A cell wall-localized cytokinin/purine riboside nucleosidase is involved in apoplastic cytokinin metabolism in Oryza sativa.

In the final step of cytokinin biosynthesis, the main pathway is the elimination of a ribose-phosphate moiety from the cytokinin nucleotide precursor by phosphoribohydrolase, an enzyme encoded by a gene named LONELY GUY (LOG). This reaction accounts for most of the cytokinin supply needed for regulating plant growth and development. In contrast, the LOG-independent pathway, in which dephosphorylation and deribosylation sequentially occur, is also thought to play a role in cytokinin biosynthesis, but the gene entity and physiological contribution have been elusive. In this study, we profiled the phytohormone content of chromosome segment substitution lines of Oryza sativa and searched for genes affecting the endogenous levels of cytokinin ribosides by quantitative trait loci analysis. Our approach identified a gene encoding an enzyme that catalyzes the deribosylation of cytokinin nucleoside precursors and other purine nucleosides. The cytokinin/purine riboside nucleosidase 1 (CPN1) we identified is a cell wall-localized protein. Loss-of-function mutations (cpn1) were created by inserting a Tos17-retrotransposon that altered the cytokinin composition in seedling shoots and leaf apoplastic fluid. The cpn1 mutation also abolished cytokinin riboside nucleosidase activity in leaf extracts and attenuated the trans-zeatin riboside-responsive expression of cytokinin marker genes. Grain yield of the mutants declined due to altered panicle morphology under field-grown conditions. These results suggest that the cell wall-localized LOG-independent cytokinin activating pathway catalyzed by CPN1 plays a role in cytokinin control of rice growth. Our finding broadens our spatial perspective of the cytokinin metabolic system.

Development of 12 sets of chromosome segment substitution lines that enhance allele mining in Asian cultivated rice.

Many agronomic traits that are important in rice breeding are controlled by multiple genes. The extensive time and effort devoted so far to identifying and selecting such genes are still not enough to target multiple agronomic traits in practical breeding in Japan because of a lack of suitable plant materials in which to efficiently detect and validate beneficial alleles from diverse genetic resources. To facilitate the comprehensive analysis of genetic variation in agronomic traits among Asian cultivated rice, we developed 12 sets of chromosome segment substitution lines (CSSLs) with the japonica background, 11 of them in the same genetic background, using donors representing the genetic diversity of Asian cultivated rice. Using these materials, we overviewed the chromosomal locations of 1079 putative QTLs for seven agronomic traits and their allelic distribution in Asian cultivated rice through multiple linear regression analysis. The CSSLs will allow the effects of putative QTLs in the highly homogeneous japonica background to be validated.

Genomic adaptation of flowering-time genes during the expansion of rice cultivation area.

The diversification of flowering time in response to natural environments is critical for the spread of crops to diverse geographic regions. In contrast with recent advances in understanding the molecular basis of photoperiodic flowering in rice (Oryza sativa), little is known about how flowering-time diversification is structured within rice subspecies. By analyzing genome sequencing data and a set of 429 chromosome segment substitution lines (CSSLs) originating from 10 diverse rice accessions with wide distributions, we revealed diverse effects of allelic variations for common flowering-time quantitative trait loci in the recipient's background. Although functional variations associated with a few loci corresponded to standing variations among subspecies, the identified functional nucleotide polymorphisms occurred recently after rice subgroup differentiation, indicating that the functional diversity of flowering-time gene sequences was not particularly associated with phylogenetic relationship between rice subspecies. Intensive analysis of the Hd1 genomic region identified the signature of an early introgression of the Hd1 with key mutation(s) in aus and temperate japonica accessions. Our data suggested that, after such key introgressions, new mutations were selected and accelerated the flowering-time diversity within subspecies during the expansion of rice cultivation area. This finding may imply that new genome-wide changes for flowering-time adaptation are one of the critical determinants for establishing genomic architecture of local rice subgroups. In-depth analyses of various rice genomes coupling with the genetically confirmed phenotypic changes in a large set of CSSLs enabled us to demonstrate how rice genome dynamics has coordinated with the adaptation of cultivated rice during the expansion of cultivation area.

Late flowering in F1 hybrid rice brought about by the complementary effect of quantitative trait loci.

Late flowering sometimes occurs in F1 hybrids between rice varieties (Oryza sativa L.), although the parental varieties show similar days-to-flowering (DTF). The genetic architecture prompting the occurrence of such late flowering is poorly understood. To clarify the genetic architecture of late flowering in F1 hybrids from a cross between rice varieties, 'Koshihikari' and 'IR64', we performed quantitative trait locus (QTL) analysis using an F2 population (selfed progeny of an F1 plant), in which heterozygous genotypes should segregate in a certain proportion in a Mendelian manner. The QTL analysis detected three significant QTLs. At one QTL (putatively Heading date 1), the 'Koshihikari' allele increased DTF, and at the other two QTLs (putatively Heading date 6 and Oryza sativa Pseudo-Response Regulator 37/Heading date 2), the 'IR64' alleles increased DTF. All alleles at these three QTLs showed partial dominance. The combination of the QTLs explained 82.2% of the total phenotypic variance of DTF in the F2 population, with contribution from epistasis between QTLs. There was no difference between DTFs of F1 hybrids and heterozygous genotypes for the three QTLs. Our results demonstrated that the complementary effects accompanied by epistasis of at least three QTLs were responsible for late flowering in F1 hybrids.

Switching genetic effects of the flowering time gene Hd1 in LD conditions by Ghd7 and OsPRR37 in rice.

Flowering time control in plants is a major limiting factor on the range of species. Day length, perceived via the photoperiodic pathway, is a critical factor for the induction of flowering. The module of GIGANTEA (GI)-CONSTANS (CO)-FLOWERING LOCUS T in the long day (LD) plant Arabidopsis is conserved in diverse plant species including the short day (SD) plant rice, where this module comprises OsGI-Heading date 1 (Hd1)-Heading date 3a. Hd1, the rice ortholog of Arabidopsis CO, has dual functions in the regulation of flowering time, promoting flowering in SD conditions and delaying it in LD conditions. We herein show genetic interactions among three LD repressor genes: Hd1, Grain number, plant height and heading date 7 (Ghd7), and Oryza sativa Pseudo-Response Regulator37 (OsPRR37). Genetic analyses, including segregation analyses, evaluations of near isogenic lines, and transformation for flowering time demonstrated that Hd1 promoted flowering time in inductive SD and non-inductive LD conditions in genetic condition of loss-of-function Ghd7 and OsPRR37 (ghd7osprr37) in rice. Functional Ghd7 or OsPRR37 may switch the genetic effects of Hd1 from the promotion to the delay of flowering times in LD conditions.

Detection of heading date QTLs in advanced-backcross populations of an elite indica rice cultivar, IR64.

IR64 is one of the world's most popular rice cultivars. To collect genetic factors involved in controlling its heading date, we developed 70 reciprocal advanced-backcross populations with a total of 6284 individuals at the BC4F2 generation from crosses between Koshihikari and IR64. We detected 29 QTLs associated with heading date on chromosomes 3, 5-8, 10, and 12. Twenty QTLs were located in the same chromosome regions as previously isolated heading date genes (Hd1, Hd6, Hd16, Ghd7, DTH8, Hd17, and Hd18). The rest were located in other chromosome regions. We found more number of QTLs than previous studies using mapping populations of IR64. Fine mapping in additional advanced-backcross populations clearly revealed that QTLs on the long arm of chromosome 7 are overlapping and seem to be a novel genetic factor for heading date because of their different locations from OsPRR37. Our results suggest that the difference in heading date between IR64 and Koshihikari is genetically controlled by many factors, and that a non-functional allele of Hd1 contributes to early heading of IR64 in the genetic background of functional alleles of other heading date QTLs and genes such as Hd6 and Hd16.

The Birth of a Black Rice Gene and Its Local Spread by Introgression.

The origin and spread of novel agronomic traits during crop domestication are complex events in plant evolution. Wild rice (Oryza rufipogon) has red grains due to the accumulation of proanthocyanidins, whereas most cultivated rice (Oryza sativa) varieties have white grains induced by a defective allele in the Rc basic helix-loop-helix (bHLH) gene. Although the events surrounding the origin and spread of black rice traits remain unknown, varieties with black grains due to anthocyanin accumulation are distributed in various locations throughout Asia. Here, we show that the black grain trait originated from ectopic expression of the Kala4 bHLH gene due to rearrangement in the promoter region. Both the Rc and Kala4 genes activate upstream flavonol biosynthesis genes, such as chalcone synthase and dihydroflavonol-4-reductase, and downstream genes, such as leucoanthocyanidin reductase and leucoanthocyanidin dioxygenase, to produce the respective specific pigments. Genome analysis of 21 black rice varieties as well as red- and white-grained landraces demonstrated that black rice arose in tropical japonica and its subsequent spread to the indica subspecies can be attributed to the causal alleles of Kala4. The relatively small size of genomic fragments of tropical japonica origin in some indica varieties indicates that refined introgression must have occurred by natural crossbreeding in the course of evolution of the black trait in rice.

Genetic dissection of pre-harvest sprouting resistance in an upland rice cultivar.

Seed dormancy is important in rice breeding because it confers resistance to pre-harvest sprouting (PHS). To detect quantitative trait loci (QTLs) for pre-harvest sprouting resistance, we used chromosome segment substitution lines (CSSLs) derived from a cross between the Japanese upland rice cultivar 'Owarihatamochi' and the lowland rice cultivar 'Koshihikari'. In the CSSLs, several chromosomal regions were associated with PHS resistance. Among these, the chromosome 9 segment from 'Owarihatamochi' had the greatest association with increased PHS resistance. Further QTL analysis using an advanced backcross population (BC4F2) derived from a 'Koshihikari' × 'Owarihatamochi' cross revealed two putative QTLs, here designated qSDR9.1 (Seed dormancy 9.1) and qSDR9.2, on chromosome 9. The 'Owarihatamochi' alleles of the two QTLs reduced germination. Further fine mapping revealed that qSDR9.1 and qSDR9.2 were located within 4.1-Mb and 2.3-Mb intervals (based on the 'Nipponbare' reference genome sequence) defined by the simple sequence repeat marker loci RM24039 and RM24260 and Indel_2 and RM24540, respectively. We thus identified two QTLs for PHS resistance in 'Owarihatamochi', even though resistance levels are relatively low in this cultivar. This unexpected finding suggests the advantages of using CSSLs for QTL detection.

Hd1,a CONSTANS ortholog in rice, functions as an Ehd1 repressor through interaction with monocot-specific CCT-domain protein Ghd7.

Flowering time is an important agronomic trait that affects crop yields. In cereals, several CCT-domain proteins unique to monocots, such as Grain number, plant height, and heading date 7 (Ghd7) gene, have been identified as key floral repressors, although the corresponding molecular mechanisms have been unknown until now. In rice, a short-day plant, Heading date 1 (Hd1) gene, a rice ortholog of Arabidopsis floral activator CONSTANS (CO), represses flowering under non-inductive long-day (LD) conditions and induces it under inductive short-day (SD) conditions. Here, we report biological interactions between Ghd7 and Hd1, which together repress Early heading date 1 (Ehd1), a key floral inducer under non-inductive LD conditions. In addition to this genetic interaction between them, Co-IP experiments further demonstrated that a Ghd7-Hd1 protein formed a complex in vivo and ChIP and luciferase reporter analyses suggested that this complex specifically binds to a cis-regulatory region in Ehd1 and represses its expression. These findings imply that Hd1, an evolutionally conserved transcriptional activator, can function as a strong transcriptional repressor within a monocot-specific flowering-time pathway through with Ghd7.

Genetic dissection of agronomically important traits in closely related temperate japonica rice cultivars.

Many quantitative trait loci (QTLs) for agronomically important traits such as grain yield, disease resistance, and stress tolerance of rice (Oryza sativa L.) have been detected by using segregating populations derived from crosses between indica and japonica subspecies or with wild relatives. However, the QTLs involved in the control of natural variation in agronomic traits among closely related cultivars are still unclear. Decoding the whole genome sequences of Nipponbare and other temperate japonica rice cultivars has accelerated the collection of a huge number of single nucleotide polymorphisms (SNPs). These SNPs are good resource for developing polymorphic DNA markers and for detecting QTLs distributed across all rice chromosomes. The temperate japonica rice cultivar Koshihikari has remained the top cultivar for about 40 years since 1979 in Japan. Unraveling the genetic factors in Koshihikari will provide important insights into improving agronomic traits in temperate japonica rice cultivars. Here we describe recent progress in our studies as an example of genetic analysis in closely related cultivars.

Detection of novel QTLs qDTH4.5 and qDTH6.3, which confer late heading under short-day conditions, by SSR marker-based and QTL-seq analysis.

Heading date is one of the most important traits in rice breeding. It is governed by multiple genes, including known quantitative trait loci (QTLs). In general, almost all japonica cultivars, including Nipponbare, head early under short-day (SD) conditions, but some indica cultivars, including Kasalath, head late. To explain this difference, we identified QTLs controlling heading date under SD conditions. We used NILs, CSSLs, and BILs from a cross between Nipponbare and Kasalath, and evaluated days to heading (DTH) under SD conditions. No NILs or CSSLs showed late heading, but two BILs (BIL-55 and BIL-78) had almost the same DTH as Kasalath. We developed an F2 population from a cross between BIL-55 and Nipponbare and performed QTL analysis using SSR markers. The late-heading phenotype was controlled by two known genes and at least two novel QTLs on chromosomes 4 and 6, named qDTH4.5 and qDTH6.3. These QTLs were confirmed by QTL-seq. The QTLs and polymorphisms detected here will provide useful information for further genetic studies and breeding under SD conditions at lower latitudes.

Metabolome-genome-wide association study dissects genetic architecture for generating natural variation in rice secondary metabolism.

Plants produce structurally diverse secondary (specialized) metabolites to increase their fitness for survival under adverse environments. Several bioactive compounds for new drugs have been identified through screening of plant extracts. In this study, genome-wide association studies (GWAS) were conducted to investigate the genetic architecture behind the natural variation of rice secondary metabolites. GWAS using the metabolome data of 175 rice accessions successfully identified 323 associations among 143 single nucleotide polymorphisms (SNPs) and 89 metabolites. The data analysis highlighted that levels of many metabolites are tightly associated with a small number of strong quantitative trait loci (QTLs). The tight association may be a mechanism generating strains with distinct metabolic composition through the crossing of two different strains. The results indicate that one plant species produces more diverse phytochemicals than previously expected, and plants still contain many useful compounds for human applications.

Hd18, Encoding Histone Acetylase Related to Arabidopsis FLOWERING LOCUS D, is Involved in the Control of Flowering Time in Rice.

Flowering time is one of the most important agronomic traits in rice (Oryza sativa L.), because it defines harvest seasons and cultivation areas, and affects yields. We used a map-based strategy to clone Heading date 18 (Hd18). The difference in flowering time between the Japanese rice cultivars Koshihikari and Hayamasari was due to a single nucleotide polymorphism within the Hd18 gene, which encodes an amine oxidase domain-containing protein and is homologous to Arabidopsis FLOWERING LOCUS D (FLD). The Hayamasari Hd18 allele and knockdown of Hd18 gene expression delayed the flowering time of rice plants regardless of the day-length condition. Structural modeling of the Hd18 protein suggested that the non-synonymous substitution changed protein stability and function due to differences in interdomain hydrogen bond formation. Compared with those in Koshihikari, the expression levels of the flowering-time genes Early heading date 1 (Ehd1), Heading date 3a (Hd3a) and Rice flowering locus T1 (RFT1) were lower in a near-isogenic line with the Hayamasari Hd18 allele in a Koshihikari genetic background. We revealed that Hd18 acts as an accelerator in the rice flowering pathway under both short- and long-day conditions by elevating transcription levels of Ehd1 Gene expression analysis also suggested the involvement of MADS-box genes such as OsMADS50, OsMADS51 and OsMADS56 in the Hd18-associated regulation of Ehd1 These results suggest that, like FLD, its rice homolog accelerates flowering time but is involved in rice flowering pathways that differ from the autonomous pathways in Arabidopsis.

Global agricultural intensification during climate change: a role for genomics.

Agriculture is now facing the 'perfect storm' of climate change, increasing costs of fertilizer and rising food demands from a larger and wealthier human population. These factors point to a global food deficit unless the efficiency and resilience of crop production is increased. The intensification of agriculture has focused on improving production under optimized conditions, with significant agronomic inputs. Furthermore, the intensive cultivation of a limited number of crops has drastically narrowed the number of plant species humans rely on. A new agricultural paradigm is required, reducing dependence on high inputs and increasing crop diversity, yield stability and environmental resilience. Genomics offers unprecedented opportunities to increase crop yield, quality and stability of production through advanced breeding strategies, enhancing the resilience of major crops to climate variability, and increasing the productivity and range of minor crops to diversify the food supply. Here we review the state of the art of genomic-assisted breeding for the most important staples that feed the world, and how to use and adapt such genomic tools to accelerate development of both major and minor crops with desired traits that enhance adaptation to, or mitigate the effects of climate change.

Genetic control of flowering time in rice: integration of Mendelian genetics and genomics.

KEY MESSAGE: Integration of previous Mendelian genetic analyses and recent molecular genomics approaches, such as linkage mapping and QTL cloning, dramatically strengthened our current understanding of genetic control of rice flowering time. Flowering time is one of the most important agronomic traits for seed production in rice (Oryza sativa L.). It is controlled mainly by genes associated with photoperiod sensitivity, particularly in short-day plants such as rice. Since the early twentieth century, rice breeders and researchers have been interested in elucidating the genetic basis of flowering time because its modification is important for regional adaptation and yield optimization. Although flowering time is a complex trait controlled by many quantitative trait loci (QTLs), classical genetic studies have shown that many associated genes are inherited in accordance with Mendelian laws. Decoding the rice genome sequence opened a new era in understanding the genetic control of flowering time on the basis of genome-wide mapping and gene cloning. Heading date 1 (Hd1) was the first flowering time QTL to be isolated using natural variation in rice. Recent accumulation of information on rice genome has facilitated the cloning of other QTLs, including those with minor effects on flowering time. This information has allowed us to rediscover some of the flowering genes that were identified by classical Mendelian genetics. The genes characterized so far, including Hd1, have been assigned to specific photoperiod pathways. In this review, we provide an overview of the studies that led to an in-depth understanding of the genetic control of flowering time in rice, and of the current state of improving and fine-tuning this trait for rice breeding.

RCN1/OsABCG5, an ATP-binding cassette (ABC) transporter, is required for hypodermal suberization of roots in rice (Oryza sativa).

Suberin is a complex polymer composed of aliphatic and phenolic compounds. It is a constituent of apoplastic plant interfaces. In many plant species, including rice (Oryza sativa), the hypodermis in the outer part of roots forms a suberized cell wall (the Casparian strip and/or suberin lamellae), which inhibits the flow of water and ions and protects against pathogens. To date, there is no genetic evidence that suberin forms an apoplastic transport barrier in the hypodermis. We discovered that a rice reduced culm number1 (rcn1) mutant could not develop roots longer than 100 mm in waterlogged soil. The mutated gene encoded an ATP-binding cassette (ABC) transporter named RCN1/OsABCG5. RCN1/OsABCG5 gene expression in the wild type was increased in most hypodermal and some endodermal roots cells under stagnant deoxygenated conditions. A GFP-RCN1/OsABCG5 fusion protein localized at the plasma membrane of the wild type. Under stagnant deoxygenated conditions, well suberized hypodermis developed in wild types but not in rcn1 mutants. Under stagnant deoxygenated conditions, apoplastic tracers (periodic acid and berberine) were blocked at the hypodermis in the wild type but not in rcn1, indicating that the apoplastic barrier in the mutant was impaired. The amount of the major aliphatic suberin monomers originating from C(28) and C(30) fatty acids or ω-OH fatty acids was much lower in rcn1 than in the wild type. These findings suggest that RCN1/OsABCG5 has a role in the suberization of the hypodermis of rice roots, which contributes to formation of the apoplastic barrier.

Genetic architecture of variation in heading date among Asian rice accessions.

BACKGROUND: Heading date, a crucial factor determining regional and seasonal adaptation in rice (Oryza sativa L.), has been a major selection target in breeding programs. Although considerable progress has been made in our understanding of the molecular regulation of heading date in rice during last two decades, the previously isolated genes and identified quantitative trait loci (QTLs) cannot fully explain the natural variation for heading date in diverse rice accessions. RESULTS: To genetically dissect naturally occurring variation in rice heading date, we collected QTLs in advanced-backcross populations derived from multiple crosses of the japonica rice accession Koshihikari (as a common parental line) with 11 diverse rice accessions (5 indica, 3 aus, and 3 japonica) that originate from various regions of Asia. QTL analyses of over 14,000 backcrossed individuals revealed 255 QTLs distributed widely across the rice genome. Among the detected QTLs, 128 QTLs corresponded to genomic positions of heading date genes identified by previous studies, such as Hd1, Hd6, Hd3a, Ghd7, DTH8, and RFT1. The other 127 QTLs were detected in different chromosomal regions than heading date genes. CONCLUSIONS: Our results indicate that advanced-backcross progeny allowed us to detect and confirm QTLs with relatively small additive effects, and the natural variation in rice heading date could result from combinations of large- and small-effect QTLs. We also found differences in the genetic architecture of heading date (flowering time) among maize, Arabidopsis, and rice.

Chromosomal locations of a gene underlying heat-accelerated brown spot formation and its suppressor genes in rice.

Brown spots on mature leaves from the heading to ripening stages in rice are considered to be lesions induced by heat stress. However, there are few studies of lesions that are induced by heat stress rather than by pathogen infections. To understand the genetic background underlying such lesions, we used the chromosome segment substitution line (CSSL) SL518, derived from a distant cross between rice cultivars Koshihikari (japonica) and Nona Bokra (indica). We observed brown spots on mature leaf blades of the CSSL, although the parents barely showed any spots. Spot formation in SL518 was accelerated by high temperature, suggesting that the candidate gene for spot formation is related to heat stress response. Using progeny derived from a cross between SL518 and Koshihikari, we mapped the causative gene, BROWN-SPOTTED LEAF 1 (BSPL1), on chromosome 5. We speculated that one or more Nona Bokra genes suppress spot formation caused by BSPL1 and identified candidate genomic regions on chromosomes 2 and 9 using a cross between a near-isogenic line for BSPL1 and other CSSLs possessing Nona Bokra segments in the Koshihikari genetic background. In conclusion, our data support the concept that multiple genes are complementarily involved in brown spot formation induced by heat stress and will be useful for cloning of the novel gene(s) related to the spot formation.

A novel rice cytochrome P450 gene, CYP72A31, confers tolerance to acetolactate synthase-inhibiting herbicides in rice and Arabidopsis.

Target-site and non-target-site herbicide tolerance are caused by the prevention of herbicide binding to the target enzyme and the reduction to a nonlethal dose of herbicide reaching the target enzyme, respectively. There is little information on the molecular mechanisms involved in non-target-site herbicide tolerance, although it poses the greater threat in the evolution of herbicide-resistant weeds and could potentially be useful for the production of herbicide-tolerant crops because it is often involved in tolerance to multiherbicides. Bispyribac sodium (BS) is an herbicide that inhibits the activity of acetolactate synthase. Rice (Oryza sativa) of the indica variety show BS tolerance, while japonica rice varieties are BS sensitive. Map-based cloning and complementation tests revealed that a novel cytochrome P450 monooxygenase, CYP72A31, is involved in BS tolerance. Interestingly, BS tolerance was correlated with CYP72A31 messenger RNA levels in transgenic plants of rice and Arabidopsis (Arabidopsis thaliana). Moreover, Arabidopsis overexpressing CYP72A31 showed tolerance to bensulfuron-methyl (BSM), which belongs to a different class of acetolactate synthase-inhibiting herbicides, suggesting that CYP72A31 can metabolize BS and BSM to a compound with reduced phytotoxicity. On the other hand, we showed that the cytochrome P450 monooxygenase CYP81A6, which has been reported to confer BSM tolerance, is barely involved, if at all, in BS tolerance, suggesting that the CYP72A31 enzyme has different herbicide specificities compared with CYP81A6. Thus, the CYP72A31 gene is a potentially useful genetic resource in the fields of weed control, herbicide development, and molecular breeding in a broad range of crop species.

Hybrid breakdown caused by epistasis-based recessive incompatibility in a cross of rice (Oryza sativa L.).

Viability and fertility in organisms depend on epistatic interactions between loci maintained in lineages. Here, we describe reduced fitness of segregants (hybrid breakdown, HB) that emerged in an F2 population derived from a cross between 2 rice (Oryza sativa L.) cultivars, "Tachisugata" (TS) and "Hokuriku 193" (H193), despite both parents and F1s showing normal fitness. Quantitative trait locus (QTL) analyses detected 13 QTLs for 4 morphological traits associated with the HB and 6 associated with principal component scores calculated from values of the morphological traits in the F2 population. Two-way analysis of variance of the putative QTLs identified 4 QTL pairs showing significant epistasis; among them, a pair on chromosomes 1 and 12 made the greatest contribution to HB. The finding was supported by genetic experiments using F3 progeny. HB emerged only when a plant was homozygous for the TS allele at the QTL on chromosome 1 and homozygous for the H193 allele at the QTL on chromosome 12, indicating that each allele behaves as recessive to the other. Our results support the idea that epistasis is an essential part of hybrid fitness.