Fine Mapping and Genetic Effect Analysis of Rf21(t) for the Fertility Restoration of Chinsurah-Boro-II-Type Cytoplasmic Male Sterile Oryza sativa (ssp. japonica) Lines

Abstract

The combination of Chinsurah Boro II (BT)-type cytoplasmic male sterility (CMS) and Rf1, the main fertility restorer gene (Rf) for CMS-BT, has been extensively utilized for the production of three-line commercial japonica hybrid seeds. The identification of new Rf genes holds significance for the breeding of BT-type restorer lines, aiming to enhance the heterosis level of BT-type japonica hybrids. In the present study, ‘02428’, a wide-compatibility japonica variety, was observed to partially restore fertility to BT-type CMS lines. Genetic analysis revealed that ‘02428’ carries a dominant Rf gene, Rf21(t), responsible for the fertility restoration of BT-type CMS lines. Leveraging bulked segregant analysis (BSA) resequencing technology and molecular markers, the Rf21(t) locus was identified, and mapped within a candidate interval of 6–12.5 Mb on chromosome 2. Using the iso-cytoplasmic restorer populations, Rf21(t) was ultimately mapped to an interval of approximately 77 kb, encompassing 12 predicted genes, including LOC_Os02g17360, encoding a PPR-domain-containing protein and LOC_Os02g17380 (Rf2), a cloned Rf for Lead-rice-type CMS. A comparative sequence analysis, gene expression profiling and gene knockout experiments confirmed that LOC_Os02g17360 and LOC_Os02g17380 are the most likely candidates of Rf21(t). Furthermore, Rf21(t) showed the dosage effect on the fertility restoration of BT-type CMS lines. This newly identified Rf21(t) represents a valuable genetic resource for the breeding of BT-type japonica restorer lines. Our findings offer practical insights for breeders interested in advancing BT-type japonica hybrid development.

1. Introduction

Cytoplasmic male sterility (CMS) is a maternally inherited trait that results in the inability to produce functional pollen, with fertility restoration achieved by nuclear-encoded genes known as fertility restorer genes (Rfs) [1]. The CMS/Rf system has been instrumental in elucidating the interactions between mitochondrial and nuclear genomes in plants and is extensively utilized for hybrid seed production [2,3,4]. Rice (Oryza Sativa L.) is a typical self-pollinating crop, so CMS is essential for utilizing heterosis. Three-line hybrid rice, produced via the CMS/Rf system, yields more than inbred varieties by 15–30% [5,6]. Despite these advancements, japonica hybrids occupy a relatively small fraction (less than 3%) of the total planting area of japonica varieties in China [7]. Consequently, the development of three-line hybrid japonica varieties holds significant potential to increase rice production in the future.

To date, more than ten types of CMS systems have been identified in rice, including CMS-BT, CMS-HL, CMS-LD, CMS-CW, CMS-WA, CMS-ID, CMS-RT102, CMS-RT98, CMS-D1, CMS-TA and CMS-FA [8,9]. CMS-BT (Chinsurah Boro II/Tai Chung 65) serves as the predominant CMS used for generating three-line japonica hybrids, while CMS-WA and CMS-HL are representative CMS types used for commercial indica hybrid seed production [6,10,11,12]. The BT-type cytoplasm originates from Oryza sativa ssp. indica Chinsurah Boro II, and the BT-type CMS is caused by orf79 in the mitochondria [13]. The restoration of pollen fertility is governed by Rf1, a fertility restorer gene located on chromosome 10, operating in a gametophytic manner [14]. In China, C57 was the first japonica restorer line breeding with Rf1 introduced from IR8, and many japonica restorer lines were developed from C57 [15]. Due to the widespread use of the Rf1 allele, the genetic diversity of BT-type restorer lines has been reduced significantly, which has been an obstacle to developing BT-type japonica hybrids. Also, the fertility levels of F₁ hybrids would be higher while the restorers carry multiple restorer gene loci, especially in the gametophytic CMS/Rf systems [16,17]. Therefore, the identification of novel Rf genes for breeding BT-type japonica fertility restorers could enhance hybrid vigor and facilitate better combinations among hybrid japonica parents, thereby expediting the advancement and adoption of BT-type japonica hybrids. To date, several genetic loci related to the restoration of fertility in different types of CMS lines have been identified in rice. Three fertility restorer genes for CMS-WA, namely Rf3, Rf4 and Rf20, have been mapped to chromosomes 1, 10 and 1, respectively [18,19,20]. Rf5 and Rf6 are the two major fertility restorer genes for CMS-HL and they have been mapped to chromosomes 10 and 8, respectively [16,21,22]. Rf17 (chromosome 4) and Rf2 (chromosome 2) are the fertility restorer genes for Chinese-wild-rice-type CMS [23] and Lead-rice-type CMS [24], respectively. Recently, OsRf19 for CMS-FA rice has been mapped to chromosome 10 [9]. Among these mapped genes, most Rfs have been cloned and characterized as organelle-targeted pentatricopeptide repeat (PPR) proteins, such as Rf1a and Rf1b in CMS-BT rice [13,25,26,27], Rf5 and Rf6 in CMS-HL rice [21,22], Rf4 in CMS-WA rice [19] and OsRf19 in CMS-FA rice [9]. In most cases, a single Rf gene can only restore the fertility of one type of CMS line. Thus, the Rf genes for the fertility restoration to one given CMS are still rare.

The variety ‘02428’ is a widely used compatibility japonica variety in breeding. In this study, we found that ‘02428’ exhibited a partial ability to restore fertility to BT-type CMS lines. To identify Rf genes in ‘02428’, genetic analysis was performed, and the target gene Rf21(t) was mapped using a method that integrates bulk segregant analysis (BSA) with conventional map-based cloning. The candidates for Rf21(t) were analyzed, and its ability to restore fertility to BT-type japonica CMS lines was evaluated. Our results not only lay the foundation for gene cloning but also provide valuable insights for breeding BT-type japonica restorer lines, thereby facilitating the development of three-line japonica hybrids.

2. Materials and Methods

2.1. Plant Materials

The variety ‘02428’, derived from the Jiangsu Academy of Agricultural Sciences, is a japonica cultivar known for its wide compatibility, primarily used as a male parent in breeding. In our study, we employed three pairs of BT-type japonica CMS lines and their corresponding maintainer lines developed by Yangzhou University, specifically ① BT-9201A and 9201B, ② BT-LiuqianxinA (BT-LqxA) and LiuqianxinB (LqxB), and ③ BT-NipponbareA (BT-NipA) and NipponbareB (NIP). However, two additional BT-type CMS lines were employed: ① BT-SuqiuA (BT-SqA) and ② BT-LingfengA (BT-LfA). For genetic analysis, BT-9201A and 9201B, and BT-LqxA and LqxB were used as female lines to cross with ‘02428’, and two pairs of F₁s (① BT-9201A/02428 and 9201B/02428; ② BT-LqxA/02428 and LqxB/02428) were generated in 2019. To reveal the genetic basis of fertility restoration, a three-way cross, BT-9201A//9201B/02428, was performed to develop a population consisting of 691 plants in 2020. To precisely map the target gene Rf21(t), we developed the BT-9201A//9201B/02428 F_2–5 populations from plants harboring the heterozygous genotype at the Rf21(t) locus from 2020 to 2023. To assess the effects of Rf21(t) on fertility restoration, NIL-Rf21(t), the near-isogenic line for Rf21(t), was developed by successive backcrossing (BC₅F_n) between NIP and ‘02428’ with marker-assisted selection in 2022. An F₂ population was developed from the cross of BT-NipA and NIL-Rf21(t), and at least 10 plants harboring the homozygous or heterozygous genotypes were identified. T8819, a stable iso-cytoplasmic line with a homozygous genotype, was obtained from the BT-9201A//9201B/02428 F₅ population and it exhibited a natural spikelet level of more than 75%.

T8819, a stable iso-cytoplasmic line with a homozygous genotype, was obtained from the BT-9201A//9201B/02428 F₅ population and exhibited a natural spikelet level of more than 75%. The Rf knockout (KO) mutants in the background of T8819 were generated using the CRISPR/Cas9 system.

All plant materials were cultivated at the experimental field of Yangzhou University in Yangzhou, Jiangsu, China, employing standard rice cultivation practices. Each field plot comprised four rows spaced 30 cm apart, with ten plants in each row spaced 13.5 cm apart.

2.2. Fertility Scoring and Genetic Analysis

Pollen fertility levels, as well as bagged and natural spikelet fertility levels, were assessed for five plants from the CMS lines and each F₁ line, along with each plant in the BT-9201A//9201B/02428 F₁ population in 2020. The natural spikelet fertility levels of 10 plants in the BT-NipA/NIL-Rf21(t) F₂ population and T8819 were assessed in 2023. Pollen grains were stained using a 1% w/v iodine/potassium iodide solution (I₂-KI) and examined under an optical microscope. Additionally, two panicles were bagged before flowering for subsequent evaluation. We counted the filled and unfilled grains of two panicles from one plant harvested 30 days after flowering, and the spikelet fertility level of one plant was measured as the average seed-setting rate of two panicles.

For the genetic analysis, the bagged spikelet fertility was used as the main criterion for the evaluation of sterile and fertile plants. Plants with <1% seed setting on bagged panicles were categorized into the sterile class, and others with >1% seed setting on bagged panicles were regarded as fertile in the three-cross F₁ population. A chi-square test (χ²) was used to evaluate the goodness of fit of observed and expected segregation ratios in the populations.

2.3. Specific-Locus Amplified Fragment (SLAF) Library Construction and Sequencing

Two phenotypically contrasting bulk DNAs, one sterile and one fertile, were formed by mixing the DNA from 35 corresponding plants of the three-way-cross F₁ population in 2020. An improved SLAF sequencing (SLAF-seq) strategy was utilized in our experiment, as described by Sun et al. [28], with small modifications. For the two DNA bulk samples and two parental samples, the HaeIII restriction enzyme (New England Biolabs, Ipswich, MA [NEB], USA) was used to digest the genomic DNA. A single nucleotide (A) overhang was added subsequently to the digested fragments using the Klenow Fragment (3′→5′exo^–) (NEB) and dATP at 37 °C, and the resulting fragments were ligated to sequencing connectors. Fragments ranging from 264 to 314 base pairs (with indexes and adaptors) in size, isolated by a gel-based method, were selected for pair-end sequencing (each end was 80 bp) performed on an Illumina HiSeq 2500 system (Illumina Inc, San Diego, CA, USA) according to the manufacturer’s recommendations at the Biomarker Technologies Corporation (Beijing, China).

2.4. Sequence Data and Linkage Analyses

SLAF marker identification and genotyping were constructed following established protocols described by Sun et al. [28]. Subsequently, SLAF-seq data analysis was conducted according to Abe et al. [29] and Hill et al. [30]. To precisely identify the region in which Rf lies, we used the χ² test to evaluate the goodness of fit of segregation ratios. Specifically, we compared the ratios of Pab:Mab = 99:1 and Paa:Maa = 50:50, where P stands for 9201B (9201A), M stands for ‘02428’, aa denotes the fertile pool, and ab denotes the sterile pool. Maa and Paa indicate the depths of the aa population derived from M and P, respectively. Mab and Pab indicate the depths of the ab population derived from M and P, respectively.

2.5. Development of the Knockout Transgenic Lines

For validation of the candidate gene of Rf21(t), we used CRISPR/Cas9 genome editing to generate a mutant allele for LOC_Os02g17360 in wild-type T8819 in 2023. The coding sequence (AGCTGGCGCTAAAGGCGCGACGG) was chosen as the target sequence and subsequently cloned into a binary plasmid (pYLCRISPR/Cas9 Pubi-MH). The resulting construct was then delivered into T8819 via Agrobacterium-tumefaciens-mediated transformation. The sequence variation in the corresponding individual KO plants was analyzed by sequencing. Each T₀ mutant self-pollinated to produce a T₁ population. Eight plants from each T₁ population were used to verify the sequence variation by sequencing.

2.6. DNA Extraction, Polymerase Chain Reaction and Sequencing

Genomic DNA was isolated from fresh leaves using the cetyltrimethylammonium bromide (CTAB) method [31] and the Genomic DNA Extraction Kit was purchased from Thermo Fisher Company (Suzhou, China). Briefly, the sample was ground and placed into a 2 mL microcentrifuge tube. A total of 650 μL of CTAB was added and this was incubated at 65 °C for 30 min. Subsequently, an equal volume of chloroform was added and centrifuged at 10,000 rpm for 10 min. The supernatant was collected, an equal volume of isopropanol was added, and incubation took place at −20 °C for 1 h. Finally, centrifugation was carried out to obtain the pellet, which was the DNA. Simple sequence repeat markers (SSR) were sourced from the Gramene database (http://www.gramene.org/, accessed on 4 June 2025), while new insertion/deletion markers (InDel) for the NIP (japonica) and ‘9311’ (indica) genome sequences were identified using the BLAST online tool (http://www.ncbi.nlm.nih.gov/, accessed on 4 June 2025). And the corresponding databases we used are Oryza sativa Japonica Group (taxid:39947) and Oryza sativa Indica Group (taxid:39946). Primers were synthesized by Tsingke Biotech Co. (Nanjing, China). Molecular marker analyses were conducted in a 1× reaction buffer containing 0.1 mM dNTP, 1.0 U Taq polymerase, 0.2 µM of primer and 20 ng of the template DNA, with the final volume adjusted to 20 µL with ultra-pure water. The polymerase chain reaction (PCR) program consisted of an initial denaturation step at 94 °C for 4 min; this was followed by 30 cycles of 94 °C for 40 s, annealing at 53 °C for 40 s, and extension at 72 °C for 50 s; the final extension step took place at 72 °C for 5 min. Amplification products were separated by 3% (w/v) agarose gel electrophoresis, stained with ethidium bromide and visualized using a Gel Doc 1000 system (Bio-Rad Laboratories, Hercules, CA, USA).

According to the mapping results, DNA fragments corresponding to candidate genes in the mapped region were amplified from 02428 and NIP genomic DNA using KOD Plus high-fidelity DNA polymerase (Toyobo, Osaka, Japan), and sequenced by Tsingke Biotech Co. (Nanjing, China). Sequence alignment was conducted using BLAST tools provided by the National Center for Biotechnology Information. The primers utilized for gene mapping and candidate sequencing are detailed in

Table A1. Primer sequences used in this study.

Marker	Forward Primer (5′-3′)	Reverse Primer (5′-3′)
RM12906	CATGCTCTTCGTCACACCATCG	GTGTGATGTGGATAATTGGCATCC
RM12914	CGGAGGGAGTAGTGGTGTATTGG	CACAATGGAGGAGCAGATGAGG
RM6639	ACCGGAAGGGATACTTATCAGC	CTTCCTGTGAAATAGTAGAGGTAGGC
RM12921	TCGTATTTCCCGGTGTCTCAGG	ACTAGTACTCGGTGCAGGGAATCG
RM6375	CGAATGGAACGACAAGAGATGC	ATAGATCGACAACGTAGATCCACAGG
RM12941	TTATGCCATGTGGTCCAATCAGC	ATTTGAACCATTTGGGCCTTGG
RM12938	CCATCTGTCTCTCCTCCTTATCTCC	CTCCCACCAAATCTAGTGAATGG
STS-2-39.6	TGCTGATGTTAATCTGTGGGTG	TGGCGGCTTGAGAGTGTTTGTAG
STS-2-55.4	TGACTATGTAAGGTTGCTGCTG	ACAGCTCCACAGCAAAGAACC
RM1347	AACAAATTAAACTGCCAAG	GTCTTATCATCAGAACTGGA
STS2-31	TTAACTGGTAACGTGAAT	CAGCAAGGCAACAACAAT
STS2-39	TGCTGATGTTAATCTGTGGGTG	TGGCGGCTTGAGAGTGTTTGTAG
STS2-101	ATGGGTTCAGACCTTAGGCG	AGGGAAGGGTTTATGTTGTTCT
STS2-12	AAGTCAAGCTCCTGTAAGATTC	TGTTGATGTGATGAAAAAAAGT
STS2-13	TAGCGTTTTACGGAGAGATTTT	GATTAGGAGTGTGACGTGGACT
STS2-20	CATGTATGACAAAAACAGAGCC	TTTAGTAGTTTGAAAAGCGTGC
STS2-21	CTTCTTCTTCGTTGCGATT	CGTGTGGAGGTTAGGCTGT
RM1358	GATCGATGCAGCAGCATATG	ACGTGTGGCTGCTTTTGC
17360ce-1	CGGTGTTCAATGTTCACTTGTT	ATGCCTCCTCCTTGCTCAG
17360ce-2	TGAGCAAGGAGGAGGCATT	AGTTCTTACAGCGTGTGAATCA
17360ce-3	CCAGTGAGATGTCGGTTATGC	AGACCTGAAGAGCCTGAGTTG
17360ce-4	TCTCCAACTCAGGCTCTTCAG	TTCGTTCCTCCTCCTCATCTC
OsActin	GGAAGTACAGTGTCTGGATTGGAG	TCTTGGCTTAGCATTCTTGGGT
qRT-02g17360	AATGCCTCCTCCTTGCTCAG	GTCCTTCTCTAGTGAACCTTCCT
KO-Identified 12159	TCGGCAAAAGAAACGAAAAG	GGGGTGGAGGAGGAGATGGG
GP15924-12323	TACTCTGGCATCTCCCCGAA	CAGCCACCACCTCATCATCC

in the Appendix A.

2.7. Quantitative Real-Time (qRT) PCR

The expression patterns of candidate genes in BT-NipA and T8819 were conducted by qRT-PCR in 2023 (BIO-RAD, CFX Connect Real-Time System). Total RNA was extracted from the youth panicles (10–15 cm) using the total RNA Isolation Kit (Vazyme, Nanjing, China) and this was reverse transcribed into cDNA using the Fast Quant RT Kit (with gDNase) (Vazyme, Nanjing, China). The amplification conditions included one cycle at 95 °C for 5 min, followed by 40 cycles at 95 °C for 15 s, 60 °C for 30 s, then 95 °C for 15 s and 60 °C for 1 min, with a melting curve from 60 °C to 95 °C. The expression of the candidate genes was determined relative to BT-NipA and T8819 using the 2^−ΔΔCt method [32]. The primer sequences are provided in Table A1.

2.8. Data Analysis

Statistical analysis was conducted using the analysis of variance (ANOVA) procedure package in SPSS 15.0.

3. Results

3.1. Evaluation of the Fertility Restoration Capability of ‘02428’

We performed I₂–KI staining to analyze the pollens of ‘02428’, different CMS lines, and the testcross F₁s. Similar to 02428, the CMS lines and the testcross F₁ plants all exhibited stainable pollen grains. There were no differences in pollens between the CMS lines and the testcross F₁s (Figure 1A,B,D,E), or between A/02428 and B/02428 F₁s (Figure 1B,C,E,F). BT-9201A and BT-LqxA showed a low natural spikelet fertility level and no bagged spikelet fertility, and they were completely sterile. The B/02428 F₁ plants exhibited a natural spikelet fertility level of >65% and a bagged spikelet fertility of >30% (

Table 1. Summary of testcross experiments.

Lines	Cytoplasm	Phenotype	Bagged Spikelet Fertility (%)	Natural Spikelet Fertility (%)
BT-9201A	BT	Sterile	0	3.15
9201B	Normal	Fertile	54.74	89.18
BT-9201A/02428 F1	BT	Fertile	12.24	41.15
9201B/02428 F1	Normal	Fertile	34.76	65.33
BT-LiuqianxinA	BT	Sterile	0	2.27
LiuqianxinB	Normal	Fertile	63.22	96.12
BT-LiuqianxinA/02428 F1	BT	Fertile	28.65	61.94
LiuqianxinB/02428 F1	Normal	Fertile	50.96	92.33
02428	Normal	Fertile	57.81	94.29

). Accordingly, the testcross F₁ plants from the CMS lines and ‘02428’ showed a natural spikelet fertility level of >40% and a bagged spikelet fertility level of >10% (Table 1). The natural spikelet fertility levels of F₁ plants from the crosses BT-9201A/02428 and 9201B/02428, and BT-LqxA/02428 and LqxB/02428 showed statistical differences. These results indicated that the restorer ‘02428’ partially restored the fertility of BT-type CMS lines.

3.2. Genetic Analysis of Fertility Restoration

The pollen fertilities and spikelet fertilities of 691 plants in the BT-9201A//9201B/02428 population were observed. The pollen grains of all plants in the BT-9201A//9201B/02428 population were stainable, similar to those of the BT-9201A/02428 F₁ plants. Therefore, the bagged spikelet fertility was used as the main criterion for the evaluation of fertile and sterile plants in our study. There were 354 sterile plants and 337 fertile plants in the BT-9201A//9201B/02428 population, yielding a segregation ratio of 1:1 between sterile and fertile plants (χ² = 0.52 < χ²_0.05 = 3.84). These results indicated that there was one dominant gene in ‘02428’ conferring fertility restoration, and this Rf was tentatively named Rf21(t). The natural spikelet levels of the fertile plants ranged from 0.97 to 79.05%, and displayed a continuous distribution along with an average level of 35.74%, which was similar to that of the F₁ plants (Figure A1). These results further suggested that Rf21(t) can partially restore the fertility to BT-type CMS lines.

3.3. Mapping of Rf21(t) by BSA Combined with the SLAF Sequence

Since both 9201B and ‘02428’ belong to the japonica cultivars, we conducted an SLAF-seq assay to rapidly narrow down the genomic position of Rf21(t) by identifying polymorphisms and performing gene mapping. SLAF-Seq of the two parental lines and two extreme fertility pools generated about 14.9 million high-quality short reads, and 53,948 high-quality SLAF tags were developed. According to the analysis of allele numbers and sequence differences between genomes, a total of 2275 polymorphic markers were identified, resulting in a polymorphism rate of 5.13%. Statistics for the number of markers on each chromosome, categorized by their positions, are presented in

Table A2. Number of polymorphic markers on 12 chromosomes.

Chromosome	Number of Polymorphic Markers	Chromosome	Number of Polymorphic Markers
Chromosome 1	261	Chromosome 7	188
Chromosome 2	326	Chromosome 8	346
Chromosome 3	132	Chromosome 9	77
Chromosome 4	296	chromosome10	120
Chromosome 5	273	chromosome11	325
Chromosome 6	249	chromosome12	182

. A distribution diagram of candidate markers on each chromosome is shown in Figure A2. Ultimately, 1385 polymorphic markers with a read depth of more than 5× and biallelic between ‘02428’ and 9201B were analyzed. A total of 52 different makers were found to fit the segregation ratio in the two pools (

Table A3. Distribution of different markers.

Chromosome	No. of Different Markers	Chromosome	No. of Different Markers
Chromosome 2	45	Chromosome 6	2
Chromosome 4	1	Chromosome 8	1
Chromosome 5	2	Chromosome11	1
Total			52

). A total of 45 different markers were found on chromosome 2, predominantly concentrated between 6 and 12.5 Mb (Figure 2). This suggests a high likelihood that Rf21(t) is located on chromosome 2.

To validate the mapping results obtained from the BSA analysis, we developed a set of markers, comprising 21 SSR and 2 InDel markers, within the candidate region. These markers were used to identify the polymorphisms between ‘02428’ and 9201B, and four polymorphic markers were detected. We used two markers, STS-2-39.6 and RM1358, to assay the individual genotype used for BSA in the 9201A//9201B/02428 population. The majority of fertile plants (33/35) showed a heterozygous genotype, while most sterile plants (31/35) exhibited a 9201B homozygous genotype at these two markers loci, suggesting a linkage between these two markers and Rf21(t) (Figure 3A). Therefore, Rf21(t) was primarily mapped on chromosome 2.

3.4. Fine Mapping of Rf21(t)

To clarify the precise genomic position of Rf21(t), STS-2-39.6 and RM1358 were used to screen recombinant individuals in the 9201A//9201B/02428 population. Among 691 plants, there were 15 and 2 recombinants detected by STS-2-39.6 and RM1358, respectively. Based on the data obtained from these recombinants, we delimited the location of Rf21(t) between STS-2-39.6 and RM1358, as illustrated in Figure 3A. We developed 21 SSR and 33 InDel markers in the region between STS-2-39.6 and RM1358, and nine polymorphic markers were obtained (Table A1). Four markers (named RM6375, RM12938, RM12941 and STS2-31) were used to genotype the 17 recombinants. Among these recombinant plants, one individual was positive for RM12941, while none showed positivity for STS2-31. Based on information on these recombinants, we delimited Rf21(t) between RM12941 and RM1358 (Figure 3B). Subsequently, we constructed several 9201A//9201B/02428 F₂ populations consisting of 2736 plants. Using markers RM12941 and RM1358, we identified eight recombinants and determined their genotypes at the Rf21(t) locus by progeny testing. With six additional polymorphic markers, we successfully mapped Rf21(t) to an approximate region spanning 77.0 kb, utilizing the reference sequence of the NIP genome, situated between the markers STS2-101 and STS2-20 (Figure 3C).

3.5. Candidate Gene Analysis of Rf21(t)

According to data from the Rice Genome Annotation Project (http://rice.uga.edu/index.shtml, accessed on 4 June 2025), the 77 kb region harbors 12 predicted open reading frames (ORFs). Among these, six ORFs are associated with known functions: one corresponds to an expressed protein, one is a hypothetical protein, and the remaining four are retrotransposon proteins (Figure 3D,

Table A4. Predicted genes in the 77 kb target region.

Locus Name	Gene Product Name
LOC_Os02g17360	PPR-repeat-domain-containing protein, putative, expressed
LOC_Os02g17370	Expressed protein
LOC_Os02g17380	EMB1303, putative, expressed
LOC_Os02g17390	3-hydroxy acyl-CoA dehydrogenase, putative, expressed
LOC_Os02g17400	Leucine-rich repeat protein, putative, expressed
LOC_Os02g17410	Hypothetical protein
LOC_Os02g17420	Retrotransposon protein, putative, unclassified, expressed
LOC_Os02g17430	Retrotransposon protein, putative, unclassified, expressed
LOC_Os02g17440	Retrotransposon protein, putative, unclassified, expressed
LOC_Os02g17450	Retrotransposon protein, putative, unclassified, expressed
LOC_Os02g17460	Tesmin/TSO1-like CXC domain-containing protein, expressed
LOC_Os02g17470	RNA-binding-protein-related, putative, expressed

). Among them, LOC_Os02g17380, encoding a protein consisting of 152 amino acids with a glycine-rich domain, has been cloned as Rf2 for Lead-rice-type CMS. We first sequenced the Rf2 allele in ‘02428’ and found that ‘02428’ carries the functional allele of Rf2. However, previous studies have proved that plants carrying BT-CMS with Rf2rf2 showed weak restoration of seed setting (7.0–15.2%) [33], which is lower than that of the testcross F₁ from ‘02428’ in the present study (>40%). Additionally, we sequenced Rf2 alleles in two other BT-CMS lines, namely BT-SqA and BT-LfA, revealing that BT-SqA carries the functional allele of Rf2. In conclusion, we speculated that Rf2 could not be the only candidate gene for Rf21(t). In other words, Rf21(t) may include multiple candidate genes within the target region.

In general, fertility restoration factors are expected to interact with RNA or protein products of CMS-causing mitochondrial genes, with many cloned Rf genes in rice encoding PPR proteins. Among the 12 predicted ORFs, LOC_Os02g17360 was predicted to encode a PPR-domain-containing protein. Thus, we analyzed the coding region of LOC_Os02g17360 in the parental lines BT-NipA, BT-SqA and ‘02428’. As a result, three base substitutions in exons of LOC_Os02g17360 were identified between parental lines. The variations were all predicted to cause amino acid changes, including cysteine to glycine, glycine to alanine and serine to alanine (Figure 3E). Additionally, the mRNA levels of LOC_Os02g17360 in young panicles in T8819, a stable iso-cytoplasmic BT-9201A//9201B/02428 F₅ line with a homozygous genotype, were higher than those of BT-NipA (Figure 3F). Taken together, LOC_Os02g17360 is predicated as the most likely candidate gene for Rf21(t) other than Rf2.

3.6. Genetic Effects of Rf21(t) Using CRISPR/Cas9-Based Mutagenesis

To validate the function of LOC_Os02g17360, we employed CRISPR/Cas9 genome editing to generate the KO mutants in T8819. As a result, a total of 15 independent T₀ mutants (KO1 to KO15) had been developed. Based on the analysis of the target sequence, we obtained two typed mutants. The KO1 allele contains a 1 bp deletion at the 286 bp site, resulting in a frameshift and premature termination of translation at the 207th amino acid. Similarly, the KO2 allele carries a 1 bp insertion at the 287 bp site, leading to a frameshift and premature termination of translation at the 184th amino acid (Figure 4A). Although both alleles had an overall appearance similar to recipient plant T8819 (Figure 4B), they all displayed a significantly decreased natural spikelet rate compared with T8819, but there were no differences in pollens between the T8819 and KO lines (Figure 4C–G), suggesting that LOC_Os02g17360 is the most likely candidate gene of Rf21(t) other than Rf2.

3.7. Ability of Rf21(t) to Restore Fertility

To evaluate the ability of Rf21(t) in terms of fertility restoration, we developed the NIL for Rf21(t) in the background of NIP with MAS and named this as NIL-Rf21(t) (Figure 5A,B). Similar to NIP, the NIL-Rf21(t) exhibited stainable pollen grains (Figure 5C,D). There were no differences in pollens between the CMS lines and the testcross F₁s (Figure 1A,B,D,E). A testcross was conducted between BT-NipA and NIL-Rf21(t), yielding BT-NipA/NIL-Rf21(t) F₁ plants with a natural spikelet fertility level of 28.05 ± 4.19. By detecting with STS2-12, 24 plants harboring the genotype of Rf21(t)Rf21(t), and 16 plants harboring the genotype of Rf21(t)rf21(t) were found in the BT-NIPA/NIL-Rf21(t) F₂ population, and the average natural spikelet fertility levels of these two types of plants were observed. Plants with the genotype of Rf21(t)Rf21(t) exhibited significantly higher natural spikelet fertility (54.17 ± 8.70) than that of the plants with the genotype of Rf21(t)rf21(t) (31.55 ± 7.66) (Figure 5E,F). These results further indicate that Rf21(t) exhibits only partial capability and displays a dosage effect on the fertility restoration of BT-CMS.

4. Discussion

CMS is essential for developing three-line hybrid rice, which has significantly increased grain yields. While significant progress has been made with indica hybrids, japonica hybrids currently account for only a small portion of the total japonica rice planting area in China. Therefore, developing three-line hybrid japonica varieties has a major potential to boost future rice production. BT-CMS is the most commonly used sterility system for japonica hybrid seed production. Although remarkable progress has been made in the breeding of BT-type hybrid japonica rice, only Rf1 has been used in the breeding of BT-type CMS with major genetic effect [11,12]. In this study, we identified a novel fertility restorer locus, Rf locus Rf21(t), originating from the wide-compatibility japonica variety 02428. Through molecular marker analysis, we successfully narrowed down the genetic region containing Rf21(t) to a compact 77 kb region on chromosome 2. Subsequently, we identified two candidates for Rf21(t). The use of Rf21(t) for breeding BT-type japonica restorers holds significant promise for enhancing genetic diversity and fertility restoration ability. This advancement is poised to contribute positively to the development of japonica hybrids in China.

In breeding practice, the japonica varieties from Japan and China serve as maintainers for BT-CMS, while BT-type CMS lines are developed through successive backcrosses. Notably, the Rf1 locus in most BT-type japonica restorers was introduced from indica using the “indicia–japonica bridging technique” in China [10,11,12,13,25,34]. There has been a prevailing notion that Rf genes for BT-CMS may not be present in japonica varieties. However, as has been reported previously, ‘02428’ was developed from Jibang and Pangxiegu, which are japonica varieties originating from Jiangsu and Yunnan provinces, respectively. Our study corroborates these findings, demonstrating that ‘02428’ indeed restores partial fertility to BT-CMS lines, aligning with previous research outcomes. These results indicated there might be Rf genes in japonica varieties, especially from Yunnan in China, which increases our knowledge about restoring and maintaining the relationship of BT-CMS lines and provides valuable information regarding the breeding of BT-type japonica hybrids.

In general, BT-type japonica CMS lines exhibited stainable abortive pollen grains, which were less stained than the normal pollen grains. However, the staining degree of pollen grains is affected by environmental factors such as temperature, and humidity. In many cases, it is difficult to distinguish the stainable abortive pollen grains from the normal pollen grains. In our study, we observed that the stainable abortive pollen grains showed no discernible difference from normal pollen grains. Consequently, spikelet fertility level was employed as the basis for genetic analysis. Taking the bagged spikelet level as the criterion, the 9201A//9201B/02428 F₁ plants had an approximate 1:1 ratio for phenotypic segregation, suggesting that the fertility restoration in this population is caused by a single locus. However, certain plants exhibited natural spikelet levels of >70% in the BT-9201A//9201B/02428 F₁ population, which was higher than that of the BT-9201A/02428 F₁ plants. These results were unexpected, surpassing our initial expectations. We hypothesize that there might be other fertility restorer genes with minor genetic effects present in the maintainer line 9201B. Further studies will be necessary to test this hypothesis thoroughly.

In the present study, the two parents, ‘02428’ and BT-9201A, are japonica varieties, and their close genetic relationship is a major obstacle for detecting enough polymorphic SSR or InDel markers to construct a high-density molecular linkage map for the preliminary mapping of Rf21(t). It is considered that SNPs should be a better choice for marker development in this study. SLAF-seq is a high-resolution strategy for the large-scale de novo discovery and genotyping of SNPs, and it has been proven to be efficient and cost-effective [22,35,36]. In the present study, a total of 5327 SNPs were identified using SLAF-seq technology, and these markers were uniformly distributed on 12 chromosomes. Based on these polymorphic markers identified, we conducted a bulked segregant analysis to identify candidate genes associated with Rf21(t). Our analysis revealed a candidate region with high confidence on chromosome 2. The markers within the candidate region showed a higher number of polymorphisms between two parents, and we ultimately mapped Rf21(t) to a region of 77 kb containing Rf2. Rf2 has been proven to be able to function in the fertility restoration of BT-CMS [33]. This compelled us to analyze the sequence variations in the Rf2 allele in ‘02428’. The sequencing results confirmed that Rf2 exists in ‘02428’. In a former study, the relationship between BT-CMS and LD-CMS and the restorer genes Rf1 and Rf2 was studied based on classical crossing experiments. A plant carrying BT-cytoplasm with Rf2rf2 only exhibited a seed setting rate of 15% [37,38]. However, BT-9201A/02428 F₁ plants and BT-NIPA/NIL-Rf21(t) F₁ plants carrying BT-cytoplasm with Rf21(t)rf21(t) exhibited a seed setting rate of >20% in our study. This suggests that Rf21(t) differs from Rf2. Additionally, we found that the sterile line BT-SqA carries Rf2. This suggests that there may be other candidates for Rf21(t) within the mapping region.

In rice, most cloned Rf genes are found to encode PPR proteins, including Rf1a and Rf1b in BT-type CMS [13], Rf5 and Rf6 in HL-type CMS [21,22], Rf4 in WA-type CMS [19,39] and OsRf19 in FA-type CMS [9]. Based on gene annotation, genomic sequence analysis and mRNA level analysis, we identified LOC_Os02g17360 as a high-priority candidate for further investigation. Moreover, the CRISPR/Cas9 edit that causes a frameshift and terminates translation in LOC_Os02g17360 segregated perfectly with the restoration of fertility in the validation population, demonstrating that LOC_Os02g17360 is one of the most likely target genes at the Rf21(t) locus as well. In the future, we will use a transgenic method such as knocking out the Rf21(t) candidate gene in different genetic backgrounds and a gene complementation experiment to further verify the function of LOC_Os02g17360. Meanwhile, we will clarify the molecular mechanism of the interaction between Rf21(t) and Rf2 using knockout experiments and other molecular experiments, such as Y2H, etc.

In the present study, the near-isogenic line for Rf21(t) has been successfully developed for evaluating the genetic effect of Rf21(t), and Rf21(t) was confirmed to show a weak fertility restoration of BT-CMS. BT-NipA/NIL-Rf21(t) F₁ plants exhibited a natural spikelet fertility level of 28.05 ± 4.19, which was lower than that of A/02428 F₁ plants. We speculated that there might be a minor Rf in ‘02428’. The F₂ plants harboring Rf21(t)Rf21(t) exhibited a higher fertility level than that of plants harboring Rf21(t)rf21(t), indicating that Rf21(t) has a dosage effect on the fertility restoration of BT-CMS. These observations imply that Rf21(t) might have significant value in breeding BT-type japonica restorers with multiple dominant Rf genes, potentially increasing the stability of seed settings in BT-type hybrid japonica rice.

5. Conclusions

In the current study, we successfully identified a dominant Rf gene, Rf21(t), responsible for the fertility restoration of BT-type CMS lines, sourced from the wide-compatibility japonica variety ‘02428’. Through comprehensive sequence analysis and mRNA levels analysis, LOC_Os02g17360 and LOC_Os02g17380 were confirmed to be the most promising candidates for Rf21(t). Additionally, Rf21(t) showed a dosage effect on the fertility restoration of BT-type CMS lines. Overall, Rf21(t) represents a valuable genetic resource for the breeding of BT-type japonica restorer lines. These findings offer valuable insights for breeders aiming to develop BT-type japonica hybrids.