Genetic mapping of the qSBN7 locus, a QTL controlling secondary branch number per panicle in rice.

Secondary branch number (SBN) is an important component affecting spikelet number per panicle (SPP) and yield in rice. During recurrent backcross breeding, four BC2F4 populations derived from the high-yield donor parent IR65598-112-2 and the recurrent parent Tainan 13 (a local japonica cultivar) showed discontinuous variations of SPP and SBN within populations. Genetic analysis of 92 BC2F4 individuals suggested that both SPP and SBN are controlled by a single recessive allele. Two parents and 37 BC2F4 individuals showing high- and low-SBN type phenotypes were analyzed by restriction-site associated DNA sequencing (RAD-seq). Based on 2,522 reliable SNPs, the qSBN7 was mapped to a distal region of the long arm of chromosome 7. Trait-marker association analysis with an additional 166 high-SBN type BC2F4 individuals and 8 newly developed cleaved amplified polymorphic sequence markers further delimited the qSBN7 locus to a 601.4-kb region between the markers SNP2788 and SNP2849. Phenotype evaluation of two BC2F5 backcross inbred lines revealed that qSBN7 increased SPP by 83.2% and SBN by 61.0%. The qSBN7 of IR65598-112-2 could be used for improving reproductive sink capacity in rice.

Introduction

Spikelet number per panicle (SPP, also known as SNP) is one of the predominant components of rice yield. Improving SPP can increase reproductive sink capacity and promote rice yield. SPP is a typical quantitative trait controlled by quantitative trait loci (QTLs), and the phenotype can be affected by epistatic interactions and environmental factors (Xing ). Therefore, SPP is difficult to improve efficiently by traditional breeding methods.

In the last 10 years, five SPP-related genes, including GN1A, Ghd7, DEP1, WFP/IPA1 and SPIKE/NAL1, have been cloned, and the molecular mechanisms underlying SPP have been described (Ashikari , Fujita , Huang , Jiao , Miura , Xue ). In addition, numerous QTLs for SPP have been identified (Balkunde , Fujita , Ikeda , Koide , Liu , Luo , Mei , Terao , Tian , Xing , Zhang ). These findings have allowed for early-generation marker-assisted selection of breeding materials, to more efficiently improve rice yield (Lai ).

SPP is associated with many morphological components of panicle architecture, including secondary branch number (SBN). SBN and SPP are positively correlated (Luo , Mei ). A few studies have identified QTLs for SBN (Ando , Ikeda , Luo , Mei ). Eleven QTLs were found associated with SBN in a study of two populations of reciprocal introgression lines. Among these QTLs, qSBN-9, contributed a large effect on both SBN and SPP (Mei ). Another major QTL for SBN, qSBN1, was detected in a study of 39 chromosome segment substitution lines derived from a cross between the japonica cultivar Sasanishiki and indica cultivar Habataki. The qSBN1 from Habataki increased SPP approximately 30% (Ando ). A total of 9 QTLs for SBN were mapped by using 265 introgression lines of common wild rice (Oryza rufipogon Griff.) in the background of indica cultivar Guichao 2 (Luo ). Three major QTLs for secondary rachis branch number (SRN; same as SBN) on chromosomes 1 and 8 were detected in an F2 segregated population derived from a cross between a japonica cultivar Koshihikari and a new plant type (NPT) cultivar NP-6 (Ikeda ).

Tainan 13 (TN13) is a local japonica cultivar with good characteristics such as early maturing and good taste. To improve its reproductive sink capacity, marker-assisted backcrossing was used to introgress GN1A from a high-yield NPT cultivar IR65598-112-2 into TN13. GN1A is a previously cloned OsCKX2 (cytokinin oxidase/dehydrogenase) gene affecting grain number per panicle (Ashikari ). During the process, SPP and SBN distinctly segregated in four BC2F4 populations (homozygous for GN1A).

In the current study, one of the segregating BC2F4 populations was investigated for genetic dissection of SBN. To achieve high-resolution mapping, restriction-site associated DNA sequencing (RAD-seq) was performed for genotyping. Genetic evidence confirmed that a QTL located on the long arm of chromosome 7 (designated qSBN7) was responsible for the SPP and SBN variations. In addition, the genetic effect of qSBN7 was validated and assessed using two backcross inbred lines (BILs) carrying different alleles. Our results suggest the potential of using the qSBN7 of IR65598-112-2 for SPP improvement and high-yield breeding in rice.

Materials and Methods

Plant materials

The breeding scheme for developing mapping populations is illustrated in Fig. 1. A BC2F2 population with the genetic background of a japonica cultivar TN13 was established by recurrent backcross breeding, with IR65598-112-2 used as a donor parent. Four BC2F2 individuals with the GN1A homozygous allele were self-pollinated to produce bulk BC2F3 seeds for phenotypic investigation. Six BC2F3 plants showing similar plant height and heading date to TN13 were self-pollinated to create six BC2F4 populations. Variations in SPP and SBN were observed in four of the six populations. One BC2F4 population (n = 800) which showed segregation of SPP, SBN, and spikelet number per primary branch (SPPB) was arbitrarily selected for subsequent analyses (Fig. 1). In this population, 92 individuals were randomly selected for evaluating of SPP, SBN, and SPPB. Totally 37 individuals, including 18 individuals showing high-SBN type (SPP > 210, SBN > 45 and SPPB > 18.5) and 19 individuals showing low-SBN type (SPP < 170, SBN < 35 and SPPB < 14.0), were used for substitution mapping by RAD-seq. An additional 166 high-SBN type individuals (SPP > 210, SBN > 45, and SPPB > 18.5) were used to construct a high-resolution map of qSBN7. Two BC2F5 BILs, BIL-Sbn/Sbn, with a homozygous qSBN7 allele of TN13, and BIL-sbn/sbn, with a homozygous qSBN7 allele of IR65598-112-2, were chosen from the same BC2F4 population based on the genotypes of two cleaved amplified polymorphic sequence (CAPS) markers, SNP2750 and SNP2962, on chromosome 7 (Table 1).

Fig. 1

Schematic of the generation of the genetic materials used in this study. MAS, marker-assisted selection; BIL, backcross inbred line; SBN, secondary branch number.

Table 1

Newly developed cleaved amplified polymorphic sequence markers used for high-resolution linkage mapping of qSBN7

ID	SNP position (Mb)a	Tab (°C)	Forward Primer (5′-3′)	Reverse Primer (5′-3′)	REc
SNP2750	27.5	55	gaacacacccttccaattttgt	ccatataatactagcctgcacagc	HpyCH4IV
SNP2788	27.9	55	agtggaacccttctgaaatcaa	attgcaatttgactggttgaca	ApeKI
SNP2830	28.3	55	tcgtgcgtgtagaggataagaa	aagcaaaccaaacacaatcctt	HypCH4IV
SBN2835	28.4	55	gtcaaggatacgatcatatctttct	tctaaagtgggttcgtggctta	Hpy188I
SBN2849	28.5	55	cggtagagcattgtggatacag	gatgtgatggaaagttggaagctt	DdeI
SBN2855	28.6	55	aggtggacaattcatccagaag	aacggcccataatctcacatag	TaqI
SBN2906	29.1	55	cccctatttaccgttacccatt	gtatccaacgtttgacctttcg	Mn1I
SNP2962	29.6	55	caactcaggcagtaacaatgga	caaaggaacagaaaacatgcaa	AvaII

Physical map position of the targeted SNP, based on the MSU Rice Genome Annotation Project Database release 7.

Ta, annealing temperature.

RE, restriction enzyme.

Genetic analysis

Four panicle traits, SPP, SBN, SPPB, and primary branch number (PBN), were manually counted in plants. The population for genetic analysis was grown in the experimental farm at Chiayi Branch Station, Tainan District Agricultural Research and Extension Station, in the second cropping season of 2014. Three panicles per plant were measured. Normality of the phenotypic distribution was investigated by generating the normal probability plot in R 3.0.2 (R Core Team 2013). The chi-square goodness-of-fit test was used to examine the hypothesis of segregation ratio. Pearson correlation coefficients between traits were analyzed by using R 3.0.2 (R Core Team 2013).

RAD library preparation and sequence analysis

Genomic DNA was extracted from leaves of mature plants as described (Murray and Thompson 1980). For substitution mapping, two parents and 37 BC2F4 individuals (18 high-SBN type and 19 low-SBN type) were genotyped by using RAD-seq. The RAD library was constructed as described (Etter ) with some modifications. DNA was digested by using PstI (New England Biolabs, MA, USA), then ligated with the adapters containing a multiplex identifier we previously designed (Hsieh ). DNA libraries were sequenced as 100-bp single-end reads on an Illumina HiSeq2500 at the VYM Genome Research Center in National Yang-Ming University (Taipei). Raw sequence data were de-multiplexed by using Stacks software (Catchen ), then mapped to the Nipponbare reference genome sequences IRGSP1.0 (Kawahara ) by using Bowtie2 software (Langmead and Salzberg 2012). SNP identification and genotype scoring involved use of Stacks package (Catchen ). The tags from IR65598-112-2 and TN13 underwent comparative analysis to identify SNPs. To improve the accuracy of data, low-quality SNPs (read depth <6 or >5 missing data per locus) were filtered. After filtering, 2,522 SNPs were retained for the 37 BC2F4 individuals. Linkage analysis involved use of the R/qtl package (Broman ). The SNP positions in this study are based on the MSU Rice Genome Annotation Project Database release 7 (http://rice.plantbiology.msu.edu/) (Kawahara ).

CAPS marker development

Raw RAD-seq reads for two parents were analyzed by using the “demultiplex reads”, “map reads to reference”, “variant detection” and “comparing variants” functions in CLC Genomics Workbench v.7.05 with the following parameters: mismatch cost 2, insertion cost 3, deletion cost 3, length fraction 0.9, and similarity fraction 0.97. All reads were aligned to the Nipponbare reference genome IRGSP1.0 (Kawahara ). A total of 109 SNPs were detected between the physical positions 26.7 and 29.6 Mb on the long arm of chromosome 7. Eight SNPs, evenly distributed around the markers 87830 (29.0 bp) and 87941 (29.2 bp), were designed as CAPS markers. Primers were designed by using Primer3 (Untergrasser ; http://bioinfo.ut.ee/primer3-0.4.0/primer3/). The primer sequences for CAPS markers and corresponding annealing temperatures are in Table 1.

Genomic DNA was extracted from 166 individuals by using a simple, fast CTAB method (Murray and Thompson 1980). PCR was performed as described (Wang ). The PCR products were separated on 2% Agrose SFR gel (Amresco, Solon, OH, USA) with the TBE buffer system and stained with ethidium bromide.

High-resolution mapping

One BC2F4 individual was genotyped with 8 CAPS markers (Table 1) to find the borders of the IR65598-112-2 introgression segment. Overall, 26 BC2F4 recombinants with a single crossover event in the interval between the markers SNP2750 and SNP2962 were self-pollinated to produce BC2F5 lines for the progeny test. Field trials for the progeny test involved three replications in the first cropping season of 2015. In each replication, 16 plants per BC2F5 line were transplanted into rows 25 cm apart with 20 cm within-row spacing. Ten individuals per BC2F5 line were visually rated as low-SBN, segregating, or high-SBN types. Moreover, six BC2F5 lines containing different recombinant breakpoints were assessed for SBN. The data were analyzed by ANOVA with SAS 9.2 (SAS Institute, Cary, NC). Comparisons among lines were assessed by Fisher’s least significant difference test at α = 0.05.

Phenotype evaluation of BILs

BIL-Sbn/Sbn, BIL-sbn/sbn and TN13 were grown with three replications in the experimental farm at Chiayi Branch Station, Tainan District Agricultural Research and Extension Station, in the first cropping season of 2015. A total of 6 agronomic traits, including plant height (PH), panicle number per plant (PN), SPP, PBN, SBN, and SPPB were evaluated at maturity. For each line, 5 individuals from the middle of a row were chosen for phenotyping. Panicles in the tallest tillers were harvested and measured. The data for each trait were analyzed by ANOVA. Comparisons among lines were assessed by Fisher’s least significant difference test at α = 0.05.

To evaluate the effect of qSBN7 on grain traits, BIL-Sbn/Sbn and BIL-sbn/sbn were grown with three replications, 52 plants (in 2 rows) per line per replication, in the experimental farm at Chiayi Branch Station, Tainan District Agricultural Research and Extension Station in the second cropping season of 2016. Traits including 1,000-grain weight (TGW), ripened grain ratio (RGR), and chalky grain ratio (CGR) were evaluated at maturity. For each line, 10 individuals were randomly chosen for measurement. Differences between lines were analyzed by Student’s t test at α = 0.05.

Results

Phenotypic variation and genetic analysis of SBN

Ninety-two BC2F4 individuals among one segregation population were evaluated for four panicle traits, including SPP, PBN, SBN and SPPB. The frequency distribution of PBN exhibited a normal distribution. The frequency distributions of SPP, SBN and SPPB exhibited discontinuous variations (Fig. 2, Supplemental Fig. 1). The 92 BC2F4 individuals were classified into low- and high-SBN type subgroups, with separation of SPP, SBN and SPPB at 210–230, 45–50, and 18.5–20.0, respectively. The ratio of number of individuals in the low- and high-SBN subgroups was 3:1 (χ21,0.05 = 0.93 < 3.84). Furthermore, significant positive correlations between SPP and SBN, SPP and SPPB, and SPPB and SBN were found. PBN was not correlated with SPP and SBN (Table 2). The results suggested that SPP, SBN and SPPB may be controlled by a single locus and the high-SBN phenotype may have a recessive pattern of inheritance.

Fig. 2

Frequency distribution of (A) spikelet number per panicle (SPP), (B) primary branch number (PBN), (C) secondary branch number (SBN), and (D) spikelet number per primary branch (SPPB) in the BC2F4 population.

Table 2

Pearson correlation coefficients between 4 panicle traits from 92 BC2F4 individuals

Traitsa	PBN	SBN	SPP	SPPB
PBN	1.000
SBN	0.127	1.000
SPP	0.166	0.965***	1.000
SPPB	−0.241*	0.893***	0.914***	1.000

PBN, primary branch number; SBN, secondary branch number; SPP, spikelet number per panicle; SPPB, spikelet number per primary branch.

0.01 < P < 0.05;

0.001 < P < 0.01;

P < 0.001.

Substitution mapping by RAD-seq

Trait-marker association was conducted to examine 17 low-SBN and 18 high-SBN BC2F4 individuals. RAD-seq resulted in ~1.23 × 108 single-end reads, from which 2,522 valid SNP markers were obtained for two parents and 37 BC2F4 individuals. Among these individuals, six chromosome intervals, including 1L (markers 12382–12807; 39.6–40.3 Mb), 4S (marker 58649; 6.7 Mb), 9S-1 (marker 100074; 1.2 Mb), 9S-2 (markers 105417–105650; 4.3–5.1 Mb), 9L (markers 106710–106787; 9.4–9.7 Mb) and 11L (markers 116181–116251; 18.2–18.4 Mb), exhibited homozygous substituted segments of IR65598-112-2; three chromosome intervals, 2L (markers 26476–26797; 32.0–32.6 Mb), 3S (marker 44580; 5.1 Mb) and 7L (markers 87830–87941; 29.0–29.2 Mp), showed segregating genotypes of IR65598-112-2 and TN13; and the remaining chromosome intervals were homozygous for TN13 (Fig. 3A). Linkage analysis using 18 high-SBN individuals and the markers targeting the three segregating segments (2L, 3S and 7L) revealed only one chromosome interval, 7L (markers 87830–87941; 29.0–29.2 Mb), co-segregating with the SBN phenotype (Fig. 3B). The SBN-associated locus on the long arm of chromosome 7 was designated qSBN7.

Fig. 3

Substitution mapping of secondary branch number (SBN) by restriction-site associated DNA sequencing (RAD-seq) and linkage mapping. (A) Graphical genotypes of 37 BC2F4 individuals. The 7 bars represent rice chromosomes, numbered at the top. The numbers on the left and right of the bars represent physical map positions (Mb) and the single nucleotide polymorphisms (SNPs) resulting from RAD-seq analysis, respectively. White squares: homozygous chromosomal segments of TN13; black squares: homozygous chromosomal segment of IR65598-112-2; grey squares: chromosomal segments still segregating in the BC2F4 population. (B) A rough map from linkage analysis based on 18 high-SBN type plants and the markers targeting three segregating segments (2L, 3S and 7L).

High-resolution mapping of qSBN7

To confirm the borders of the IR65598-112-2 introgression segment on chromosome 7, a BC2F4 individual with high-SBN phenotype was genotyped by using 8 additional CAPS markers that locate downstream of the marker 86544 (26.7 Mb) (Table 1). The markers SNP2750 (27,504,237 bp) and SNP2962 (29,627,810 bp) were identified to be the polymorphic markers closest to either end of the IR65598-112-2 introgression segments (Fig. 4A). Genotyping 166 high-SBN type plants (BC2F4) with SNP2750 and SNP2962 identified 26 recombinants, which were further genotyped with 6 additional CAPS markers in the interval. These BC2F4 recombinants were classified into 6 groups (A–F groups) based on the positions of recombination breakpoints (Fig. 4B). According to progeny test using BC2F5 lines, all groups showed a high-SBN phenotype. The A, B, C and D groups placed the qSBN7 locus to a region upstream of SNP2962, SNP2906, SNP2855 and SNP2849, respectively. The E and F groups located qSBN7 to a region downstream of SNP2788 and SNP2750, respectively. Thus, qSBN7 was located within a region of 601.4 kb between the markers SNP2788 (27,889,190 bp) and SNP2849 (28,490,603 bp) (Fig. 4B).

Fig. 4

Linkage map around the qSBN7 locus on chromosome 7. (A) The qSBN7 was mapped to the distal region of the long arm of chromosome 7 in 37 BC2F4 individuals. (B) High-resolution mapping of qSBN7 in additional 26 recombinants. White bars: homozygous TN13 chromosome segment; black bars: homozygous IR65598-112-2 chromosome segment; grey bars: heterozygous chromosome segment. The numbers in parenthesis are the mean SBN for each recombinant group. Different letters indicate significant differences based on Fisher’s least significant difference test at α = 0.05.

Phenotype evaluation of BILs for qSBN7

To characterize the effect of qSBN7, agronomic traits of BIL-sbn/sbn and BIL-Sbn/Sbn were compared (Table 3, Supplemental Table 1). For panicle traits, no significant difference in PN or PBN was observed between the two BILs. Nevertheless, values for PH, SPP, SBN and SPPB were significantly higher for BIL-sbn/sbn than BIL-Sbn/Sbn (Table 3). Compared to BIL-Sbn/Sbn, BIL-sbn/sbn showed an 83.2% increase in SPP and a 61.0% increase in SBN (Fig. 5). Moreover, SPP and SBN values were significantly higher for BIL-Sbn/Sbn than TN13 perhaps because of different alleles of GN1A in IR65598-112-2 and TN13. For grain traits, BIL-sbn/sbn showed significantly lower TGW and RGR values. No significant difference was observed for CGR (Supplemental Table 1).

Table 3

Agronomic traits of TN13 and two backcross inbred lines (BILs) carrying different alleles at qSBN7

Line	PH (cm)	Panicle traitsa
		PN	SPP	PBN	SBN	SPPB
TN13	78.7 ± 4.6 a	10.4 ± 2.0 a	143.3 ± 5.7 a	11.1 ± 2.3 a	27.2 ± 2.0 a	13.4 ± 3.1 a
BIL-Sbn/Sbn	79.5 ± 6.9 a	10.3 ± 1.9 a	167.5 ± 19.0 b	11.8 ± 2.3 a	34.1 ± 2.9 b	14.4 ± 1.8 a
BIL-sbn/sbn	85.7 ± 5.2 b	9.7 ± 2.0 a	306.8 ± 35.6 c	12.5 ± 0.8 a	54.9 ± 4.7 c	24.8 ± 4.4 b

PH, plant height; PN, panicle number per plant; SPP, spikelet number per panicle; PBN, primary branch number; SBN, secondary branch number; SPPB, spikelet number per primary branch. Data are mean ± SE. Different letters indicate significant differences based on Fisher’s least significant difference test at α = 0.05.

Fig. 5

Phenotype characteristics of TN13 and two backcross inbred lines (BILs) carrying different alleles at qSBN7.

Discussion

An increase in productivity has been one of the principal goals in rice breeding. Many major QTLs for yield-related traits have been identified and used for developing high-yield rice (Miura ). At the beginning of this study, we aimed to alter SPP in a japonica cultivar, TN13, by using marker-assisted backcrossing. The GN1A allele of IR65598-112-2 was introgressed into TN13. Of note, four BC2F4 populations with a homozygous GN1A allele of IR65598-112-2 showed segregation of SPP and SBN, which suggested the existence of another QTL(s) controlling SPP and SBN. The donor parent, IR65598-112-2, is an NPT rice cultivar with low tiller number, high SPP, high SBN, and strong culms (Khush and Peng 1996). The phenotypes of IR65598-112-2 are similar to plants with the IPA1/WFP allele at OsSPL14. High expression of OsSPL14 in the reproductive stage promotes panicle branching and grain yield in rice (Jiao , Miura ). OsSPL14 was not considered a causative QTL in our materials, because among F2 individuals derived from the cross of IR65598-112-2 and TN13, there was no significant association detected between SPP/SBN and an OsSPL14-linked marker (data not shown).

The current study located a novel QTL for SBN (qSBN7) at a region between 27.9–28.5 Mb on chromosome 7, and the SBN trait was completely co-segregated with the markers SNP2830 and SNP2835. Although several QTLs for SBN have been identified (Ando , Ikeda , Luo , Mei ), none was localized to the same region of qSBN7. A major QTL, qTSN7.1, for total spikelet number per panicle (TSN) (Koide ), located close to qSBN7 and resembled the gene effect of qSBN7. However, its position on chromosome 7 (24.0–26.0 Mb) (Koide ) is different from that of qSBN7 (27.9–28.5 Mb). It is worth noting that a pleiotropic QTL, qGL7, was mapped to the long arm of chromosome 7 (28.2–28.5 Mb) (Bai ), which shares a small overlapping region with qSBN7. As a major QTL for grain size, qGL7 simultaneously affects grain weight, length, width, thickness, and spikelet number per panicle (Bai ). The qSBN7 allele of IR65598-112-2 increased SBN and SPP and also reduced TGW and RGR. Whether qGL7 and qSBN7 are identical, allelic, or linked remains to be resolved.

qSBN7 is a pleiotropic QTL affecting multiple panicle traits. The homozygous IR65598-112-2 allele at qSBN7 significantly improved PH, SPP and SBN in the genetic background of TN13. High SPP potentially enhances the reproductive sink capacity of rice. Several SPP-related genes and QTLs have been identified (Ashikari , Fujita , Huang , Jiao , Miura , Xue ). Previous studies indicated that favorable alleles at two SPP-related genes, GN1A and SPIKE/NAL1, found widely in modern rice varieties, could improve rice grain yield (Wang , Yan ). GN1A is a major gene encoding OsCKX2. The nearly isogenic line (NIL) carrying the Habataki allele at GN1A in the Koshihikari genetic background showed a 20.7% increase in grain number per panicle (Ashikari ). SPIKE/NAL1 has pleiotropic effects on several components of plant architecture. The NIL carrying the IR68522-10-2-2 allele at SPIKE/NAL1 achieved a 13% to 36% increase in yield (Fujita ). This study revealed a newly discovered and fine-mapped SPP/SBN-related QTL, qSBN7, for high-sink capacity breeding in rice. Because qSBN7 has a similar effect as GN1A and SPIKE/NAL1 and it negatively affected TGW and RGR, the sink–source balance should be considered when pyramiding qSBN7 with other SPP- and grain-filling-related genes. In addition, the phenotype of SPP is easily affected by genetic and environmental factors (Xing ). To facilitate the use of qSBN7 in rice breeding, future work should focus on understanding the effect of qSBN7 in diverse genetic backgrounds and/or environments.