A Novel Variation in the FRIZZLE PANICLE (FZP) Gene Promoter Improves Grain Number and Yield in Rice

Abstract

Secondary branch number per panicle plays a crucial role in regulating grain number and yield in rice. Here, we report the positional cloning and functional characterization for SECONDARY BRANCH NUMBER7 (qSBN7), a quantitative trait locus affecting secondary branch per panicle and grain number. Our research revealed that the causative variants of qSBN7 are located in the distal promoter region of FRIZZLE PANICLE (FZP), a gene previously associated with the repression of axillary meristem development in rice spikelets. qSBN7 is a novel allele of FZP that causes an ∼56% decrease in its transcriptional level, leading to increased secondary branch and grain number, and reduced grain length. Field evaluations showed that qSBN7 increased grain yield by 10.9% in a temperate japonica variety, TN13, likely due to its positive effect on sink capacity. Our findings suggest that incorporation of qSBN7 can increase yield potential and improve the breeding of elite rice varieties.

GRAIN number, one of the predominant components of rice yield, can be efficiently increased by nitrogen fertilizer (Doberman and Fairhurst 2001). To achieve high grain productivity, farmers usually apply excessive fertilizers, which lead to soil acidification, groundwater contamination, and increased costs. Pyramiding quantitative trait loci (QTL) for grain number and other yield-related traits is an alternative means of improving productivity without causing adverse environmental effects. To date, several QTL for grain number, such as GN1A, DEP1, WFP/IPA1, SCM2, SPIKE (qTSN4), GNP1, qNPT1, and SGDP7/COS1, have been identified from natural variations and applied in rice breeding (Ashikari et al. 2005; Huang et al. 2009, 2018; Jiao et al. 2010; Miura et al. 2010; Ookawa et al. 2010; Fujita et al. 2013; Wu et al. 2016; Bai et al. 2017; S. Wang et al. 2017).

Grain number is associated with several morphological components of panicle architecture, among which secondary branch number per panicle is one of the most crucial (Mei et al. 2006; Luo et al. 2009). To identify new genes for grain number, a backcross population with segregating secondary branch number per panicle and grain number phenotypes was developed by using the donor parent IR65598-112-2 and recurrent parent TN13 (S.-S. Wang et al. 2017). Genetic analysis and rough genetic mapping suggested that qSBN7, a single locus located on the long arm of chromosome 7, underlay the secondary branch and grain number variant, and that the allele of IR65598-112-2 was recessive (S.-S. Wang et al. 2017).

IR65598-112-2 is a new plant-type cultivar, one of several released in the 1990s by the International Rice Research Institute with a higher grain number per panicle, larger leaves, strong culms, and few unproductive tillers. The high-yield potentials of these cultivars render them suitable for rice breeding and genetic analysis of rice yield. However, so far only a few QTL for yield potentials have been identified from new plant-type cultivars (Fujita et al. 2013; S. Wang et al. 2017).

To reveal the genetic determinant underlying secondary branch number per panicle and grain yield in rice, we undertook map-based cloning and a functional characterization of qSBN7. Natural variations in the promoter region of FRIZZLE PANICLE (FZP), a gene encoding an APETALA2/ETHYLENE response factor (ERF), was identified as the genetic basis of qSBN7. This finding provided insight into the molecular mechanisms of the complex rice-yield trait. Phenotypic evaluations also showed that qSBN7 had pleiotropic effects on secondary branch number per panicle, grain number, grain length, 1000-grain weight, and the percentage of filled grains. Our results suggest that qSBN7 increases reproductive sink capacity in rice and can be applied in breeding new elite cultivars.

Materials and Methods

Primers

Primers used for high-resolution mapping, DNA sequencing, genotyping of four grain-number genes, vector construction, expression analysis, near-isogenic line (NIL) development, and transgenic plant identification are listed in Supplemental Material, Table S1.

Development of NILs

Two BC₃F₂ populations with the genetic backgrounds of TN13 and TCS10 were generated by recurrent backcross breeding using IR65598-112-2 as the donor parent. Two backcross inbred lines (BILs) carrying homozygous qSBN7 alleles of IR65598-112-2, BIL_TN13^sbn and BIL_TCS10^sbn, were selected from respective BC₃F₂ population based on the genotypes of the SNP2830.5 marker. A whole-genome survey of the two BILs was conducted using restriction site-associated DNA (RAD) sequencing. Seven and three introgression segments of IR65598-112-2 were detected in BIL_TN13^sbn and BIL_TCS10^sbn, respectively (Figure S1, A and B). To develop NIL_TN13^sbn and NIL_TCS10^sbn, BIL_TN13^sbn and BIL_TCS10^sbn were backcrossed to their recurrent parents two times. One BC₅F₁ plant was selected and self-pollinated to develop NIL_TN13^sbn, using eight Kompetitive Allele-Specific PCR (KASP) markers (TN13_C2_8, TN13_C3_6, TN13_C6_5, TN13_C7_22, TN13_C8_20, TN13_C9_16, TN13_C11_21, and SNP2831) located at the introgression segments of IR65598-112-2 in BIL_TN13^sbn (Figure S1A). Similarly, One BC₅F₁ plant was selected and self-pollinated to develop NIL_CS10^sbn, using five KASP markers (TCS10_C2_15, TCS10_C2_25, TCS10_C4_14, TCS10_C7_20, and SNP2831) located at the introgression segments of IR65598-112-2 in BIL_TCS10^sbn (Figure S1B). Custom KASP SNP assays and KASP Genotyping Master Mix were supplied by LGC Genomics (Middlesex, UK). KASP analysis was carried out according to the manufacture’s protocol, with a 5-µl total reaction volume, on a CFX96 Connect Real-Time PCR Detection System (Bio-Rad, Hercules, CA).

Plant materials and growing conditions

For quantitative (q)RT-PCR analysis and transgenic analysis, BIL-sbn with the homozygous qSBN7 allele of IR65598-112-2 was developed as previously described (S.-S. Wang et al. 2017) (previously named BIL-sbn/sbn). The 66 accessions used in sequence analysis of FZP are listed in Table S2. All transgenic plants and the 66 accessions used for DNA sequencing were grown in a closed greenhouse at the Biotechnology Center in Southern Taiwan of the Agricultural Biotechnology Research Center, Academia Sinica, Tainan, Taiwan (23°10’N, 120°29’E). The others were grown in paddy fields at Chia-Yi, Taiwan (23°42’N, 120°28’E). The rice plants were cultivated according to conventional management practices in a well-irrigated paddy field. The amount of nitrogen fertilizer used was 160 kg ha^-1 in each field.

Fine mapping of qSBN7

Through backcross breeding, we generated a BC₂F₅ mapping population with segregating panicle phenotypes using the donor parent IR65598-112-2 and recurrent parent TN13. A total of 1652 BC₂F₅ individuals were genotyped using the SNP2788 and SNP2855 markers. Twenty-four recombinants each, with a single crossover event in the interval between SNP2788 and SNP2855, were obtained and self-pollinated to produce BC₂F_5:6–BC₂F_5:9 lines for a progeny test. All recombinants were further genotyped using five additional markers: Indel2823, Indel2829, SNP2830.7, SNP2830.9, and SNP2835. For the progeny test, 52 individuals per BC₂F_5:6 line were visually rated as low-secondary branch number (SBN) types, segregating, or high-SBN types. Furthermore, nine homozygous recombinants were selected based on the location of their crossover sites and evaluated for secondary branch number per main panicle in the BC₂F_5:9 generation.

Sequence polymorphisms of FZP in rice germplasms

Genomic DNA samples extracted from the 66 accessions were used as DNA templates for PCR. Primers used in DNA sequencing of the 9.3-kb candidate region and genotyping of four grain-number genes (GN1A, IPA1, DEP1, and SPIKE) are listed in Table S1. To avoid false nucleotide polymorphisms caused by PCR amplification, three independent PCR amplicons resulting from each combination of primer pairs and DNA templates were mixed and then submitted to Sanger sequencing. The DNA sequences were analyzed using Chromas software version 2.23 (http://www.technelysium.com.au).

Vector construction and plant transformation

To generate the p66 plasmid, a DNA fragment spanning from 2466-bp upstream of the FZP transcription start site to 1307-bp downstream of its stop codon was amplified from TN13 plants using the Vector_p40F and Vector_p40R primers, then cloned into the binary pCAMBIA1300 vector (Cambia) treated with BamHI. This mediated vector was designated as p40. Next, a DNA fragment between 1375-bp and 7583-bp upstream of the FZP transcription start site was amplified from TN13 plants using the Vector_p66F and Vector_p66R primers, then cloned into the binary p40 vector treated with EcoRI. To generate the pRNA1 plasmid, a 21-bp artificial microRNA sequence of FZP (amiRNA) sequence (5′-TAATGATATATGCATGATCGC-3′ for the 3′-UTR of FZP) was designed using Web MicroRNA Designer 3 software (http://wmd3.weigelworld.org/cgi-bin/webapp.cgi). The amiRNA precursor was generated by gene synthesis. The sequence of the amiRNA precursor is shown in Figure S2. The precursor was amplified using the Vector_RNAi F and Vector_RNAi R primers, and then cloned into the binary vector pCAMBIA1301 (Cambia), which was digested using BglII and PmlI. All vectors were constructed using the In-Fusion HD Cloning Kit (Clontech Takara Bio). Vector p66 was introduced into BIL-sbn and vector pRNA1 into TN13 by Agrobacterium-mediated transformation (Toki et al. 2006) at the Transgenic Plant Laboratory, Institute of Plant and Microbial Biology, Academia Sinica (Taiwan). The copy number of each independent T0 plant was evaluated using TaqMan Copy Number Assays (Applied Biosystems, Foster City, CA). We chose the rice tubulin α-1 chain as an endogenous control gene (Kim et al. 2015), and hptII in the p66 and pRNA1 plasmids as the target gene. Real-time PCR data were analyzed using CopyCaller software version 2.0 (Applied Biosystems) according to the manufacturer’s instructions. To generate p66 and pRNA1 transgenic lines, a hptII-specific marker (Table S1) was used to exclude the wild-type plants in the T1 and T2 generations. For p66, 4 out of 10 T1 lines were selected and then self-pollinated to generate T2; 10 out of 32–36 plants from each T2 line (total four independent T2 lines) were selected by the hptII marker and then used for phenotyping. For pRNA1, 6 out of 12 T1 lines were selected, then self-pollinated to generate T2; 10 out of 32–36 plants from each T2 line (six independent T2 lines in total) were selected by the hptII marker and then used for phenotyping.

RNA isolation and qRT-PCR analysis

We extracted total RNA using the RNeasy Plant Mini Kit (QIAGEN, Valencia, CA). First, 1 µg of total RNA was treated with RNase-free DNase I (Ambion) and then submitted to complementary DNA (cDNA) synthesis using a SuperScriptTM III First-Strand Synthesis System Kit (Invitrogen, Carlsbad, CA). Transcriptional levels of LOC_Os07g47330 (FZP) and LOC_Os07g47340 were detected in a real-time PCR analysis. All assays were conducted with three biological and three technical replicates. Real-time PCR was performed on a CFX96 Connect Real-Time PCR Detection System (Bio-Rad). Each real-time PCR reaction was performed with a final volume of 10 µl, consisting of 5 µl of 2× SsoAdvanced Universal Probes Supermix (Bio-Rad), 0.25 µl of each primer (50 µmol/liter), 0.25 µl of probe (10 µmol/liter), 1 µl cDNA of template, and 3.5 µl of ddH₂O. The rice ubiquitin gene UBQ5 (LOC_Os01g22490) was used as the internal control.

Next-generation sequencing and sequence analysis

To obtain the full length of the insertion located 6102-bp upstream of FZP, the genomic DNA of BIL-sbn was extracted using the protocol described by the hexadecyl trimethyl-ammonium bromide (CTAB) method (Murray and Thompson 1980). A HiSeq4000 platform (Illumina) was used for next-generation sequencing. DNA libraries were sequenced as 150-bp pair-end reads at Tri-I Biotech Inc. (Taipei, Taiwan). A total of 75.3 Gb of raw reads were obtained and de novo assembled at Genomics company (Taipei, Taiwan). To investigate the whole-genome backgrounds of BIL_TN13^sbn and BIL_TCS10^sbn, the RAD libraries of TN13, TCS10, IR65598-112-2, BIL_TN13^sbn, and BIL_TCS10^sbn were constructed as previously described (Etter et al. 2011). A total of 7.6 Gb of raw reads were obtained on an Illumina HiSeq4000 at Tri-I Biotech Inc. and then analyzed as previously described (S.-S. Wang et al. 2017).

Analysis of grain quality

To evaluate eating-quality traits, 150–200-g grains from each replication were milled to yield brown rice grains. Chalky grain ratio was evaluated using 1000 brown rice grains and a rice quality selector (Satake Co.) following the operating manual. The amylose content and protein content were measured using a grain composition analyzer (Kett Co.) in accordance with the operations manual.

Data availability

Materials and plasmids are available upon request. Whole-genome sequences of BIL_sbn are deposited in the Sequence Read Archive (SRA) database (SRA accession: PRJNA607167; https://www.ncbi.nlm.nih.gov/sra/PRJNA607167). RAD sequencing data are deposited in the SRA database (SRA accession: PRJNA607003; https://www.ncbi.nlm.nih.gov/sra/PRJNA607003). All other relevant data are within the article and its tables and figures. Supplemental material, including FZP sequences of 66 rice accessions (File S1), and the data in supplemental figures and supplemental tables, available at figshare: https://doi.org/10.25386/genetics.11919255.

Results

High-resolution mapping and identification of qSBN7

TN13 is a temperate japonica cultivar with ∼19.5 secondary branches and 111.2 grains per panicle; IR65598-112-2 is a tropical japonica (javanica) cultivar with ∼82.1 secondary branches and 371.3 grains per panicle (Figure 1, A–E). To identify the causal gene of qSBN7, we conducted high-resolution linkage analysis on a segregating population consisting of 1652 BC₂F₅ individuals derived from IR65598-112-2 and TN13. The qSBN7 locus was narrowed to an ∼9.3-kb segment between the Indel2829 and SNP2830.9 markers (Figure 1F and Figure S3). According to Michigan State University Rice Genome Annotation Project release 7 (http://rice.plantbiology.msu.edu/), this region covers the full length of LOC_Os07g47330 and the last two exons of LOC_Os07g47340 (Figure 1G). We compared the sequences of TN13 and IR65598-112-2 across the 9.3-kb candidate region and found no difference in the coding regions of the two candidate genes, except for two SNPs (c.-4066C > T and c.-6383A > G) and an 8461-bp indel (c.-6101 > -6102insCACTACCA…) in the upstream region of LOC_Os07g47330 (Figure 1G). The 8461-bp insertion encodes a putative CACTA transposon (Figure S4), and thus may affect the expression of LOC_Os07g47330 and LOC_Os07g47340, but the transposon itself is unlikely to be the causative gene.

Figure 1.

Map-based cloning of qSBN7. (A) Plant architecture of IR65598-112-2. Bar, 20 cm. (B) Plant architecture of TN13. Bar, 20 cm. (C) Panicle structure of IR65598-112-2 and TN13. Bar, 4 cm. (D) Number of secondary branches per panicle. (E) Number of grains per panicle. (F) A high-resolution map delimiting the qSBN7 locus to a 9.3-kb region between the Indel2829 and SNP2830.9 markers. (G) The structure of two putative genes predicted in Rice Genome Annotation Project release 7. A C/T SNP, 8461-bp insertion, and A/G SNP were located 4066, 6102, and 6383 bp upstream of LOC_Os07g47330, respectively. Values in (D) and (E) are means ± SD (n = 10 plants). Chr., chromosome; No., number.

qRT-PCR analysis was conducted to reveal the transcript levels of the two candidate genes at the seedling and three-panicle developmental stages in TN13 and BIL-sbn (a BC₂F₄-derived line carrying the homozygous qSBN7 locus of IR65598-112-2 in the TN13 genetic background). No significant differences were detected between the expression of LOC_Os07g47340 in TN13 and BIL-sbn at all stages (Figure 2A). However, significantly higher expression of LOC_Os07g47330 was observed at the 1-mm panicle stage in TN13 than in BIL-sbn (Figure 2B). At this stage, the primary rachis branch meristem produces lateral branches (Ikeda et al. 2004), suggesting that differential expression of LOC_Os07g47330 may cause variations in secondary branch number. LOC_Os07g47330 is the previously identified FZP gene in rice. FZP encodes an ERF transcription factor (Komatsu et al. 2003) and is the ortholog of the maize BRANCHED SILKLESS 1 gene (Chuck et al. 2002).

Figure 2.

Expression analysis of two candidate genes and complementation test of qSBN7. (A) Transcript levels of LOC_Os07g47340 at seedling and three-panicle development stages. (B) Transcript levels of LOC_Os07g47330 at seedling and three-panicle development stages. (C) Transcript levels of FZP in BIL-sbn and four independent transgenic lines at the 1-mm panicle stage. (D) Panicle structure of BIL-sbn. Bar, 4 cm. (E) Panicle structure of p66-01. Bar, 4 cm. (F) Panicle structure of TN13. Bar, 4 cm. (G) Grains of BIL-sbn, p66-01, and TN13. Bar, 5 mm. (H) Number of secondary branches per main panicle. (I) Number of grains per main panicle. (J) Grain length. Values in (A–C) and (H–J) are means ± SD [n = 3 independent trials and three plants per trial in (A–C); n = 10 plants in (H–J)]. Data in (D–J) were collected from plants grown in paddies under greenhouse conditions in the 2016 second crop season. The plant spacing was 15 cm between plants and 20 cm between rows. Student’s t-test was used to examine P-values. **Significant at 1% level; *significant at 5% level. BIL, backcross inbred line; n.s., not significant; No., number.

FZP was the gene responsible for qSBN7

To examine whether FZP underlay the secondary branch and grain-number variant, a 9.8-kb genomic DNA fragment containing the FZP locus of TN13 (designated as p66) (Figure S5) was introduced into BIL-sbn. The transgenic plants carrying a single copy of p66 showed significantly lower secondary branch number per panicle, lower grain number, and increased grain length (Figure 2 and Figure S6). Comparison of FZP transcript levels in BIL-sbn and the four independent T2 lines showed that three lines with rescued phenotypes (p66-01, p66-20, and p66-27) exhibited significantly higher FZP expression than BIL-sbn (Figure 2C). These results indicated that FZP was the gene responsible for the observed qSBN7 effects.

Negative correlation between the level of FZP expression and qSBN7 phenotypes

Next, we investigated whether the differential expression of FZP correlated with the levels of secondary branch number per panicle, grain number, and grain length. We generated a silencing transgenic line by introducing pRNA1, which contains a 21-bp amiRNA1, into TN13. Compared with TN13, most of the independent T1 lines with single-copy amiRNA1 showed increased secondary branch per panicle and grain number, and reduced grain length, with some variations (Figure 3 shows comparisons between TN13 and a representative silencing line, pRNAi-19; data from all 12 T1 lines are given in Figure S7, A–C). However, we observed no significant difference in primary branch number per panicle among the transgenic plants and TN13 (Figure 3B and Figure S7D)

Figure 3.

Transgenic analysis for FZP through gene silencing. (A) Panicle structures of TN13, BIL-sbn, and four transgenic T1 lines. Bar, 4 cm. (B) Comparison of primary branch number per main panicle between TN13 and pRNAi-19. (C) Comparison of secondary branch number per main panicle between TN13 and pRNAi-19. (D) Comparison of grain number per main panicle between TN13 and pRNAi-19. (E) Comparison of grain length between TN13 and pRNAi-19. (F) FZP transcript levels in TN13, BIL-sbn, and six independent T2 transgenic lines. Values in (B–F) are means ± SD [n = 10 plants in (B–E); n = 3 independent trials and three plants per trial in (F)]. Data in (A–E) were collected from plants grown in paddies under greenhouse conditions in the 2016 second crop season. The plant spacing was 15 cm between plants and 20 cm between rows. Data in F were collected from plants grown in paddies under greenhouse conditions in the 2017 first crop season. The plant spacing was 23 cm between plants and 23 cm between rows. Student’s t-test was used to examine P-values. **Significant at 1% level; *significant at 5% level. BIL, backcross inbred line; n.s., not significant; No., number.

We then performed qRT-PCR to clarify the transcript levels of FZP in TN13 and six randomly chosen independent T2 lines. FZP was expressed in the six T2 lines at lower levels than in TN13 (Figure 3F), and its expression was negatively correlated with secondary branch number per panicle (r = −0.883) (Figure S8). We noted that pRNAi-19, pRNAi-03, and BIL-sbn, which exhibited similar FZP expression levels (∼36–44% of FZP expression in TN13), all showed increased secondary branch per panicle and grain numbers, and similar panicle types (Figure 3 and Figure S7). In pRNAi-01, the line exhibiting a stronger silencing effect (FZP expression decreased to 24.6%), we found an increased secondary branch number per panicle, almost no grains, and a frizzy panicle (floret formation replaced by continuous branching), which was similar to previously reported fzp mutants¹⁴ (Figure 3A). Our findings confirmed that differential expression of FZP affected secondary branch number per panicle, grain number, and grain length, but not primary branch number per panicle. They also suggested that qSBN7 was a hypomorphic allele for lower expression of FZP.

Sequence polymorphisms of FZP in rice germplasms

To investigate functional allelic variations in FZP, 66 accessions consisting of 22 temperate japonica, 15 tropical japonica, and 29 indica cultivars were selected for sequencing analysis of a 1.8-kb genomic DNA region spanning a 402-bp promoter, 5′-UTR, the coding region, and 3′-UTR (Figure S9, A and B). Since IR65598-112-2 and TN13 have two SNPs, and one insertion difference, in the distal promoter region of FZP, the genotypes near these three polymorphic sites in the 66 accessions were also examined. According to the panicle structure of p66 and pRNA1 transgenic lines, FZP affected secondary branch number per panicle by regulating secondary branch number per primary branch (Figure S6E and Figure S7E). Therefore, secondary branch numbers per primary branch of the 66 accessions were evaluated. Eleven haplotypes were identified, with only four synonymous polymorphisms and one missense polymorphism found in the coding region (Figure S9C). Two indica cultivars carrying the same missense polymorphism, IR61608-3B-20-2-2-1-2 (haplotype 10) and Mudgo (haplotype 11), showed different levels of secondary branch numbers per primary branch, indicating that this was a neutral mutation (Figure S9C and Table S2). Additionally, no amino acid changes in ERF elements were detected (Figure S9C), indicating that the FZP protein was highly conserved in rice.

Nucleotide variants found in the promoter and UTR regions may regulate the transcriptional level of FZP, resulting in differential numbers of secondary branches and grains. A comparison between the genotypes and phenotypes of the 66 accessions (Figure S9C and Table S2) suggested that the polymorphisms at sites −6383, −220, −136, 1211, and 1276 were unlikely to be the key regulatory elements for the trait. On the other hand, haplotypes 6 (IR65598-112-2 and IR76904-7-19) and 7 (BSI325) shared the same variants at sites −6102, −4066, and 1103, and exhibited relatively high numbers of secondary branches per primary branch. Site 1103 (= Indel2829 marker) at the 3′-UTR was excluded based on the result from high-resolution mapping of qSBN7 (Figure S9A).

Among individual accessions in different rice populations (Table S2), IR76904-7-19 (haplotype 6) and IR65598-112-2 (haplotype 6) showed higher numbers of secondary branches per primary branch than other tropical japonica cultivars investigated in this study. Similarly, BSI325 (haplotypes 7) showed higher numbers of secondary branches per primary branch than other temperate japonica cultivars tested. However, among the tested indica cultivars, Kasalath and Zhuan (which shared the same variants at sites −6102 and −4066) did not show significantly higher numbers of secondary branches per primary branch than the rest of the indica cultivars (Table S2). To understand whether other trait-related polymorphisms occurred in the distal promoter region, we sequenced an 8331-bp region upstream of FZP for IR76904-7-19, BSI325, Kasalath, and Zhuan. Our results revealed that whereas IR76904-7-19 and BSI325 were monomorphic to IR65598-112-2, Kasalath and Zhuan9 differed from IR65598-112-2 by an 18-bp duplication (5′-GCACGCACGCACGGACGC-3′) located 5308-bp upstream of FZP (Table S2). This indicated that natural variants located in the distal promoter region of FZP played a regulatory role in the production of secondary branches per panicle and grains in rice, and qSBN7 was a rare allele in the collected accessions.

qSBN7 enhanced sink capacity and yield in a favorable genetic background

To evaluate the potential of qSBN7 for high-yield breeding, we generated two NILs, NIL_TN13^sbn and NIL_TCS10^sbn, that carried the homozygous qSBN7 allele of IR65598-112-2 in the TN13 (temperate japonica) and TCS10 (indica) genetic background, respectively (Figure S1, C and D). Phenotypic characterization of TN13 and NIL_TN13^sbn in rice paddies showed no significant morphological differences at vegetative stages (Figure 4A). For panicle traits, NIL_TN13^sbn showed higher numbers of secondary branches per panicle, grains per panicle, and grains per secondary branch than TN13, but also shorter grain length (Figure 4). Similarly, there were no significant phenotypic differences among TCS10 and NIL_TCS10^sbn before flowering. However, NIL_TCS10^sbn showed severely degenerated panicle and abortive grains (Figure S10), which were similar to the phenotypes of the mutant lines (pRNAi-21) exhibiting a stronger silencing effect of FZP (Figure 3A).

Figure 4.

Phenotypic characterization of TN13 and NIL_TN13^sbn. (A) Plant structure of TN13 and NIL_TN13^sbn. Bar, 20 cm. (B) Panicle structure of TN13 and NIL_TN13^sbn. Bar, 4 cm. (C) Grain size of TN13 and NIL_TN13^sbn. Bar, 5 mm. (D) Number of primary branches per main panicle. (E) Number of secondary branches per main panicle. (F) Number of grains per main panicle. (G) Grain length. (H) Number of grains per secondary branch. A completely randomized design with three replications was used in all trials. Data in (D–H) were collected from plants grown in paddies under natural conditions. The plant spacing was 15 cm between plants and 30 cm between rows. Values in (D–H) are means ± SD (n = 10 plants). Student’s t-test was used to examine P-values. **Significant at 1% level. NIL, near-isogenic line; n.s., not significant; No., number.

Additional field experiments were conducted to determine whether qSBN7 affects important agronomic traits, grain quality, and yield components in the TN13 genetic background. No significant difference in days to heading, plant height, amylose content, protein content, chalky grain ratio, and panicle number per plant were observed between TN13 and NIL_TN13^sbn (Table 1 and Table S3). However, NIL_TN13^sbn showed significantly increased grain number, sink capacity, and grain yield, and decreased 1000-grain weight and filled-grain percentage, than in TN13 (Table 1). NIL_TN13^sbn exhibited 10.9% higher grain yield than in TN13 (Table 1). Taken altogether, our findings indicated that qSBN7 is a pleiotropic QTL for panicle traits, and it could increase rice production without affecting grain quality in a favorable genetic background such as TN13.

Table 1. Comparison of agronomic traits between TN13 and NIL_TN13^sbn.

Traitsa	TN13	NIL_TN13sbn	P-valueb
Days to heading	76.7 ± 1.5	76.0 ± 1.7	0.643
Plant height (cm)	93.2 ± 1.2	95.3 ± 3.2	0.338
Panicles per plant	15.6 ± 1.2	16.3 ± 1.2	0.493
Grains per panicle	99.2 ± 4.8	170.0 ± 16.1	0.002
Percentage of filled grains	85.6 ± 1.1	65.8 ± 9.2	0.025
1000-grain weight (g)	26.1 ± 0.8	21.1 ± 0.8	0.002
Sink capacity per plantc (g)	39.9 ± 1.4	58.1 ± 7.3	0.013
Grain yield per plant (g)	34.2 ± 1.5	37.9 ± 1.3	0.045

NIL, near-isogenic line.

^aThe experiments were carried out in paddy fields at Chia-Yi, Taiwan, in the first crop season of 2019. A completely randomized design with three replications was used in all trials. Plant spacing was 15 cm between plants and 30 cm between rows, with a single plant per hill. Data are shown as mean ± SD (n = 6 plants).

^bThe P-value indicates the significance of the difference between TN13 and NIL_TN13^sbn based on a Student’s t-test.

^cSink capacity per plant (maximum grain weight per plant) = (number of grains per plant) × (1000-grain weight)/1000.

Discussion

Grain number is an essential agronomic trait for rice production, because it determines reproductive sink capacity and thus affects grain yield. Several genes for grain number have been identified from natural variations (Ashikari et al. 2005; Huang et al. 2009, 2018; Jiao et al. 2010; Miura et al. 2010; Ookawa et al. 2010; Fujita et al. 2013; Wu et al. 2016; S. Wang et al. 2017), each with distinct pleiotropic effects on other agronomic traits. For example, the dep1 allele (DEP1 allele of Shennong 265) enhanced grain and secondary branch number per panicle, but reduced 1000-grain weight, panicle length, and plant height (Huang et al. 2009). The ipa1 allele (IPA1 allele of Shaoniejing) increased grain number, 1000-grain weight, and plant height, but reduced tiller number (Jiao et al. 2010). These genes with differential effects are perhaps suitable for different ecological conditions and genetic backgrounds, and can be applied to achieve different breeding goals. Further identification of yield-related genes will allow greater flexibility in rice-breeding programs.

Identifying advantageous alleles from natural variants can facilitate the breeding of high-yield rice. We performed high-resolution mapping for a secondary branch per panicle and grain number locus, qSBN7. Fine genetic mapping and a complementation test for qSBN7 revealed that the causal gene was FZP, which encodes an ERF transcription factor gene (Komatsu et al. 2003). Grass inflorescence development involves meristem determinacy/indeterminacy decisions (Bommert and Whippleb 2018). FZP appears to be a key regulator of the competitive balance between the formation of axillary and floral meristems. Komatsu et al. (2003) showed that FZP functions to repress axillary meristem formation in rice spikelets. Mutation of FZP resulted in exceptionally high numbers of secondary and higher-order branches without normal spikelets (Komatsu et al. 2003; Yi et al. 2005; Kato and Horibata 2012; Bai et al. 2016). FZP was recently identified as the causal gene of two grain-number loci in rice, Small Grain and Dense Panicle 7 (SGDP7) and CONTROL OF SECONDARY BRANCH 1 (COS1) (Bai et al. 2017; Huang et al. 2018). An 18-bp duplication ∼5.3-kb upstream of FZP (SGDP7) caused reduced expression of FZP, and increased grain number and grain yield (Bai et al. 2017). A 4-bp deletion ∼2.7-kb upstream of FZP (COS1) caused reduced expression of FZP, and increased secondary branches and grain yield (Huang et al. 2018). In this study, we found a novel and advantageous allele of FZP. Compare to cultivated rice (TN13), qSBN7 of IR65598-112-2 is also a hypomorphic allele of FZP that causes an ∼56% decrease in its transcriptional level. Similar to SGDP7, qSBN7 probably increased the transition from spikelet meristem to secondary branch meristem on primary branches while sufficiently preventing the formation of tertiary branches on secondary ones. Thus, qSBN7 can simultaneously increase secondary branch and grain numbers, making it suitable for breeding elite rice varieties.

Sequence polymorphisms of FZP in 66 accessions showed that the FZP protein was highly conserved in rice, and that variants located in the distal promoter region were responsible for modulating the production of secondary branches per panicle and grains. It is notable that among the 66 accessions, only IR65598-112-2, IR76904-7-19, and BSI325 carried the qSBN7 allele, suggesting that qSBN7 has not been widely used in modern rice-breeding programs. In addition to the two polymorphic sites (−6102 and −4066) identified in the qSBN7 allele, the 18-bp duplication previously identified in the SGDP7 allele (Bai et al. 2017) was found at site −5308 (in Kasalath and Zhuan 9) (Table S2). This duplication contained two BES1 transcription factor-binding sites (CGTGCG), which were reported to be regulated by a brassinosteroid signal transduction gene, OsBZR1(He et al. 2005; Bai et al. 2007). Curiously, although both of the variants of qSBN7 (−6102 and −4066) and the 18-bp duplication (Bai et al. 2017) could repress the FZP expression and increase secondary branch number in rice, two indica cultivars, Kasalath and Zhuan (which contained both the variants of qSBN7 and the 18-bp duplication), did not show significantly higher numbers of secondary branches per primary branch than the rest of the indica cultivars. How different variations in the promoter region of FZP affect its expression and the regulation of secondary branch number per primary branch remain to be resolved.

Transgenic analysis and allelic evaluation of qSBN7 revealed its pleiotropic effects of increasing secondary branch per panicle and grain number, and reducing grain length, 1000-grain weight, and percentage of filled grains (Figure 2, Figure 4, and Table 1). Other studies have noted trade-offs among yield components. It has been reported that improving grain number per panicle could increase competition for assimilate supply, resulting in a reduced filled grain percentage and 1000-grain weight. Thus, introgression lines carrying QTL for grain number usually showed increased panicle size but not enhanced grain yield (Ohsumi et al. 2011; Takai et al. 2014; Fukushima et al. 2017). In this study, qSBN7 could improve grain number and grain yield in TN13 by ∼71.4% and 10.9%, respectively (Table 1). TN13 is a variety exhibiting high lodging resistance, long and erect flag leaf, and a low number of grains. The architectures imply that TN13 may have relatively higher photosynthesis efficiencies but poor sink capacity (Li et al. 1998; Horton 2000). When qSBN7 was introgressed into TN13, the improved sink capacity together with the inherently high source capacity in TN13 successfully improved the grain yield. In contrast, TCS10 is a high-grain-number variety that carries a grain number QTL, GN1A (Table S2). Although pyramiding qSBN7 with GN1A may further promote the development of spikelet meristems, most of the spikelets in NIL_TCS10^sbn were aborted (Figure S10), perhaps due to a shortage of carbohydrates. Source, sink, and translocation capacities all play important roles in grain yield (Ohsumi et al. 2011; Adriani et al. 2016). The qSBN7 allele of IR65598-112-2 had no effect on source size, but the efficiencies of photosynthesis and translocation of carbohydrates appeared to be critical during the grain-filling stage. In recent years, some QTL for source-related traits have been identified and used for rice breeding (Sun et al. 2014; Hu et al. 2015). Pyramiding the beneficial qSBN7 allele and source-related QTL may be the key to successful development of high-yield rice.