a defect in starch synthesis specific in peripheral endosperm (Zhang et al. 2016).

Unlike Arabidopsis PTST1, which transiently binds to starch via the CBM48 domain (Seung et al. 2015), OsLESV could strongly bind to both amylose and amylopectin via its C-terminal Trp-rich region (Fig. 4, H to K), resembling Arabidopsis LESV (Feike et al. 2016; Singh et al. 2022; Liu et al. 2023). Mutagenesis of those Trp residues located in OsLESV dramatically abolished its binding to amylopectin (Fig. 4L; Supplementary Fig. S6, C to E). More strikingly, we found that OsLESV physically interacts with ISA1 via its N-terminal region (Fig. 5, A to D). When ISA1-GFP was overexpressed in wild-type rice protoplasts, a small proportion of ISA1 was localized to starch granules (Supplementary Fig. S12A). This observation could be explained by the presence of endogenous OsLESV protein. Supporting this notion, loss of OsLESV function largely impaired the recruitment of ISA1 onto SGs (Fig. 5, E and F; Supplementary Fig. S12, B to D). Our findings suggest that OsLESV may function as a scaffold protein to help deliver ISA1 onto starch granules.

OsLESV functions cooperatively with FLO6 to regulate starch biosynthesis and endosperm development

FLO6 is identified as a non-enzymatic interacting partner of ISA1 in plants, although its precise function in starch metabolism has been unclear (Peng et al. 2014). flo6 and flo9 exhibited a similar defect in starch granule initiation (Fig. 2; Peng et al. 2014). Furthermore, we found that either OsLESV or FLO6 depletion influenced the distribution of ISA1 onto starch granules (Fig. 5, E and F; Supplementary Fig. S13, D and E). These observations indicated that OsLESV and FLO6 might function, at least in part, in a common starch metabolism pathway in rice endosperm. This hypothesis is further supported by the strong physical interaction of OsLESV with FLO6 via their N-terminal regions in vitro and in vivo (Fig. 6, A to C).

Furthermore, the co-expression of OsLESV and FLO6 significantly enhanced the interactions of either protein with ISA1 (Fig. 6, D and E; Supplementary Fig. S14). An in vitro starch-binding assay indicated that OsLESV exhibited a stronger ability to bring ISA1 to the proximity of starch granules than FLO6, and that co-incubation of OsLESV and FLO6 had a significantly stronger effect than either protein alone (Fig. 6F). Genetically, our evidence showed that loss of both OsLESV and FLO6 functions dramatically disrupted starch biosynthesis, and generated a severely shrunken grain (Fig. 6H; Supplementary Fig. S16). These results suggested that OsLESV acts synergistically together with FLO6 to control starch biosynthesis and endosperm development.

Based on these findings, we proposed a working model for the role of OsLESV and FLO6 in regulating starch biosynthesis and endosperm development (Fig. 7). According to this model, OsLESV and FLO6 are able to recruit ISA1 from the amyloplast stroma through their N-terminal domains. By virtue of the Trp-rich C-terminal domain of OsLESV and the CBM48 domain of FLO6, ISA1 is brought to the surface of starch granules where it is required to remove misplaced starch branches, thereby ensuring the formation of the semi-crystalline amylopectin matrix. Loss of OsLESV function considerably compromises the distribution of ISA1 onto starch granules, whereas loss-of-function mutations of both OsLESV and FLO6 substantially decrease the protein abundance of ISA1, and dramatically disrupt starch biosynthesis and endosperm development.

Figure 7.

Working model for the OsLESV-FLO6 molecular module in facilitating ISA1 binding to starch granules in rice endosperm. In the WT endosperm, OsLESV and FLO6 form a functional protein complex to recruit ISA1 from the stroma to starch granules, where ISA1 is responsible for the removal of misplaced branches in amylopectin. Loss of OsLESV function considerably compromises the distribution of ISA1 onto starch granules, whereas loss-of-function mutations of both OsLESV and FLO6 substantially decreased the protein abundance of ISA1, and dramatically disrupted starch biosynthesis and endosperm development. Loss of ISA1 function accumulates large amounts of phytoglycogen instead of starch, and thus disrupts the formation of higher-order amylopectin structure and amylopectin crystallizing. The thick solid arrow indicates an effective targeting of ISA1 to starch granules, while the thin solid arrow and dashed arrows indicate weak and disturbed binding of ISA1 to starch granules, respectively. In addition, the dotted arrow indicates a possible functional role of OsLESV and FLO6 in the delivery of other unknown cargos onto starch granules during starch biosynthesis. Insets inside of starch granule model denote the I₂-KI staining of transverse sections of corresponding grains.

We also noted the stronger flo6 flo9 phenotype relative to isa1 (Fig. 6H), suggesting extra function roles of OsLESV and FLO6 in modulating starch synthesis apart from targeting ISA1 to starch granules. Supporting this notion, our genetic evidence showed that either OsLESV or FLO6 genetically interacts with ISA1 to regulate starch biosynthesis and endosperm development (Supplementary Fig. S17). Similar genetic interactions were recently reported for LESV and ISA in Arabidopsis leaves as well as for FLO6 and ISA1 in barley endosperm (Liu et al. 2023; Matsushima et al. 2023). In addition, putative interaction relationships of FLO6 with SSIVb and GBSSI&II were also reported in rice endosperm (Zhang et al. 2022). Based on these findings, we proposed that OsLESV and FLO6 might also target other unknown cargos essential for starch biosynthesis to starch granules during starch biosynthesis.

Why does the targeting of ISA1 onto starch granules require the assistance of multiple proteinaceous non-enzymatic factors? A previous study of rice endosperm suggested that ISA1 can form homohexamers (∼530 kD) and hetero-oligomers (∼450 kD) (Utsumi and Nakamura 2006). It is a possible explanation that OsLESV interacts with FLO6 to form a more powerful scaffolding protein complex to efficiently facilitate the delivery of ISA1 homohexamers to starch granules. The more severe phenotypic defect observed in flo9 than flo6 suggested that OsLESV might play a more important role than FLO6 in modulating starch biosynthesis and endosperm development. In contrast to those carbohydrate binding module (CBM) domain-containing enzymes that could directly bind to starch by themselves, such a complex targeting strategy for ISA1 delivery to starch might be more flexible to finely tune starch biosynthesis by maintaining a proper quantity of ISA1 onto starch granules during endosperm development.

Multiple physiological roles of LESV-like proteins in influencing starch biosynthesis in plants

LESV represents a type of plant-unique protein, and its functional loss generally disrupted the metabolism of both transient and storage starch in plants including Arabidopsis and rice (Feike et al. 2016; Singh et al. 2022; Liu et al. 2023; Fig. 1; Supplementary Fig. S5). Two functional roles of LESV in starch metabolism have been proposed based on in vitro biochemical and genetic data (Singh et al. 2022; Liu et al. 2023). One is the role in modification, in which LESV functions in modulating the glucan structure on the starch granule surface to influence the activity of starch metabolic enzymes (Singh et al. 2022). Another physiological role of Arabidopsis LESV is to help the assembly of amylopectin helices into starch granules (Liu et al. 2023).

In this study, we identified a rice mutant of OsLESV, and found that its functional defect dramatically disrupted the biosynthesis of storage starch in endosperm, which is distinguished from starch metabolism in vegetative organs such as leaves. Our combined biochemical, cytological, and genetic evidence pointed to a possible role of LESV-like protein as a scaffolding protein to recruit ISA1 from amyloplast stroma onto starch granule in rice endosperm (Fig. 7). Although these possible roles of LESV in modulating starch metabolism appeared to be distinguished from each other, they are not mutually exclusive. We could not rule out the possibility that OsLESV could also influence the glucan structure of the starch granule surface, whereby strengthens the binding of ISA1 to starch granules.

Unexpectedly, we also found that in contrast to FLO6 and ISA1 that accumulated in both the S and LBP fractions (Fig. 4G), a considerable amount of OsLESV was present in the TBP fraction (Fig. 4G). Intriguingly, OsLESV homologs from Arabidopsis and potatoes are also found to be encapsulated inside starch granules (Helle et al. 2018; Liu et al. 2023). Consistent with the model proposed for Arabidopsis LESV in amylopectin phase transition (Liu et al. 2023), we found that depletion of OsLESV generated substantial amounts of WSG compared to the wild type (Fig. 1E), which is most likely owing to the failure in amylopectin phase transition. Further studies should be aimed to investigate the possible multiple physiological roles of LESV in modulating storage starch metabolism in cereal crops.

Materials and methods

Plant materials and growth conditions

The flo9 mutant was isolated from an N-methyl-N-nitrosourea-induced mutant population of the rice (Oryza sativa) japonica variety W017. Reciprocal crosses between flo9 and W017 were used for genetic analysis. An F₂ mapping population was obtained by crossing flo9 with the indica variety Nanjing 11. The flo6 mutant in the Nipponbare (japonica) background and the isa1 mutant in the Kitaake (japonica) background were used in this study. We crossed flo9 with flo6 and identified the flo6 flo9 double mutant by genotyping from the segregated F₂ grains. The primers used for genotyping are listed in Supplementary Data Set 1. Plants were grown in paddy fields under natural conditions or in a greenhouse at the Chinese Academy of Agricultural Sciences, Beijing. Developing seeds at different stages of development were separately collected and frozen in −80 °C freezer for use. Nicotiana benthamiana plants were grown in a climate chambers at ∼21 °C for 4 wk with a 14 h light/10 h dark photoperiod and an illumination intensity of 120 to 150 µmol m⁻² s⁻¹ using the light-emitting diode lamps.

Microscopy

For scanning electron microscopy analysis, mature grains were transversely cut with a knife, followed by examination with a HITACHI S-3400N scanning electron microscope (Tokyo, Japan).

For visualization of SGs in developing grains, semi-thin sections were prepared as described previously (Peng et al. 2014). Briefly, transverse sections of the wild type and flo9 developing grains at 6, 9, and 12 DAF were fixed in 0.1 m phosphate buffer (pH 7.4) containing 2.5% (v/v) glutaraldehyde for 12 h at 4 °C. Fixative samples were dehydrated through a gradient ethanol series (30%, 50%, 70%, 90%, and 100% [v/v]), and sequentially embedded in LR White resin (London Resin, 14388-UC). Semi-thin sections (approximately 1 μm in thickness) were prepared with an ultramicrotome (Leica microsystems, RM2265), and stained with I₂-KI (0.5% [w/v]) or periodic acid-Schiff reagent (Wu et al. 2016) for 3 min. After washing 3 times with distilled water, the sections were examined under a Nikon ECLIPSE80i microscope.

For immune-gold electronic microscopy, ultra-thin sections were prepared from high-pressure frozen/freeze-substituted samples of flo9/ProUbi:OsLESV-GFP transgenic endosperm cells, followed by immunogold labeling as described previously (Ren et al. 2014).

Cell wall labeling was conducted according to a method previously described (Ren et al. 2014). The thick sections (100 μm) prepared from wild type and flo9 developing endosperm were stained with Calcofluor white (a non-specific dye for β-glucan, Sigma-Aldrich) at room temperature for 5 min, and washed 3 times with sodium phosphate (PBS) buffer. Images were taken using a laser scanning confocal microscope (Zeiss LSM980). A 405-nm laser excitation and 475 to 500-nm prism filter set used for Calcofluor white emission. The range of laser intensity was 0.5% to 3%. Images were taken at 400 Hz, with a picture size of 1,024 × 1,024 pixels and a below 800 gain score.

Physicochemical properties of rice grains

Soluble and insoluble glucans were prepared from dehulled rice grains as described previously (Nakagami et al. 2017). Total starch, amylose, and WSG contents were enzymatically measured using starch assay kits (Megazyme, K-TSTA-100A), following the manufacturer’s protocol. The amylopectin content was calculated as the difference between total starch content and amylose content. Lipid and protein contents of mature grains were determined as described previously (Kang et al. 2005; Liu et al. 2009). The starch pasting properties were determined with a Rapid Visco Analyzer (TecMaster RVA, Perten; Peng et al. 2014). For CLD of total glucans, the sifted rice powder (∼0.2 g) was suspended in 10 mL of methanol and boiled for 10 min, followed by centrifugation at 2,500 g for 10 min at room temperature. The resulting precipitate was washed twice with 5 mL of 90% (v/v) methanol to completely remove free sugars. Total glucans were obtained and used to assess CLD following the methods described previously (Fujita et al. 2003; Peng et al. 2014).

The gelatinization behavior of endosperm starch in urea solution was examined as described previously (Nishi et al. 2001). Briefly, approximately 20 mg of rice powder was separately mixed with 1 mL of various concentrated urea solution (0 to 9 m, pH 6.0) in a 1.5 mL Eppendorf tube. After incubation with shaking at 25 °C overnight, the mixture was centrifuged at 8,000 × g for 20 min at room temperature, followed by standing for 1 h. Images were taken with a scanner (Scanmaker 560).

Iodine staining and starch content analysis of leaves

Four-week-old seedlings were grown under a 10-h light/14-h dark diurnal cycle in a climate chamber (HP1500GS; Ruihua, Wuhan, China) at 30 °C and 60% relative humidity. Light was provided by fluorescent white-light tubes (400 μmol m⁻² s⁻¹). For iodine staining, leaves harvested at the end of the day and night were decolorized and stained in Lugol’s solution as described previously (Wang et al. 2020). The starch content was determined in the leaves harvested during the diurnal cycle, following the method as described previously (Wang et al. 2020).

Map-based cloning

Floury grains were selected from the F₂ population of flo9 and Nanjing 11 for preliminary mapping, using over 180 polymorphic simple sequence repeat markers covering the rice genome. To finely map the OsLESV locus, new genetic markers were developed by comparing the corresponding genomic sequences of the japonica variety Nipponbare and the indica variety 93-11 (Supplementary Data Set 1).

Generation of transgenic plants

Full-length coding sequence of OsLESV was amplified and cloned into the vector pCUbi1390 (at the HindIII and BamHI sites) driven by its native 1972-bp promoter to generate the ProOsLESV:OsLESV construct. GFP with linker sequences was inserted into 30-bp upstream of OsLESV stop code to generate OsLESV-GFP fusion under the control of its native regulatory elements including promoter, intron, and downstream regulatory region, and subsequently integrated into pCAMBIA2300 vector (at the EcoRI and SmaI sites) to generate the ProOsLESV:gOsLESV-GFP construct (Ren et al. 2020; Supplementary Fig. S7A). Full-length coding sequence of OsLESV was cloned into the binary vector pCAMBIA1305-GFP (at the BamHI site) driven by the maize (Zea mays) UBIQUITIN promoter (Ren et al. 2014) to generate ProUbi:OsLESV-GFP construct. To generate OsLESV or FLO6 knockout construct, a 20-bp gene-specific spacer sequence was inserted into the CRISPR/Cas9 expression vector (at the BsaI site) according to a previously described method (Miao et al. 2013).

All above constructs were individually introduced into W017, Kitaake, flo9, or isa1 mutant by Agrobacterium tumefaciens-mediated transformation (Hiei et al. 1994). Primers were listed in Supplementary Data Set 1.

Phylogenetic analysis

Sequences of OsLESV homologs were retrieved from the National Center for Biotechnology Information (NCBI, http://www.ncbi.nlm.nih.gov). The phylogenetic tree was built by the neighbor-joining method (Tamura et al. 2011), using MEGA 5.0 software (http://www.megasoftware.net/). Bootstrap values from 1,000 replicates are given. The amino acid sequences of OsLESV homologs were aligned using ClustalX (Thompson et al. 1994; http://www.clustal.org). The alignment files are provided in Supplementary Files 1 and 2.

Subcellular localization

For subcellular localization of OsLESV and its variants, the full-length and truncations of OsLESV were N terminally fused to GFP in the vector pAN580 (at the XbaI and BamHI sites) to generate the OsLESV-GFP, OsLESV(ΔcTP^1–38)-GFP, OsLESV(ΔCT^346–618)-GFP, OsLESV(ΔNT^39–345)-GFP, and OsLESV(ΔNT^39–345); (16W-16A)-GFP constructs, respectively. The sequence of the OsLESV(ΔNT^39–345); (16W-16A) was synthesized in GeneScript Biotech (Nanjing China; https://www.genscript.com.cn/). Full-length coding sequence of GBSSII was cloned into the pAN583 vector (at the XbaI and BamHI sites) with a mCherry fluorescent tag and used as an SG marker (Dian et al. 2003). These transient expression constructs were introduced into rice protoplasts following a PEG-mediated transfection method (Chen et al. 2006). Confocal imaging was performed using a laser scanning confocal microscope (Zeiss LSM980). For GFP observations, the excitation laser wavelength was set at 488 nm, and the emission laser was collected at 495 to 550 nm. For mCherry observations, the excitation laser wavelength was set at 561 nm, and the emission laser was collected at 570 to 620 nm. The range of laser intensity was 0.5% to 3%. Images were taken at 400 Hz, with a picture size of 1,024 × 1,024 pixels and a below 800 gain score.

Glucan-binding assay

Glucan-binding assays were performed in vitro following previous methods with minor modifications (Kerk et al. 2006; Lohmeier-Vogel et al. 2008; Peng et al. 2014). Full-length coding sequences of OsLESV, FLO6, and ISA1, as well as OsLESV variants were separately cloned into the pGEX4T-2 vector (at the EcoRI and BamHI sites; GE Healthcare). The expression of recombinant protein and free GST was induced in Escherichia coli BL21 (DE3) cells with 0.5 mm isopropyl β-D-thiogalactopyranoside at 16 °C for 18 h, and then purification was performed with GST beads (Beaver; 70601-100) according to the manufacturer’s instructions. Recombinant proteins (∼1 μg) were separately incubated with excess amylopectin (∼30 mg; Sigma-Aldrich, St. Louis, MO, USA), or amylose (∼30 mg; type III; Sigma-Aldrich) suspended in a binding buffer (50 mm HEPES-NaOH [pH 7.4], 2 mm MgCl₂, 1 mm DTT, 0.1% [w/v] BSA, and 0.01% [v/v] Triton X-100) with a total volume of 250 μL, by mixing end-over-end for 30 min at 20 °C. The unbound fraction was collected in the supernatant by centrifugation at 5,000 × g for 30 s. The pellets were washed 3 times with the binding buffer, and re-suspended in 250 μL of elution buffer (50 mm HEPES-NaOH, pH 7.4, 2 mm MgCl₂, 1 mm DTT, and 4% [v/v] SDS) as the bound fraction. Sephadex G-10 beads (Sigma-Aldrich), a non-starch substance was used as a negative control (Abt et al. 2020; David et al. 2022). The unbound and bound fractions were subject to SDS-PAGE, followed by immunoblotting using anti-GST antibodies.

For in vitro starch-binding assays with different protein combinations, 1 μg of each recombinant protein in different combinations was incubated with excess amylopectin (∼100 mg) in the binding buffer in a total volume of 500 μL. The subsequent procedures were performed as described above. Proteins in different fractions were detected by SDS-PAGE, followed by immunoblotting using anti-OsLESV, anti-FLO6, and anti-ISA1 antibodies. Free GST protein was used as a negative control. Sephadex G-10 beads were also used as a non-starch control for non-specific binding or for protein precipitation. Each experiment was performed three times independently with similar results.

Far-UV CD spectroscopy

The CD spectra of recombinant GST-OsLESV-CT (as control) and GST-OsLESV-CT(16W-16A) proteins were examined at 25 °C in PBS buffer at a concentration of 0.2 mg/mL with a JASCO J-710 spectropolarimeter (Tokyo, Japan). Path length of the cell was 1 mm. Recombinant protein concentration was evaluated from absorption measurements at 280 nm using a double-beam λ-25 spectrophotometer (Perkin Elmer, Norwalk, CT, USA). CD spectra and molar ellipticity were obtained over the wavelength range of 200 to 260 nm. The results are indicated by mean residue ellipticity [θ] MRW.

Antibodies

Polyclonal antibodies against starch-synthesizing enzymes, including ADP-glucose pyrophosphorylase (AGPase) subunits (AGPS2b and AGPL2, dilution 1:2,000), SSIIa (dilution 1:1,000), BEI (dilution 1:3,000), BEIIb (dilution 1:3,000), phosphorylase 1 (PHO1, dilution 1:3,000), and GBSSI (dilution 1:5,000), were produced in rabbits at Yingji Biotech (https://immunogen.bioon.com.cn/) as described previously (Long et al. 2018).

To produce polyclonal antibodies against OsLESV, FLO6, and ISA1, we cloned partial coding sequences of OsLESV(amino acids 1 to 190), FLO6 (amino acids 95 to 240), and ISA1 (amino acids 103 to 429) into pET-28a expression vector (at the EcoRI and BamHI sites; Novagen) for recombinant protein productions. After purification using the His beads (Beaver; 70501-100), approximately 1 mg of each recombination protein was obtained and injected into rabbits for polyclonal antibodies production at ABclonal biotechnology (Wuhan, China; https://abclonal.com.cn/). These antibodies were diluted at 1:2,000. Anti-actin (Abmart, M20009L), anti-GST (MBL, PM013-7), anti-His (MBL, D291-7), anti-GFP (Roche, 11814460001), anti-HA (MBL, M180-7), anti-cMYC (MBL, M192-7), and anti-Flag (Sigma, A8592) antibodies are commercially available and diluted at 1:5,000.

Protein extraction and immunoblotting

Total proteins were extracted from the developing endosperm samples or mature grains, and subject to immunoblot analysis (Wang et al. 2010; Ren et al. 2020). Briefly, developing rice grains were dehulled and homogenized in an ice-cold lysis buffer (50 mm Tris-MES, pH 7.5, 1 mm MgCl₂, 0.5 m sucrose, 10 mm EDTA, 5 mm DTT, 0.1% [v/v] Nonidet P-40, and 1× Complete Protease Inhibitor Cocktail [Roche]) in tissue:buffer ratio of 1:20 (w/v). After incubation for 2 h at 4 °C with shaking, the supernatant was collected by centrifugation at 12,000 × g for further analyses.

The fractioned rice powder from mature grains was prepared as previously described (Takahashi et al. 2019). Briefly, ∼10 g of brown rice was polished to 90% (w/w) of the original weight by removing embryo, pericarp, and aleurone layer using a rice polisher (KETT, Japan). The resulting polished rice (∼9 g) was further polished to 70%, 50%, and 30% of the original weight (∼10 g), respectively. Rice powder was sampled at each step, namely 90% to 70%, 70% to 50%, and 50% to 30% rice powder. The remaining polished rice (∼3 g) was ground into flour using a sample mill (FOSS, CT410) and sampled as 30% to 0% rice powder. Total protein extraction and immunoblot analysis were performed as described above with modification in the lysis buffer (4% SDS [w/v], 4 m urea, 5% [v/v] β-mercaptoethanol, and 125 mm Tris-HCl [pH 6.8]).

The soluble protein, the SG loosely bound protein, and the SG tightly bound protein were extracted from the wild type, flo9, and flo6 developing endosperm (9 DAF), as described previously (Fujita et al. 2006) with minor modifications. Approximately 0.1 g of developing endosperm was homogenized in 500 μL buffer A (50 mm Tris-HCl [pH 7.5], 8 mm MgCl₂, 12.5% [v/v] glycerol, and Protease Inhibitor Cocktail [Roche]). After centrifugation, the supernatant was used as the soluble fraction (S). The loosely and tightly bound fractions were prepared from the residual pellet as described previously (Fujita et al. 2006). Three independent experiments were performed. The integrated density of protein bands was calculated by the Image J software (https://imagej.nih.gov/ij/). Antibodies against ISA1, OsLESV, FLO6, FLO4, BEI, and GBSSI were used for detection.

RNA extraction and RT-qPCR analysis

Total RNA was isolated from various tissues using a ZR Plant RNA MiniPrep Kit (ZYMO Research, Irvine, California, USA) following the manufacturer’s protocol. First-strand cDNA was synthesized from 2 μg of total RNA with a QuantiTect reverse transcription kit (Qiagen, Hilden, Germany) following the manufacturer’s protocol. RT-qPCR was performed on an ABI prism 7500 Real-Time PCR System using an SYBR premix Ex Taq Kit (TaKaRa) with rice ACTIN I gene as an internal control. The relative expression level was normalized from 3 biological replicates data via 2^−△△Ct method (Livak and Schmittgen 2001). The primers are listed in Supplementary Data Set 1.

Protein-protein interaction assays

For Y2H assays, full-length or truncated coding sequences of OsLESV, FLO6, and ISA1 were separately cloned into both pGADT7 and pGBKT7 vectors using an infusion cloning kit (at the EcoRI and BamHI sites; Clontech). Various combinations of plasmids were cotransformed into yeast (Saccharomyces cerevisiae) strain AH109, followed by incubation and interaction screening according to the manufacturer’s protocols (Clontech). Yeast soluble proteins were extracted following yeast protocols handbook (Clontech). Briefly, cell pellets collected from overnight liquid cultures were re-suspended in 100 μL (per 7.5 OD₆₀₀ units of cells) of ice-cold TCA buffer (20 mm Tris-HCl, [pH 8.0], 50 mm Ammonium acetate, 2 mm EDTA, Protease Inhibitor Cocktail [Roche]). Each cell suspension was transferred into a fresh 1.5-mL centrifuge tube containing 100 μL of acid-washed glass beads (425 to 600 mm; Sigma #G-8772) and homogenized by a Bead-Beater at the highest speed for 10 min with cooling at intervals. Transfer the supernatant above the settled glass beads to fresh 1.5-mL ice-cold centrifuge tubes. The residual unbroken cells were disrupted again. Soluble proteins were collected from combined supernatant by centrifugation at 15,000 × g for 10 min at 4 °C. Protein concentration was quantified by a Bradford-based protein assay (Bio-Rad) reagent (http://www.bio-rad.com), and ∼15 μg of each sample was loaded onto the SDS-PAGE gel.

For pull-down assays, the full-length coding sequences of FLO6 and ISA1 were separately cloned into the vector pET30a (at the EcoRI and BamHI sites; Novagen) to generate the His-FLO6 and His-ISA1 constructs, respectively. The recombinant proteins were purified using the His beads (Beaver; 70501-100), according to the manufacturer’s instructions. Equal amounts (2 μg) of GST and GST-OsLESV were separately incubated with 20 μL of GST beads (Beaver; 70601-100) in 1 mL of binding buffer (50 mm Tris-HCl [pH 7.5], 100 mm NaCl, 0.5% [v/v] Triton X-100, and Protease Inhibitor Cocktail [Roche]) at 4 °C for 1 h with gentle rotation. Approximately 2 μg of purified His-tagged FLO6 or ISA1 combination protein was added and incubated for another 2 h. The beads were washed at least 3 times with the binding buffer. Eluted proteins were separated by SDS-PAGE and detected by anti-GST (dilution 1:5,000) and anti-His (dilution 1:5,000) antibodies.

For yeast three-hybrid assay, full-length coding sequence of ISA1 was cloned into the pGADT7 vector (at the EcoRI and BamHI sites; Clontech) to generate AD-ISA1, and both full-length coding sequences of FLO6 and OsLESV were cloned into the pBridge vector (at the EcoRI and BamHI sites; Clontech) to generate FLO6-OsLESV-pBridge or OsLESV-FLO6-pBridge constructs, respectively. Yeast transformation and screening were conducted as described in Y2H. β-galactosidase activity was measured by a liquid culture assay using chlorophenol red-β-D-galactopyranoside (CPRG) following the manufacturer’s protocols (Clontech).

For LCI assays, the full-length coding sequences of OsLESV, FLO6, ISA3, and FLO7 were cloned into the pCAMBIA-nLUC vector (at the KpnI and SalI sites; Chen et al. 2008) to generate OsLESV-nLUC, FLO6-nLUC, ISA3-nLUC, and FLO7-nLUC, respectively. The coding sequences of ISA1, FLO6, ISA3, and FLO7 were cloned into the pCAMBIA-cLUC vector (at the SacI site; Chen et al. 2008) to generate ISA1-cLUC, FLO6-cLUC, ISA3-cLUC, and FLO7-cLUC, respectively. All constructs were introduced into Agrobacterium tumefaciens strain EHA105. Various combinations of strains were co-infiltrated into N. benthamiana leaves as described previously (Waadt and Kudla 2008). After 2 to 3 days, the relative luciferase (LUC) activity was measured by Tanon-5200 chemiluminescent imaging system (Tanon science and technology), as described previously (Chen et al. 2008). ISA3 and FLO7 act as negative controls for the assays.

For in vivo CoIP assay, full-length coding sequence of OsLESV was cloned into the pCAMBIA1305-GFP vector (at the XbaI and BamHI sites; Ren et al. 2014) to generate OsLESV-GFP. Full-length coding sequence of ISA1 was cloned into pCAMBIA1300-221-Flag vector (at the KpnI and BamHI sites; Ren et al. 2014) to generate ISA1-Flag. Various combinations of plasmids were transiently co-expressed in rice protoplasts as previously described (Chen et al. 2006). After incubation overnight, total protein was extracted from protoplasts with 500 μL of ice-cold protein extraction buffer (50 mm Tris-HCl [pH 7.5], 150 mm NaCl, 10 mm MgCl₂, 1 mm EDTA, 5 mm DTT, 0.1% [v/v] Nonidet P-40, 10% [v/v] glycerol, and Protease Inhibitor Cocktail [Roche]), followed by incubation with 20 μL of anti-GFP mAb-Magnetic beads (MBL, D153-10) for 1 h at 4 °C with shaking. The beads were washed 3 times with extraction buffer, and the bound protein was eluted with a reducing buffer, followed by SDS-PAGE and immunoblotting using anti-GFP (dilution 1:5,000) and anti-Flag (dilution 1:5,000) antibodies.

Statistical analysis

The statistical results are indicated as means ± SD, where n represents the number of biological replicates. GraphPad Prism 5.0 and statistical software SPSS 13.0 (SPSS) were used for statistical analysis. Detailed statistical analysis data are provided as Supplementary Data Set 2.

Accession numbers

Sequence data from this article can be found in the GenBank/EMBL libraries under the following accession numbers: OsLESV (Os11g0586300), FLO6 (Os03g0686900), ISA1 (Os08g0520900), GBSSI (Os06g0133000), GBSSII (Os07g0412100), PUL (Os04g0164900), PHO1 (Os03g0758100), AGPS2b (Os08g0345800), AGPL2 (Os01g0633100), SSI (Os06g0160700), SSIIa (Os06g0229800), BEI (Os06g0726400), BEIIb (Os02g0528200), and FLO4 (Os05g0405000). Accession numbers for the sequences used in phylogenetic tree constructed were listed on the tree.

Author contributions

J.W., Y.R., and W.Z. designed the research; H.Y., W.Z., and Y.W. performed most of the experiments. J.J. screened the flo9 mutant material and cloned the gene; Y.Z. constructed the GFP-fused genomic complementation vector; Y.W., Y.F., Z.S., H.X., X.L., Y.X.W., X.H., W.X.Z., and C.Z. generated the transgenic plants and planted the rice materials; X.Z. performed some transgenic experiments; X.W., X.T., R.W., J.Z., Y.C., X.Y., J.C., X.G., J.X., L.J., S.L., C.L., X.Z., and H.W. provided technical assistance; H.Y., Y.R., and W.Z. analyzed the data and wrote the article.

Supplementary data

The following materials are available in the online version of this article.

Supplementary Figure S1. Physicochemical properties of the flo9 grains.

Supplementary Figure S2. Light microscopy images of cell structure and SGs in the WT and flo9 developing endosperm.

Supplementary Figure S3. Characterization of OsLESV knockout transgenic lines.

Supplementary Figure S4. Structure and phylogenetic tree analysis of the OsLESV protein.

Supplementary Figure S5. Altered starch turnover in flo9 leaves.

Supplementary Figure S6. Expression and purification of recombinant full-length OsLESV and variants with different domain truncations.

Supplementary Figure S7. Functional complementation of flo9 by expressing genomic OsLESV integrated with the GFP tag.

Supplementary Figure S8. Functional complementation of flo9 by overexpressing OsLESV-GFP.

Supplementary Figure S9. Immunogold localization of OsLESV-GFP in developing rice endosperm cell.

Supplementary Figure S10. Spatial-temporal expression patterns of starch, OsLESV, ISA1, and FLO6.

Supplementary Figure S11. Negative controls of Y2H assays.

Supplementary Figure S12. OsLESV is required for the localization of ISA1 onto SGs.

Supplementary Figure S13. FLO6 physically interacts with ISA1 and its deletion affects ISA1 binding to starch.

Supplementary Figure S14. OsLESV can enhance the interaction of FLO6 with ISA1.

Supplementary Figure S15. Purification and immunoblot analysis of recombinant proteins extracted from E. coli cells.

Supplementary Figure S16. Characterization of the flo6 and isa1 mutants as well as determination of carbohydrate contents in flo9, flo6, isa1, and flo6 flo9 mutants.

Supplementary Figure S17. Characterization of flo9 isa1 and flo6 isa1 double mutants.

Supplementary Figure S18. Amino acid sequence alignment of carbohydrate binding module 48 (CBM48) domain of representative CBM48 domain-containing proteins.

Supplementary Table S1. The pasting properties of endosperm starch of WT and flo9.

Supplementary Table S2. Genetic analysis of the flo9 mutant.

Supplementary Data Set 1. Primers used in this study.

Supplementary Data Set 2. Details of the statistical analysis in this study.

Supplementary File 1. Multiple sequence alignment for Supplementary Fig. S4C.

Supplementary File 2. Multiple sequence alignment for Supplementary Fig. S18.

Funding

This work was supported by grants from the National Key Research and Development Program of China (2021YFF1000200), National Natural Science Foundation of China (31830064, 91935301, and 32001518), Innovation Program of Chinese Academy of Agricultural Sciences, International Science & Technology Innovation Program of Chinese Academy of Agricultural Sciences (CAAS-ZDRW202109), Natural Science Foundation of Jiangsu Province, Major Project (BK20212010), and Jiangsu Science and Technology Development Program (BE2021359). This work was also supported by the Central Public-Interest Scientific Institution Basal Research Fund, China (Y2021YJ18).

Dive Curated Terms

The following phenotypic, genotypic, and functional terms are of significance to the work described in this paper:

• ISA1 Gramene: AT2G39930

• ISA1 Araport: AT2G39930

• ISA3 Gramene: AT4G09020

• ISA3 Araport: AT4G09020

• starch CHEBI: CHEBI:28017

• PTST1 Gramene: At5g39790

• PTST1 Araport: At5g39790