Natural Variation and Expression of OsCFF1 Affect the Formation of Rice Chalkiness

Abstract

Chalkiness, as a major undesired-index for rice quality evaluation, is the opaque/white structure in rice endosperm, and determines the rice grade and price to a large extent. Reducing chalkiness remains one of the most important goals for breeders and producers. In this study, we identified a rice chalkiness formation-related gene (OsCFF1) that encodes a glycosyltransferase (GT). Through screening of OsCFF1 alleles in 60 rice cultivars, four genotypes were identified, including WT and three natural variations. Cultivars with natural variations that cause the function loss of OsCFF1 usually produce fewer chalky grains. Higher OsCFF1 expression during the grain-filling period is usually associated with a higher degree of chalkiness. Notably, the chalkiness in gene-edited rice kernels was decreased by more than 90% after knocking out OsCFF1 through CRISPR/Cas9, indicating a quality reversal from inferior rice with high chalkiness to superior rice with very little chalkiness. We also found that OsCFF1 might catalyze the glycosylation of GA_4, and the corresponding mechanism requires in-depth exploration. In summary, OsCFF1 is a key factor controlling the formation of rice chalkiness and demonstrates enormous potential for application in high-quality rice breeding.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12284-026-00932-6.

Introduction

Chalkiness in rice (Oryza sativa L.) refers to the opaque component of the endosperm, which primarily comprises loosely arranged starch granules and protein bodies. As one of the most concerning appearance quality indicators, chalkiness negatively affects rice processing, taste, and cooking quality. Two core indices for rice quality evaluation are the chalkiness degree and chalky rate, which determine the rice grade and price to some extent. In the past few decades, significant progress has been made in breeding high-quality rice cultivars with low chalkiness. Nevertheless, decreasing rice chalkiness remains a very important goal pursued by breeders and consumers, since high temperature during the filling period is still a major factor promoting chalky rice grain formation and decreasing the quality of rice due to global warming (Liu et al. 2024; Li et al. 2025).

Chalkiness is a complex quantitative trait in rice. Its formation is related to multiple processes in the starch/protein biosynthesis and metabolic pathways. The Waxy (Wx) gene encodes an amylose enzyme, namely, bound starch synthase I (GBSSI). The deletion of Wx significantly reduced amylose content, resulting in an opaque or powdery endosperm (Mikami et al. 1999; Tu et al. 2025). Upregulation of α-amylase genes caused an increase in the chalkiness degree (Hakata et al. 2012). NF-YB1 directly interacted with three sucrose transporters (SUTs) to regulate sucrose entry into endosperm and chalkiness formation (Bai et al. 2016; Xu et al. 2016). Protein metabolism-related genes PDIL1-1 (Kim et al. 2012), BiP1 (Wakasa et al. 2011), OsVPS9A (Liu et al. 2013), and OsRAB5A (Fukuda et al. 2013), which regulate protein metabolism, also contributed to chalkiness formation in rice. Overexpression of Chalk5 interfered with pH homeostasis of the endosperm membrane transport system, which in turn affected the biosynthesis of protein bodies and created gaps in the storage sites of endosperm, resulting in higher chalkiness (Li et al. 2014; Samonte et al. 2024). Malfunction of the GIF1 and UGPase1 genes influenced the process of cell wall biosynthesis and chalkiness formation (Wang et al. 2008; Woo et al. 2008). In addition, rice perceives environmental signals through phytohormones. An increase in indole-3-acetic acid (IAA), cytokinin (CTK), and gibberellin (GA) content led to an increase in rice chalkiness, whereas brassinosteroid (BR) could reduce chalkiness (Zhang et al. 2020). However, the regulatory mechanisms of rice chalkiness formation are far from clear, and identifying more chalkiness dominators is urgently needed for rice quality improvement.

Glycosylation is catalyzed by glycosyltransferases (GTs), which are widely distributed in plants. GTs belong to a superfamily consisting of 114 families, of which the GT1 family has more members (Mackenzie et al. 1997; Meech et al. 2019). In plants, GT1 commonly uses UDP sugar as a donor to catalyze the glycosylation of receptor molecules, which is known as UGT (Mackenzie et al. 2005). Many acceptors could be catalyzed by UGT, including monosaccharides, oligosaccharides, and polysaccharides, as well as many bioactive molecules, proteins, and nucleic acids (Bowles et al. 2005). Most GT1 members have a conserved PSPG box at the C-terminus that serves as a recognition domain for sugar donors (Shao et al. 2005). Moreover, glycosylation modification is an important approach for balancing phytohormone dynamics during plant growth and development (Zhang et al. 2020; Palaniyandi et al. 2015). Some UGTs mediated glycosylation of phytohormones, promoting the transition between active and inactive forms (Gachon et al. 2005; Jin et al. 2013; Hu et al. 2022; Thompson et al. 2017; Tognetti et al. 2010; Zhang et al. 2017; He et al. 2023).

In this study, we identified a new glycosyltransferase, OsCFF1, by using comparative transcriptomics. Natural variations of OsCFF1 cause partial fragment missing, frameshift, or early termination of translation in rice cultivars. On the whole, higher expression of OsCFF1 in filling grains resulted in a higher chalkiness degree, and functional loss of OsCFF1 led to a sharp reduction in chalkiness. OsCFF1 mediates rice chalkiness formation, thereby decreasing rice quality. OsCFF1 additionally exhibited glycosyltransfer activity by using GA₄ as the acceptor, and the molecular mechanism deserves in-depth investigations. Accordingly, this study not only found that natural variation and expression of OsCFF1 affected the formation of rice chalkiness, but also contributed to the development of molecular markers and high-quality rice breeding.

Materials and Methods

Plant Materials and Growth Conditions

The rice materials Xiangzaoxian 24 (X24) with high chalkiness and its female parent Xiangzaoxian 11 (X11) with high chalkiness and its male parent Xiangzaoxian 7 (X7) with low chalkiness were provided by the Hunan Rice Germplasm Bank. In addition, we collected rice cultivars (Tab. S1) to determine the chalkiness, and detect the natural mutations and gene expression of OsCFF1 in filling grains. The rice cultivars were cultivated in a field (Changsha, Hunan Province, China, 28° 11′ 49″ N, 112° 58′ 42″ E) and each cultivar included 20 plants. Gene-edited and overexpressed plants were cultured in pots placed in a net house under natural conditions. 5 pots were used for each material.

Chalkiness Determination in 2D (Area/Area, A/A)

Thirty milled polished rice kernels were placed at intervals on a dark surface, and the image was captured using a scanner or camera in supported formats such as JPG, BMP, and PNG. The image was imported into Chalkiness 2.0 software (Chen et al. 2011), which was previously developed by our team. The software would automatically recognize each grain contour area (circled with a red line) and the chalky area (circled with a white line). After clicking the "Calculate" button, chalkiness-related parameters, including total grain number, relative chalkiness area, chalkiness degree and chalkiness rice rate were displayed. The test report was automatically generated into an Excel format file.

Chalkiness Determination in 3D (Volume/Volume, V/V)

Grain scanning by micro-CT and chalkiness determination were conducted according to the method described by Su and Xiao (Su et al. 2020). To ascertain the volume of rice grain and its chalkiness (the white opaque part of the seed), 30 grains array was embedded in super-light clay and loaded into a sample tube of SCANCO μCT 100 (Switzerland) with a cabinet cone-beam, sealed, air-cooled X-ray source with 5–30 µm spot size, and 1.25–200 µm nominal isotropic of resolution. X-ray CT scan parameters were set at 10 μm of spatial-resolution, 50 kV of voltage, and 120 μA of electric current. The signals of transmitted X-ray were captured and imaged by an X-ray camera. The cross-section images were exported as DCM format files containing full and precise spatial-information of samples. Micro-CT integration tools and Materialise Mimics Innovation Suite Research (V 19.0) were employed to segment rice chalkiness area in cross-sections. The precise size- parameters (including volume) were displayed after 3D reconstruction.

Starch Granule Analysis by Scanning Electron Microscope (SEM)

Five mature rice grains for each group with uniform appearance were collected and manually dehulled. Dehulled kernels were cut across the middle parts (containing chalkiness) with a razor blade. The cut surface of the middle part was coated with gold by an ion sputtering device (JFC-1600, JEOL, Tokyo, Japan). Starch granule morphology in the central and peripheral regions of the endosperm was examined using a scanning electron microscope (JSM-6380LV, Tokyo, Japan) at an accelerating voltage of 15 kV. The operation procedures were carried out according to the manufacturer’s protocol.

Bioinformatic Analysis

Bioinformatic analysis of the OsCFF1 was conducted by using online tools, including evolutionary tree analysis (https://CAZy.org), multiple gene alignment (Clustal Omega), secondary structure (https://npsa.lyon.inserm.fr/cgi-bin/npsa_automat.pl?page=/NPSA/npsa_sopma.html), and tertiary structure prediction analysis (https://swissmodel.expasy.org/interactive). Plant-PLoc was employed to predict plant protein subcellular location (http://www.csbio.sjtu.edu.cn/bioinf/plant/). Protein structure of OsCFF1 was predicted by Alphafold3 (https://alphafoldserver.com) (Malhotra et al. 2025).

Gene Cloning

Genomic DNA from each rice cultivar was extracted and purified from fresh rice seedlings using the CTAB method (Zhang et al. 2013). We designed two pairs of primers (5'-AAAGGACGAGCTATTCCACC-3' and 5'-GATACAATTGGGAAATTCCATACC-3'; 5'-AAGAAGACGGTGGTTCTCTAC-3' and 5'CTCATGTCTGAATTGCTCAACGTG-3') to amplify the OsCFF1 gene (Os09g0482860) in each cultivar. The PCR reaction program was as follows: 95 °C for 5 min, followed by 35 cycles of 95 °C for 30 s, 55 °C for 30 s, and 72 °C for 2 min, then incubated at 72 °C for 5 min. Finally, the PCR products were sequenced and aligned.

Real-Time PCR

Ten plant materials were collected for RT-PCR. The tissues/organs of X7 and X11 were gathered at different growth stages (seedling stage, tillering stage, 4 days after flowering and 16 days after flowering), and filling grains of other cultivars were gathered at 4 DAF. Three biological replicates were performed for each cultivar (n = 10). Total RNA was extracted from 100 mg of fresh samples using Trizol reagent (Invitrogen). gDNA was removed, and the purified RNA was reverse-transcribed to cDNA using poly (dT) oligo primers according to the manufacturer's instructions for the M-MLV kit (Invitrogen). The primers used for RT-qPCR were designed as follows: primer for the OsCFF1 gene (forward: 5'- GAAGACGGTGGTTCTCTACC -3', reverse: 5'- CGTGGTCCAGGAAGACGTCG -3') and primer for the actin gene (forward: 5'- CTTCATAGGAATGGAAGCTGCGGGTA -3', reverse: 5'- CGACCACCTTGATCTTCATGCTGCTA -3'). RT-qPCR was performed on 96-well plates using the StepOnePlus Real-Time PCR System (Applied Biosystems) with the following cycling conditions: 95 °C for 1 min, followed by 40 cycles of 95 °C for 5 s and 60 °C for 30 s. Relative gene expression levels were quantified using the comparative ΔΔC_t method. The threshold cycle (C_t) for each reaction was determined automatically using the analysis software with default parameters (Applied Biosystems). The specificity of the RT-qPCR reactions was determined using melting curve analysis.

Plasmid Construction and Plant Transformation

Vector was constructed for the CRISPR/Cas9-mediated gene editing. Using the online resource (http://skl.scau.edu.cn/home/), two specific target sites were designed according to the CDS of OsCFF1: GCTTCGTTCTCTCCATGACCGACGG and AGGGACTCCCAGCTTCCTCGCGG. The knockout vector was subsequently transformed into X11 (high chalkiness) using Agrobacterium-mediated transformation. To construct the overexpression vector, the CDS sequence of OsCFF1 was inserted into pBWA (V) HU with the Ubi promoter. The overexpression vector was transformed into X7 (low chalkiness) using Agrobacterium-mediated transformation. An expression vector for the GFP fusion protein promoted by the 35S promoter was constructed. The pEGAD vector was used as the backbone, the CDS of OsCFF1 was linked to the pEGAD vector through BsaI and Eco31I cleavage enzymes. Tobacco leaves were injected using the Agrobacterium-mediated transformation method, and fluorescence signal detection was performed using laser confocal microscopy to observe the distribution of green fluorescence in the tobacco epidermis. The subcellular localization of OsCFF1 was analyzed through fluorescence.

Transcriptome Analysis and Candidate Gene Screening

Rice grains of X24, X11, and X7 during the filling stage were used for 3-generation high-throughput transcriptome sequencing by using Illumina NovaSeq6000. Total RNA was extracted using a TRIzol® reagent kit according to the manufacturer’s instructions. The quality and quantity of total RNA were estimated using a NanoDrop™ (Thermo Fisher Scientific, Waltham, MA, USA). mRNA was purified using poly-T and oligo-attached magnetic beads. First- and second-strand cDNA were synthesized using random hexamer primers, M-MuLV Reverse Transcriptase (RNase H⁻), and DNA Polymerase I. cDNA fragments of 370–420 bp in length were selected and purified using the AMPure XP system (Beckman Coulter, Brea, CA, USA). PCR amplification was then performed. PCR products were purified using AMPure XP beads, and the library was finally obtained. The library was analyzed by paired-end sequencing (PE150) on the platform according to standard operations, and Fastp was used for quality control (Chen et al. 2018). The original data were first filtered to obtain clean reads and then assembled (Grabherr et al. 2011). These assembled fragments without N-terminal obtained by reading overlap were used as the assembled Unigene, and then use Blast2 GO and KOBAS to obtain Gene Ontology (GO) function and Kyoto Encyclopedia of Genes and Genomes (KEGG) annotation (Conesa et al. 2005; Mao et al. 2005).

Known reference gene sequences and annotation files were used as the database (http://www.mbkbase.org/R498/R498_Chr.fasta.gz). The number of reads mapped to genes in each sample was obtained by using the Htseq-count software (Anders et al. 2015), and fragments per kilobase of transcript per million fragments mapped (FPKM) values for each gene were calculated using Cufflinks (Trapnell et al. 2012). The negative binomial distribution test method in the DESeq software package (Love et al. 2014) was employed to assess the significance of differences in reads, and the Basemean value was utilized to estimate gene expression levels. Genes with a P-value ≤ 0.05 and |Log₂Foldchange (FC)|≥ 1 were identified as significantly differentially expressed genes, i.e., candidate genes.

Gibberellin Determination by LC–MS/MS

According to the method based on magnetic solid-phase extraction and in situ derivatization (Luo et al. 2021), phytohormone was extracted from filling rice grains. About 100 mg of fresh rice grains was collected into a 2 mL centrifuge tube. After being frozen in liquid nitrogen, the grains were ground into powder with one cold ZrO₂ mill ball by ball milling (TissueLyser II, QIAGEN, Germany) at 20 Hz for 4 min. Isotope-labeled standard [²H₂]GA₄ (Olomouc, Czech Republic) was added to a 2 mL centrifuge tube, and the tube was held for 30 min to allow the standard substance to integrate well into the rice grain powder. 200 μL of cold methanol was added into sample tube, and the phytohormone was extracted at 4 °C for 12 h after vortexing and mixing well. Supernatant was transferred to another 2 mL centrifuge tube after centrifugation at 12,000 g for 10 min at 4 °C. The supernatant was added with 5 mg magnetic nanoparticles and 50 μL of derivatizing agent (20 mM EDC), then incubated at 40 ◦C for 90 min. Both the extraction and derivatization were completed in one step. The nanoparticles were collected from the suspension by magnetic separation and re-dispersed by 50 μL desorption solution. After desorption for 20 min, the eluent was collected and determined by LC-MS/MS.

LC-MS/MS was equipped with an Agilent 1290 series (Agilent Technologies, USA) and an Agilent 6460 triple quadrupole mass spectrometer (Agilent Technologies, USA). The analytes were separated on a Waters ACQUITY UPLC HSS T3 column (100 mm × 2.1 μm i.d., 1.8 μm, Waters, Milford, MA, USA). The column oven temperature was kept at 40 °C, and the sample injection volume was 10 μL, the flow rate of the mobile phase was 0.3 mL/min. The elution gradient program of the positive ion mode was performed. The multiple reaction monitoring (MRM) was introduced for determining targeted compounds. The drying and collision gas was nitrogen. The ionization source conditions were as follows: the flow rate of the nebulizer gas was 8 L/min, the source temperature of the mass spectrometer was 300 °C, the nebulizer pressure was 50 psi, and the capillary voltage was 3500 V.

Heterologous Protein Expression and Purification

A heterologous expression vector for His-labeled protein was constructed by using standard molecular biology techniques. pFolia40108 vector was used as the backbone, and the CDS of OsCFF1 was integrated into pFolia40108. Recombinant plasmid pFolia-OsCFF1 was transformed into the Agrobacterium strain C58C1. Agrobacterium suspension was infiltrated into the abaxial surface of tobacco leaves by using a needleless syringe. After 3–5 days, total protein in infiltrated tobacco leaves was extracted using a lysis buffer with protease inhibitors. The expression level of his-labeled recombinant OsCFF1 was evaluated by Western blotting with anti-His/HRP antibodies. The recombinant OsCFF1 with his-tag was purified via Ni–NTA chromatography.

Enzymatic Activity Assays

Each enzymatic reaction mixture (20 μL in total) contained reaction buffer (Tris–HCl 50 mM, pH 7.5, 10% glycerol [v/v], and 10 mM β-mercaptoethanol), donor substrate (250 mM UDP-glucose), acceptor substrate (1 mM), and recombinant protein (2 μg OsCFF1 per reaction) in the initial screening, referring to previous methods (Song et al. 2016; Sheikh et al. 2017; Jing et al. 2019). The reaction mixture was pipetted into a 48-well assay plate (Corning) and incubated at 30 °C for 30 min. The reaction was stopped by mixing the reaction solution with the same volume of UDP-Glo™ assay reagent, and free UDP was detected by using the UDP-Glo™ Glycosyltransferase Assay Kit (Promega). Data were fitted to the Michaelis–Menten equation using a non‐linear regression program (Enzyme Kinetics, SigmaPlot) to calculate V_max and K_m values. Protein concentration was determined using the protein assay dye reagent (Bio-Rad) with bovine serum albumin as a standard. Three independent biological replicates were carried out in each test.

Results

X24 and its Parents Displayed Significantly Different Chalkiness Degrees Because of the Expression Level of OsCFF1

In our previous study, we found that an early indica rice cultivar, Xiangzaoxian 11(X11), showed different levels of chalkiness compared to its parents (X24 and X7). The mature rice grains of X11 and X24 showed a 100% chalky rate with more than 40% and 38% chalkiness degree, respectively, whereas X7 only had a 5% chalky rate and 0.13% chalkiness degree (Fig. 1A and B) (Xie et al. 2021). To reveal the reason for this, we further screened the gene expression data based on transcriptome sequencing in filling grains at 8, 12, and 16 days after flowering (DAF), and identified a UGT gene, OsCFF1, that was significantly correlated with chalkiness degree. The mRNA levels (FPKM) of OsCFF1 in X7 were significantly lower than those in X11 and X24 during the grain-filling period (Fig. 1C). The full-length coding sequence (CDS) of OsCFF1 is 1410 bp. OsCFF1 encodes a glycosyltransferase that contains 469 amino acid residues and belongs to the GT1 family (Fig. 1D). OsCFF1 also has a conserved PSPG box, known as the catalytic domain, located between 345 and 388 aa (Fig. 1E).

Fig. 1.

Identification of OsCFF1 through comparative transcriptomics. A Grains of X7 and its parents (X24 and X7); B X7 and its parents (X24 and X7) show different chalkiness; C Expression level of OsCFF1 in filling grains shows a significant difference; D OsCFF1 is a member of the glycosyltransferase family according to the phylogenetic analysis; E OsCFF1 owns the conserved PSPG box of GTs. Data are presented as mean ± SD of three independent biological replicates (n = 30)

OsCFF1 Displays Natural Variation in Rice Cultivars

We collected nearly 60 rice cultivars from the rice germplasm resource bank and cloned the full-length of OsCFF1. According to the alignment of homologous genes, OsCFF1 mainly contained four genotypes (Fig. 2A and Table 1). Most cultivars are classified as genotype I (WT). The others represented three types of natural variations (Table 1). Eight rice cultivars belong to Genotype II. Genotype II involved in a simultaneous mutation of five single nucleic acids (T²³⁶ to C, A⁴⁴⁵ to G, C⁶¹⁸ to G, G⁸¹⁴ to T and G⁸³⁵ → A). The mutation from G⁸¹⁴ to T resulted in the termination of translation and the loss of the PSPG box (Fig. 2A and Table 1). This shortened protein would lose glycosyltransferase activity. Genotype III contained a missing large segment from sites 159 to 335, resulting large fragment loss and a frame shift. In addition, a single nucleotide mutation in OsCFF1 was also screened in Yuzhenxiang (Fig. 2A and Table 1).

Fig. 2.

Connection between genotypes of OsCFF1 and rice chalkiness in cultivars. A Genotypes of OsCFF1 in screened cultivars. B Rice cultivars containing genotype II variation show a low chalkiness phenotype; B The expression of OsCFF1 was highly positively correlated with the rice chalkiness degree in rice cultivars containing the full-length of OsCFF1. Pearson Correlation Coefficient (R²) = 0.8372, P-value < 0.01. n = 10 effective panicles per cultivar

Table 1.

Natural variations in OsCFF1

Genotypes	Natural variations	Consequence	Cultivars
*II	T236 → C (Val → Ala), A445 → G (Thr → Ala), C618 → G (Asn → Lys), G814 → T (Gly → stop codon), G835 → A	early termination and PSPG box loss	Zhengdao No.5, Nipponbare, Zhonghual1, IR661- 1 Aidaogu 151, Chengnongshuijing, Laoguangtou83, etc
III	nucleotide loss from 159 to 335	Frameshift and PSPG box loss	Xiangnuo and Biwusheng (glutinous rice), etc
IV	T937 → G (Val → Gly), etc	SNP	Yuzhenxiang, etc

^*Genotype I is defined as wild type, since most cultivars contain this full-sequence

Variations in OsCFF1 are Tightly Related to Rice Chalkiness

To analyze the correlation between OsCFF1 and rice chalkiness, we screened the expression levels of OsCFF1 in filling grains and the chalkiness degree in the collected rice cultivars. Genotype II includes five single-base mutations. The fourth mutation (G⁸³⁵ → A) resulted in early termination of translation and the absence of the PSPG box, leading to functional loss (Fig. 2A). Cultivars containing genotype II variation (such as Zhonghua11, Nipponbare, and Zhenshan97B) showed a low chalkiness phenotype (Fig. 2B). Genotype III variation missed a long fragment (from 159 to 335). We screened genotype III only in two glutinous rice cultivars (Xiangnuo and Biwusheng). Conversely, these two cultivars revealed a relatively high chalkiness degree for other reasons. Glutinous rice itself has high chalkiness content because of starch components. Most cultivars had a full-length OsCFF1, genotype I (WT). The mRNA levels in the filling grains were detected using real-time PCR. The results showed that the expression was highly positively correlated with the rice chalkiness degree (Fig. 2C). Therefore, OsCFF1 is likely an important decisive factor involved in rice chalkiness formation.

Expression Pattern of OsCFF1

OsCFF1 is a glycosyltransferase belonging to the GT1 family and has a typical PSPG domain (Fig. 1D and E). We further detected the spatiotemporal expression pattern of OsCFF1 in different tissues/organs of X7 and X11 using real-time PCR. In seedlings and tillering rice, the expression in leaves was higher than that in roots and sheaths, and the mRNA level in X7 was significantly lower than that in X11 (Fig. 3A and B). After flowering, higher gene expression was observed in the stems and leaves (Fig. 3C and D). In the early filling stage (4 DAF), the expression of OsCFF1 in X11 grains was significantly higher than that in X7 grains (Fig. 3C), but the difference in expression gradually decreased along with the filling process (Fig. 3D). Throughout the growth and development periods, the expression remained low in the sheath (Fig. 3A–D). OsCFF1 is predicted to be located on the cell membrane by Plant-PLoc, without a transmembrane domain and signal peptide. Using AlphaFold3, the three-dimensional structure of OsCFF1 was predicted, and asparagine (¹⁷N), glutamate (²¹E), arginine (⁵⁵R), serine (⁵⁹S), and aspartate (²⁵¹D) were predicted to form an active pocket (Fig. 3E). In addition, we constructed an OsCFF1-GFP fusion expression vector and used it to perform transient expression in tobacco leaf mesophyll cells. Fluorescence observations showed that OsCFF1 is a membrane-localized protein (Fig. 3F).

Fig. 3.

Expression pattern of OsCFF1 in rice. Relative expression of OsCFF1 in seedlings (A) and tillering rice (B); Tissue-specific expression of OsCFF1 in 4 DAF (C) and 16 DAF (D); E 3D structure of OsCFF1 generated by Alphafold3; F Fluorescence observation of OsCFF1-GFP fusion protein in tobacco leaf mesophyll cells. Data are presented as mean ± SD of three independent biological replicates (n = 10)

Overexpression of OsCFF1 Results in Higher Chalkiness

To further investigate the biological function of OsCFF1, we constructed an overexpression vector and transformed it into X7 (low chalkiness rice) (Fig. 4A). Compared with the WT, the overexpression materials showed no significant differences in growth and development phenotypes and agronomic traits (Fig. 4A, Fig. S1). We detected the starch packaging in the mature kernel of WT and OE by using SEM. Comparing with WT(X7), the endosperm of OE line carried round loosely packed starch granules with air spaces (Fig. 4B). In OsCFF1 overexpressing rice, the chalkiness degree and chalky rate were significantly increased (Fig. 4C–E). The chalkiness degree in the transformants was up to 22.3% from 5.5% (A/A) of WT(X7), and 15.9% from 3.8% (V/V) of WT(X7) (Fig. 4D and E).

Fig. 4.

Overexpression of OsCFF1 in X7. A Transformant of OsCFF1 overexpression; B Starch granule analysis by using SEM. C Chalkiness determination by micro-CT in X7-OE, the darker areas (red arrow) represent chalkiness; D Chalkiness degree (area/area) in rice kernel; E Chalkiness degree (volume/volume) in rice kernel. Data are presented as mean ± SD of three independent biological replicates (n = 30)

OsCFF1 Deletion Significantly Decreases Rice Chalkiness

We also constructed a knockout vector and transformed it into X11 (high chalkiness) (Fig. 5A and B). Compared with WT, the gene-edited materials showed no significant differences in growth and development phenotypes and agronomic traits (Fig. S1). Using the CRISPR/Cas9 system, two independent knockout lines (ko1 and ko2) were screened (Fig. 5A and B). The starch packaging in the mature kernel was detected by using SEM. The mature kernel in the ko1 plant was filled with large and tightly packed starch granules (Fig. 5C and D). The cross-sections of ko1 rice grains captured by micro-CT showed higher density compared with WT (Fig. 5E). The chalky rate and chalkiness degree in both ko1 and ko2 kernels significantly decreased (Fig. 5F and G). Based on 2D detection, chalkiness (projected area) in the ko1 and ko2 kernels was decreased by 91.3% and 62.5%, respectively. Based on 3D determination (Su and Xiao 2020), chalkiness (volume) in the ko1 and ko2 kernels decreased by 95.8% and 69.9%, respectively (Fig. 5F and G).

Fig. 5.

OsCFF1 knockout decreases chalkiness. A Two targets of gene-editing; B Possible proteins after gene-editing; C Chalkiness in rice kernel; D Starch granule analysis by using SEM. E The cross-section generated by micro-CT, the darker areas (red arrow) represent chalkiness; F Chalkiness degree (area/area) in rice kernel; G Chalkiness degree (volume/volume) rice kernel. H Expression of chalkiness-related genes in filling grains of X7 and X11. I Relative expression of GS9, SPL6, PHO1 and FLO18 in filling grains of X7 and ko1. J Relative expression of GS9, SPL6, PHO1 and FLO18 in filling grains of X11 and OE. Data are presented as mean ± SD of three independent biological replicates (n = 30)

In addition, we analyzed the potential translational consequences of OsCFF1 gene editing. In ko1, G³⁷⁹ was deleted, resulting in a frame-shift mutation (Fig. 5A and B). The mutant nucleic acid sequences contained several small CDSs. The longest predicted protein product also lost more than 120 amino acids from the N-terminus (Fig. 5B). Mutations in ko2 resulted in the loss of 42 amino acids at the N-terminus (Fig. 5B). The lower amino acid removal in ko2 may have partially retained the bioactivity of glycosyltransferases; consequently, ko2 exhibited higher chalkiness levels than ko1. Therefore, we speculate that OsCFF1 regulates the formation of rice chalkiness and that the expression of OsCFF1 determines the level of chalkiness.

The expression profiling of chalkiness-related genes was screened in filling grains of X7 and X11 according to the transcriptome data. The mRNA levels of GS9, SPL6, PHO1 and FLO18 in X11 were significantly higher than that in X7 (Fig. 5H). GS9 is a transcriptional activator to regulate rice grain shape and appearance quality. Introducing the gs9 allele into rice cultivars significantly improves grain shape and appearance quality (Zhao et al. 2018). OsSPL16 controls the grain size, shape and quality in rice (Xie et al. 2021). The loss-of-function mutation in Basmati rice was associated with the formation of a more slender grain and better quality of appearance (Wang et al. 2012). FLO18, encoding a mitochondrion-targeted P-type PPR protein with 15 PPR motifs, plays an important role in mitochondrial function and endosperm development (Yu et al. 2021). Loss-of-function ospho1;2 mutants produced small and shrunken seeds with reduced starch content, resulting in grain chalkiness (Ma et al. 2021). Furthermore, we detected the expression of these four genes in filling grain of different materials through RT-PCR. The results showed that GS9 and SPL16 were down-regulated in filling grains of OsCFF1 knock-out rice (Fig. 5I). Conversely, the two genes were down-regulated in overexpression plants (Fig. 5J). This indicates that OsCFF1 is associated with both GS9 and SPL16 in regulating rice chalkiness formation.

OsCFF1 Affects the Amylose Content and GA₄ Level

We determined the contents of amylose, soluble protein and phytohormones in grains of ko1 and overexpression rice. Comparing with WT, the soluble protein content in filling grains of ko1 was significantly increased, but the OE line showed no significant difference (Fig. S2). Amylose content in OE grains was decreased, and that in ko1 was significantly increased (Fig. 6A and B). Interestingly, GA₄ level in filing grain was sharply increased in OsCFF1 knock-out rice, and the opposite result was obtained in the OsCFF1 overexpression plant (Fig. 6C and D). We employed AlphaFold3 to predict the 3D structure of OsCFF1 and performed molecular docking between OsCFF1 and different phytohormones. The results showed that the binding energy was below -8 kcal/mol between OsCFF1 and GA₄, indicating a stable binding in the active pocket (Fig. 6E). In comparison, the values of binding energy between OsCFF1 and other phytohomones (IAA, SA, CTK, IBA and ABA) were significantly higher than this value (Fig. 6F). To figure out whether OsCFF1 mediates the glycosylation of GA, we heterologously expressed OsCFF1 in tobacco leaves and the recombinant protein was purified through Ni–NTA. Subsequently, the enzymatic activity was assayed by using UDP-glucose as the donor substrate, and phytohormones (GA₄, IAA, SA, CTK, IBA and ABA) as the acceptor substrate. The results showed that OsCFF1 could efficiently catalyze the glycosylation of gibberellin, but the glycosyltransferase activity for other phytohomones is less than 10% of that for GA₄ (Fig. 6G). Recombinant OsCFF1 glucosylates GA₄ with a K_m value of 18.56 ± 3.57 μM and a V_max value of 0.132 ± 0.009 nKat/mg (R² = 0.976). Accordingly, the glycosylation of GA₄ may be mediated by OsCFF1 and leads to the transformation of GA₄ into an inactive form. Of course, related regulatory factors and molecular mechanisms remain to be further analyzed.

Fig. 6.

Enzymatic activity assay of OsCFF1. A, B The contents of amylose in rice grains. C, D The contents of GA₄ in filling grains of different rice materials. E Imaging of OsCFF1 binding GA₄ in silicon. F Evaluating the binding energy between OsCFF1 and different ligand through molecular docking. G Glucosyltransferase activity of OsCFF1 towards different acceptor substrates with UDP-glucose. Data are presented as mean ± SD of three independent biological replicates

Discussion

OsCFF1 is an Important Regulator for Rice Chalkiness Formation

Chalkiness refers to the opaque portions in rice endosperm, which primarily comprise loosely packed starch granules and protein bodies (PBs). As a complex quantitative trait, chalkiness is regulated by multiple genes and environmental factors. Starch synthesis genes, such as Wx (Mikami et al. 1999; Tu et al. 2025), SSIIIa (Fujita et al. 2007), and OsAPL in rice (Zhang et al. 2012), affected the chalkiness formation. Similarly, protein metabolism-related genes, such as PDIL1-1 (Kim et al. 2012), BiP1 (Wakasa et al. 2011; Yasuda et al. 2009), OsVPS9A (Liu et al. 2013), and OsRAB5A in rice (Fukuda et al. 2013), may also contribute to chalkiness. This trait is also correlated with cell wall biosynthesis processes. In rice, the malfunction of the GIF1 (Wang et al. 2008), and UGPase1 genes disrupts cell wall biosynthesis, resulting in the formation of chalky grains (Woo et al. 2008). The interaction between OsNLP3 and OsNF-YA8 accelerates the grain chalkiness formation (Mondal et al. 2024). In contrast, two IAA receptor genes, OsAFB3 and OsAFB5, negatively regulate the formation of chalkiness (Shi et al. 2025). CHALK10, encoding an F-box protein, also suppresses rice chalkiness (Zhou et al. 2025). Additionally, abiotic stresses, including high temperatures (Yamakawa et al. 2007; Maccarrone et al. 2001), humidity levels (Ishimaru et al. 2019), and strong sunlight (Jiang et al. 2002), can promote chalkiness formation during the grain-filling stage. Collectively, rice chalkiness results from genetic and environmental interaction. To date, few quantitative trait loci (QTLs) related to chalkiness have been identified and functionally characterized (Woo et al. 2008). Previous research also identified a few major QTLs that may be responsible for rice chalkiness formation (Tan et al. 2000; Wan et al. 2005).

In this study, we identified a novel glycosyltransferase OsCFF1, which may be a determinant to mediate the formation of rice chalkiness, and thus determine rice quality. OsCFF1 in rice cultivars displays natural variations including SNPs and large deletions. Many natural mutations lead to the loss of the PSPG (putative sugar-binding motif), but these cultivars have not been found to have serious growth and developmental defects. One reason may be gene functional redundancy under specific environmental conditions, since GTs belong to a superfamily consisting of 114 subfamilies (Mackenzie et al. 1997; Meech et al. 2019). OsCFF1 is one member of subfamily GT1, and other genes may partially compensate for the loss of OsCFF1. In addition, a certain member of GT1 usually has receptor preference and a specific function (Zhang et al. 2020, 2017; Palaniyandi et al. 2015; Gachon et al. 2005; Jin et al. 2013; Hu et al. 2022; Thompson et al. 2017; Tognetti et al. 2010; He et al. 2023). We speculate that OsCFF1 may mediate the glycosylation of a certain bioactive molecule to regulate chalkiness formation. The chalkiness degree was positively associated with its expression level. Overexpression of OsCFF1 in the low-chalkiness cultivar X7 significantly increased both the chalkiness degree and chalky rice rate. Conversely, knockout of OsCFF1 in the high-chalkiness cultivar X11 nearly eliminated chalkiness in the grains. Given that glycosyltransferases primarily function through glycosylation, we propose that glycosyltransferase-mediated glycosylation represents a novel regulatory mechanism underlying chalkiness formation in rice. This finding advances our understanding of the molecular mechanisms underlying rice chalkiness formation, although more biochemical and molecular evidences are still required.

OsCFF1 is a Candidate with Great Practical Value in High-Quality Rice Breeding

Rice serves as the staple food crop for over half of the global population, providing nearly 75% of the caloric intake of the population in Southeast Asia (Fitzgerald et al. 2009). Breeding high-quality rice remains a critical objective for both breeders and consumers. A deeper understanding of the factors governing grain quality will enable the development of integrated breeding strategies that simultaneously enhance quality and yield. Elucidating the genetic regulation of high yield and superior cooking/eating quality is essential for targeted rice improvement (Fitzgerald et al. 2009; Cai et al. 2025). For example, the gene WBR7 modulates yield and grain quality in rice through sugar metabolism (Shi et al. 2024). The gene OsSG34 influences grain size and quality through lipid metabolism (Jin et al. 2024). CRISPR/Cas9-based editing of NF-YC4 exhibits up to a 17% increase in seed protein of rice (Wang et al. 2025).

This quality degradation has far-reaching consequences for food systems (Lin et al. 2016), potentially diminishing farmers’ incentives and threatening food security. Specifically, reduced market value decreases grower profitability, making low-chalkiness varieties economically preferable for cultivation purposes. As a key appearance trait, chalkiness significantly affects consumer acceptance of rice. Global rice markets primarily value two characteristics: the chalkiness proportion and the broken grain ratio. High-chalkiness rice faces market rejection owing to impaired processing performance and cooking quality (Xi et al. 2016; Butardo et al. 2019), resulting in price reductions (Laborte et al. 2015). Our study revealed that OsCFF1 mediated the formation of rice chalkiness. OsCFF1 has some natural mutations, and its expression level also differs significantly among rice cultivars. Overall, the mRNA content in filling grains was closely correlated with the level of chalkiness. OsCFF1 in a high-chalkiness cultivar (X11) was knocked out using the CRISPR/Cas9-based editing method, and the chalkiness degree was reduced to less than 5% from over 40%. In addition, the agronomic traits of knockout plants were not significantly different, and grain yield was not reduced. Accordingly, OsCFF1 is a good candidate for developing molecular markers and creating low-chalkiness germplasm, and shows great applicability in high-quality rice breeding.

Conclusion

In summary, this study identified OsCFF1, which codes a glycosyltransferase, as a novel positive regulator of rice chalkiness formation. OsCFF1 contains four types of genotypes among rice cultivars, and the missing mutation of the reserved PSPG box leads to a decrease in chalkiness. The expression level of OsCFF1 during the filling process was related to the chalkiness degree in cultivars to some extent. Overexpression of OsCFF1 caused the chalkiness to increase. Notably, the chalkiness degree in the ko1 and ko2 kernels was decreased by 91.3% and 62.5%, respectively, and chalkiness volume decreased by 95.8% and 69.9%, respectively, indicating that the inferior rice with high chalkiness becomes superior rice with less chalkiness. Phytohormonal profiling in filling rice grains revealed that the content of GA₄ was significantly changed in different rice plants. Recombinant OsCFF1 showed the glycosylation activity for GA₄. OsCFF1 may affect the rice chalkiness formation through regulating the dynamic equilibrium between active-GA and inactive-GA in the filling stage, but this hypothesis still needs more supporting studies and evidence. Overall, this study identified a key factor, OsCFF1, which natural variation and expression affect the formation of rice chalkiness. OsCFF1 also shows a great potential for application in high-quality rice breeding.

Author Contributions

W.L. and L.X. conceived and supervised the study. Q.P., L.M., Q.X., and M.W. performed the experiments and data analysis. D.D. and Y.W. performed the *in silico* analysis. L.M., Q.P and Y.S. prepared the manuscript. W.L., Y.S., and L.X. discussed and revised the manuscript accordingly. L.M. and Q.P. contributed equally to this study. All authors read and approved the manuscript.

Funding

This study was funded by the Natural Science Foundation of Hainan Province (326MS0191) and Scientific Research Foundation of Hainan Tropical Ocean University (No. RHDRCZK202527 and RHDRCZK202532).

Data Availability

The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding authors.

Declarations

Competing Interests

The authors declare no competing interests.