**3. Discussion**

Since its generation in 1973, hybrid rice has predominated *indica* rice production in China. Numbers of elite combinations have been developed and released as commercial varieties, with yields roughly 20% higher than their inbred counterparts [15]. Recent molecular research has investigated the number of hybrid combinations to construct a model system for studying the molecular mechanism of heterosis for three- and two-line hybrids. A study using the yield components data and an ultra-high-density SNP bin map of an immortalized F2 population derived from the cross between Zhenshan97 and Minghui63 demonstrated that the relative contributions of the genetic components vary with traits. The results indicate that overdominance/pseudo-overdominance are most important to the heterosis

of grain number per panicle, 1000-grain weight, and grain yield per plant. In heterosis of panicle number and 1000-grain weight, the dominance × dominance interaction is important. Among these yield-related traits, single-locus dominance has relatively small contributions [16]. An integrated analysis using the RIL population and RILBCF1 population derived from the cross between PA64S (which has a mix genetic background of *indica* and *javanica*) and 93-11 showed that heterosis was mainly detected in grain number per panicle and panicle number [8]. Huang et al. sequenced 10,074 F2 lines derived from 17 representative hybrid combinations, and they found that a small number of genomic loci from female parents explain a large proportion of the yield advantage of hybrids over their male parents [9]. Taken together, these studies improve our understanding of heterosis. However, research on heterosis between *japonica* and *japonica* crosses is limited. The genetic diversity of *japonica* was not as high as that of *indica*, making heterosis among *japonica* inconspicuous compared with that among *indica*. Moreover, several factors have limited the utilization of heterosis between *indica* and *japonica*, which includes sterility. Thus, we used a BIL population derived from an *indica* and *japonica* cross to improve the genetic diversity but maintain the genetic background close to *japonica*. The results of the present study showed that heterosis between Sasanishiki and Habataki mainly came from a complicated quantitative and components-specific phenotype. Here, we clearly demonstrate yield heterosis, mainly by the outperformance of grain number per panicle and panicle number. Yield components survey also showed that the grain number per panicle of BILF1 population was significantly higher than that of the BIL population. Our previous study demonstrated that the introgression of *indica* pedigree in the *japonica* genome contributed to the increase of rice production in northern China [17], and thus the present study confirmed that the *indica* pedigree could increase the grain number per panicle in *japonica*.

Using high-throughput sequencing, we conducted QTL mapping of the BIL and BILF1 populations, and the number of QTLs that are responsible for yield and yield-related heterosis was determined. A total of 10 QTLs for all of the traits were mapped independently in BIL population (Figure 4), and eight loci for the respective phenotypes were detected. Among the eight QTLs, four of these loci overlapped with QTLs detected in the BIL population, and the remaining four QTLs were detected exclusively in the BILF1 population, indicating a mode of dominance or super-dominance in the heterozygote. At the gene level, Gao et al. demonstrated that a heading-time-regulated gene *Day to Heading 8 (DTH8)* is a candidate locus for yield heterosis in Liang-you-pei 9 (LYP9) [18], and Li et al. and Huang et al. confirmed that *DTH8* corresponds to yield heterosis [8,9]. These results suggest that the heading time gene strongly participates in yield heterosis in hybrid rice. The present study detected that the heading-time gene *Ghd7* is also a candidate locus for both yield heterosis and heading time regulation. A study using the BILs derived from the cross between Habataki and Koshihikari identified five QTLs (*Gn1*–*Gn5*) that were related to grain number per panicle on chromosome 1, 4, 10, and 12 [19]. The present study confirmed that *Gn1* and *Gn3* correspond to grain number per panicle. Moreover, *Gn3,* on the short arm of chromosome 10, also corresponded to grain number per panicle in BILF1s. Thus, this QTL may be further assessed in terms of heterosis.

#### **4. Materials and Methods**
