Identification of Heterotic Loci with Desirable Allelic Interaction to Increase Yield in Rice

Heterosis denotes the superiority of a hybrid plant over its parents. The use of heterosis has contributed significantly to yield improvement in crops. However, the genetic and molecular bases on heterosis are not fully understood. A large number of heterotic loci were identified for 12 yield-related traits in one parental population of chromosome segment substitution lines (CSSLs) and two test populations, which were interconnected by CSSLs derived from two rice genome-sequenced cultivars, Nipponbare and Zhenshan 97. Seventy-five heterotic loci were identified in both homozygous background of Zhenshan 97 and heterogeneous background of an elite hybrid cultivar Shanyou 63. Among the detected loci, at least 11 were colocalized in the same regions encompassing previously reported heterosis-associated genes. Furthermore, a heterotic locus Ghd8NIP for yield advantage was verified using transgenic experiments. Various allelic interaction at Ghd8 exhibited different heterosis levels in hetero-allelic combinations of five near-isogenic lines that contain a particular allele. The significant overdominance effects from some hetero-allelic combinations were found to improve yield heterosis in hybrid cultivars. Our findings support the role of allelic interaction at heterotic loci in the improvement of yield potential, which will be helpful for dissecting the genetic basis of heterosis and provide an optional strategy for the allele replacement in molecular breeding programs in hybrid rice.


Background
Heterosis or hybrid vigor refers to the phenomenon for which a hybrid markedly outperforms its parents. The use of heterosis in crops such as rice and maize have contributed significantly to the improvement of yield. Massive efforts have been made in exploring the genetic mechanisms of heterosis, leading to propose three main genetic models for heterosis (Chen 2013;Zhang et al. 2021), which include dominance (Xiao et al. 1995) and overdominance (Li et al. 2001;Gjuvsland et al. 2010;Larièpe et al. 2012) at a single-locus level and epistasis at two or more loci (Yu et al. 1997;Zhou et al. 2012). The dominance and overdominance effects reflected as an allelic interaction at a single locus in hybrids may result in an alteration of expression profiles or protein-protein interaction (Shao et al. 2019). At the same time, epistasis is referred to as the modification of a gene by one or several other genes (Birchler et al. 2010;Schnable and Springer 2013). However, the genetic and molecular mechanisms of heterosis in plants are not fully understood.
Numerous quantitative trait loci (QTLs) have been identified for heterosis in various plants (Li et al. 2001; the identification of heterotic loci or genes support the significant role of the allelic interactions at both single locus and multiple loci in plant heterosis (Huang et al. 2015;Li et al. 2016;Liu et al. 2020). For example, SIN-GLE FLOWER TRUSS (SFT) in tomato (Lycopersicon esculentum Mill.) is firstly identified as a single overdominant gene for yield (Krieger et al. 2010). In rice (Oryza sativa L.), Heading date 3a (Hd3a) has been identified to be responsible for heterosis in a large-scale F 2 population . The gene Ideal Plant Architecture1 (IPA1) that regulates plant architecture enhancing yield heterosis in rice, displays a strong overdominance effect from IPA1 heterozygotes . Similarly, the ortholog gene of IPA1 in maize (Zea mays L.), unbranched 3 (ub3) has also been identified as the major candidate gene for heterosis advantage in three F 2 populations (Liu et al. 2020). The other gene OsMADS1 showed an incomplete dominance for grain size in rice and near-isogenic lines containing a particular alleles could highly increase grain yield by more than 8% (Wang et al. 2019a). Since the development of genome re-sequencing of germplasms allows us capture various allelic variations at any gene of interest, the intriguing issue arises of how to explore optional allelic combinations that could produce higher heterosis performance in hybrids. However, few cases where diverse effects of various allelic interactions at a heterotic locus have been reported.
Rice is the staple crop, contributing for nearly half of worldwide population's food consumption (Elert 2014). The increase of rice yield is critical to ensuring global food security. Exploitation of interspecies hybrid vigor between indica/xian and japonica/geng has been a promising approach to enhance rice yield potential (Qian et al. 2016;Tao et al. 2016). The genetic dissection of hybrid vigor could facilitate this efficient exploitation. However, the precisely identification of heterotic loci has been limited in many segregating populations such as F 2 and recombinant inbred lines because of some epistatic or high-order gene interactions in a complex genetic background (Yu et al. 1997;Li et al. 2001). With the advantage of chromosome segment substitution lines (CSSLs), each containing only one or a few introduced donor segments in the same background, the genetic effect and contribution of a gene can be confirmed without the effect of other loci in genome (Ali et al. 2010;Zhang et al. 2021). Thus, CSSLs and their derived populations could provide one of the optimal approaches for dissecting the genetic factor of heterosis at the single-locus level . In this study, the objectives are to detect heterotic loci using CSSLs and their derived testing populations of backcross and testcross, and to unravel allelic interaction at a given heterotic loci using near-isogenic lines in rice.
As a result, a number of heterotic loci (HLs) were identified for yield and yield-related traits in the backcross (BC) and testcross (TC) populations. Among them, a major heterotic gene (Ghd8) was verified through transgenic experiments. Moreover, different interaction effects arose from various allelic combinations of Ghd8 were found associated with yield heterosis. Our findings of heterotic loci with favorable allelic combinations provide new insights into the genetic basis of heterosis. This will be useful for improving yield by hybrid rice breeding.

Plant Materials
Three mapping populations were developed and used in this study. The first population of 146 CSSLs was developed from a cross between two genome-sequenced rice cultivars, japonica Nipponbare (NIP) as the donor and elite indica Zhenshan 97 (ZS97) as the recurrent parent, using a backcross scheme of at least four times backcrossing with a marker-assisted selection (MAS) approach. The backcross population contained 146 F 1 hybrids (CSSL × ZS97) derived by backcrossing each CSSL with the recurrent parent ZS97. The elite restorer line Minghui 63 (MH63) as the male parent was testcrossed with each CSSL to generate the testcross population. The parental lines (MH63, NIP, and ZS97) and hybrid cultivar Shanyou 63 (SY63, MH63 × ZS97) were used as controls in the phenotype experiments.

Experimental Design and Phenotypic Evaluation
Two experiments were conducted to dissect of the genetic basis of heterosis in rice. First, three mapping populations were used to analyze the allelic effect at a single locus. A CSSL and BC were used to identify the mid-parent heterotic loci (HL MP ) in the homozygous background of ZS97. At the same time, a TC along with a check (SY63) were used to detect the over-standard heterotic loci (HL OS ) in a similar heterozygous background of SY63. Second, to determine the interaction of different allelic combinations at Ghd8, an half-diallel mating design was used to generate ten allelic combinations with five parental lines (NIL-Ghd8 ZS97 , NIL-Ghd8 NIP , NIL-Ghd8 9311 , NIL-Ghd8 ACC10 , and NIL-Ghd8 MH63 ). The five near-isogenic lines (NILs) carrying different Ghd8 alleles were developed independently from the cross of four parents (NIP, 9311, ACC10, and MH63) as donors and ZS97 as the recurrent parent using a MAS backcross scheme. These NILs have the common background of ZS97.
All the lines were grown at the experimental station of Huazhong Agricultural University in Wuhan (30.48° N,114.2° E), China. A randomized complete block design was carried out with two replications for three mapping populations in 2006 (E1) and 2007 (E2), respectively. The same field experimental design with three replications was used for the ten allelic combinations and NILs. Each line was planted in four rows with 10 individuals per row at a spacing of 16.7 × 26.6 cm. The eight plants in the middle of each row were harvested individually at maturity and used for scoring traits. The field was managed according to local standard practices.

Transgenic Test of Ghd8 Effects on Heterosis
The transgenic experiments were performed to analyze Ghd8 effects on heterosis. Homozygous complementary transgenic lines with the NIP alleles of Ghd8 (Ghd8 NIP ) introduced into ZS97 (here named as NIL-Ghd8 ZS97 ) were developed previously (Yan et al. 2011). An F 1 hybrid was then generated by crossing each transgenic line with the corresponding negative control line or ZS97.

DNA Extract and Genotype of Hybrids
Genomic DNA was extracted from young leaves using the CTAB method (Murray and Thompson, 1980) with minor modifications. Genotyping of the 146 CSSLs by using a RICE 6 K chip generated a total of 5,102 highquality single nucleotide polymorphisms (SNPs), which were evenly distributed on all 12 chromosomes (Sun et al. 2015). A genetic bin map with 518 bins was constructed based on the recombination breakpoints in the CSSLs. The genotypes of the BC and TC were deduced from each corresponding CSSL. The insertion/deletion (Indel) marker PID2 (F: TAG AGA TGA AAT GGA GGT G; R: GTC TCA TGT TCT TCA ACA TG) was used to identify the genotypes of all the allelic combinations, except that NIP/ACC10 hybrid was determined by Indel marker PID3 (F: CTT ATC TAT CAA GGT GCT C; R: TGC ACA CAT GTA ATG CAA AC), and MH63/ZS97 hybrid was identified by simple sequence repeat (SSR) marker RM5556 (F: GTA AGC CAT TTG CAC GGA CAAGG; R: GAG CTC AGG ATC ATC CCT ACA TGC ). PID2 was also used to identify the genotypes of complementary transgenic hybrids. Polymerase chain reaction was performed following the procedure of Panaud et al (1996). The Indel and SSR markers were separated by 4% polyacrylamide gel electrophoresis and visualized by silver staining.

QTL Analyses
To decrease multicollinearity among the bin markers, the linear ridge regression method was used for the QTL analysis with the bin-map in the CSSLs as described previously (Sun et al. 2015) and for the QTL analysis of MPH and OSH values. A significance level of P < 0.05 was set as the threshold in the three mapping populations to declare the presence of a putative QTL in a given bin. If several adjacent bins showed P values lower than the threshold, the QTL was tentatively located in the bin (peak bin) with the lowest P value (Sun et al. 2015). The phenotypic variance explained by each QTL was decomposed using "relaimpo" package of R ("lmg" function). QTL nomenclature followed the principles suggested in a previous report by McCouch (2008).

Phenotypic Performance of CSSLs, BC, and TC Populations
Three mapping populations (CSSLs, BC, and TC) exhibited wide phenotypic variances with continuous distribution for 12 yield-related traits ( Fig. 1; Additional file 1: Table S1), showing a quantitative trait inheritance pattern. Most of the lines in the CSSLs and BC populations had similar phenotypic performance as ZS97, except for several lines that showed significantly higher or lower values than ZS97 ( Fig. 1; Additional file 1: Table S1), indicating that these lines carry either the introduced homozygous NIP or heterozygous NIP segments associated with the measured traits. Moreover, wide variation and continuous distribution of MPH and OSH values were observed ( Fig. 2; Additional file 1: Table S1). Several hybrids (of BC and TC) also exhibited heterosis values in two directions significantly higher or lower than corresponding controls for yield-related traits (Fig. 2). The TC population within the complex heterozygous background revealed similar variation in heterosis values to BC population within the background of ZS97 ( Fig. 2; Additional file 1: Table S1).
Correlation analysis was performed for these 12 traits among three populations (Additional file 2: Figure S1). Significantly positive correlations were observed in all of the pairwise traits except for seed setting ratio between

Detection of QTLs in CSSLs
The linear ridge regression method was used for the QTL mapping in CSSLs to decrease multicollinearity among markers. A total of 341 QTLs for 12 yield-related traits were identified in the CSSLs across two environments (E1 and E2) (Fig. 3a), and they were distributed on all 12 chromosomes (Additional file 1:  Table S2).

Detection of HL MP for Yield Traits
The CSSL and BC populations were used to identify HL MP in the homozygous background of ZS97. Every line in BC population contains only one or a few introduced heterozygous NIP segments (NIP/ZS97) within the homozygous background of ZS97. Therefore, the HL MP mainly present a non-additive genetic effect between NIP and ZS97 alleles at a given single locus. A total of 307 HL MP were detected for 12 traits across two environments (Fig. 3a). Among them, 42 QTLs were detected in both two environments. Most of HL MP (57.0%) increased MPH for yield and yield-related traits (Additional file 1:  Table S3).
In addition, all three genetic components (additive, dominance, and overdominance effects) at HL MP were estimated (Fig. 3b). The majority (97%) of HL MP exhibited an overdominance or dominance effect ( Fig. 3b; Additional file 1: Table S3). These data indicate that overdominance and dominance effects play a crucial role in MPH.

Detection of HL OS for Yield Traits
Each line in the TC population contains one or a few introduced heterozygous (NIP/MH63) segments in the otherwise uniform heterozygous background of a widely used hybrid cultivar SY63. Therefore, each HL OS effect represents an interaction effect between NIP and MH63 alleles at a given locus. HL OS detected in TC are summarized in Table S4 (Additional file 1: Table S4). A total 310 HL OS were identified for the 12 traits across two environments, which were distributed on all 12 chromosomes. Fifty-seven loci were detected in both two environments. Among them, 55.4% of the loci showed that NIP/MH63 hetero-allelic interaction increased over-standard heterosis. Twenty-eight HL OS for HD were detected across two environments. Twenty-five HL OS affecting PH was identified; among these, qTPH1.  For each bar diagram, the x-axis represents the physical location along each numbered chromosome. The y-axis represents the P value for single nucleotide polymorphism (SNP) association. Dashed lines indicate the declaration thresholds. Some major heterosis-related genes reported in previous studies are indicated by red dots in c, d values. Twenty-five loci were detected for SS, with 16 of them showing that the heterozygous alleles decreased the phenotypic values. For TGW, a major locus qTTGW5.3 explained 8.5% of the phenotypic variance in E1. For YD, 24 HL OS were identified, explaining 45.1% and 38.9% of the phenotypic variance in E1 and E2, respectively. qTYD1.2 had the most significant effect (Additional file 1: Table S4), explaining 16.2% of the phenotypic variance.

Positive Effect of Ghd8 on Yield Heterosis
Among the HLs, a major loci on chromosome 8 (3.80-4.37 Mb) was identified for MPH and OSH of four traits (Fig. 3c, d; Additional file 1: Table S3-4), and it was located in the same region of the QTL for six yieldrelated traits detected in the CSSLs (Additional file 1: Table S2), which contains a known functional gene (Ghd8 NIP ) (Yan et al. 2011). To validate the effect of this HL, NIL-Ghd8 NIP that carries an introduced NIP segment encompassing Ghd8 (Fig. 4a) was selected and crossed with NIL-Ghd8 ZS97 to produce F 1 hybrids. The MPH effects at Ghd8 were assayed in 11 yield-related traits. The heterozygotes at Ghd8 showed significant MPH for GN, PW, SN, and YD over the two parental lines (NIL-Ghd8 NIP and NIL-Ghd8 ZS97 ) across two-year trials (Fig. 4b). As GN and PW are highly dependent on SS, but SS is easily affected by high temperature in summer during the experiments, a representative yield component, SN, which is highly correlated with GN and PW, is used as the example to assess the heterotic effect. The heterozygous Ghd8 showed high MPH for YD (9.7%) and SN (6.9%) across the two-year trials (Fig. 4b). Moreover, an overdominance effect (|d/a|= 2.94) of Ghd8 on YD and a complete dominance effect (|d/a|= 0.99) on SN were found in the NIL-Ghd8 ZS97 /NIL-Ghd8 NIP hybrid (Fig. 4c). These results confirm that Ghd8 is a heterotic locus with an overdominance or dominance effect increasing yield and yield traits.

Allelic Interaction of Ghd8 Associated with Heterosis
In a previous study, three alleles (Ghd8 NIP , Ghd8 9311 and Ghd8 ACC10 ) were reported to be functional, while Ghd8 ZS97 and Ghd8 MH63 being loss-of-function alleles due to a premature stop codon occurred (Fig. 5a) (Wang et al. 2019b). To investigate the interaction effect of Ghd8 on yield MPH, five NILs (NIL-Ghd8 ZS97 , NIL-Ghd8 NIP , NIL-Ghd8 9311 , NIL-Ghd8 ACC10 , and NIL-Ghd8 MH63 ) were developed, with each carrying an introduced segment covering Ghd8 from a particular donor in the same background of ZS97 (Fig. 5b), and a half-diallel mating design with five NILs that contain a particular allele (Ghd8 ZS97 , Ghd8 NIP , Ghd8 9311 , Ghd8 ACC10 , and Ghd8 MH63 ) was used to generate 10 hetero-allelic combinations. The combination (Ghd8 ACC10 /Ghd8 MH63 ) revealed the highest MPH for YD (29.0%) and SN (20.4%) (Fig. 5c, d; Additional file 1: Table S6). Another five combinations (Ghd8 ZS97 /Ghd8 ACC10 , Ghd8 ZS97 /Ghd8 NIP Ghd8 9311 /Ghd8 ACC10 , Ghd8 9311 /Ghd8 MH63 , Ghd8 NIP /Ghd8 ACC10 ) showed significantly higher YD heterosis over the check Ghd8 ZS97 /Ghd8 MH63 (Fig. 5d). However, three allelic combinations (Ghd8 NIP /Ghd8 9311 , and Ghd8 ZS97 /Ghd8 9311 , Ghd8 NIP /Ghd8 MH63 ) produced low or no YD and/or SN advantage over the check Ghd8 ZS97 /Ghd8 MH63 (Fig. 5c, d; Additional file 1: Table S6). These results indicate that the interaction of various Ghd8 alleles could cause different heterosis levels of yield. Fig. 4 Validation of the Ghd8 effect on mid-parent heterosis (MPH) using near-isogenic lines (NILs) and complementation transgenic lines. a Graphical genotype of NIL-Ghd8 NIP showing a single introduced Nipponbare segment encompassing Ghd8 in the ZS97 background. b The heterosis advantage of Ghd8 for the yield traits in hybrids of NIL-Ghd8 NIP and NIL-Ghd8 ZS97 in a two-year trial. Error bar represents the mean ± SE (n = 3). c The dominance degree of Ghd8 effect on spikelet number (SN) and grain yield per plant (YD) in hybrids. If the effect has an excessive degree of dominance, it is set to a score of 5.0. d Phenotypes of the complementation transgenic lines and their hybrids at maturity. Ghd8 + , homozygous-positive transgenic line containing the alleles Ghd8 NIP ; Ghd8 − , homozygous-negative control lines; F 1 denotes the hybrid derived from the cross of the independent complementation transgenic plant and corresponding negative plant. Scale bars, 10 cm. e, f MPH of Ghd8 for SN (E) and YD (F) in the F 1 hybrids across a two-year trial. The error bar represents the mean ± SE (n = 3)

Discussion
In the present study, we identified 98 QTLs for yield traits and HL MP for MPH in the same or overlapping regions in both CSSLs and BC population with the homozygous background of ZS97 (Fig. 3a). These commonly detected loci are in the accordance with the high correlations among the yield-related traits (Additional file 2: Fig. S1).
As comparison, a larger number of loci were found only in the CSSLs not in the BC or TC ( Fig. 3a; Additional file 1: Table S2-S4), suggesting that most HLs were caused by allelic interaction from heterozygotes. These results also suggest that independent genetic effects (a, d) at that loci are reflected in the homozygous NIP in CSSLs and heterozygous NIP alleles in BC. Seventy-five heterotic loci (HL OS ) were found co-localized in the same or overlapping regions of HL MP , indicating that both heterozygous NIP/ZS97 and NIP/MH63 alleles at the detected loci significantly affect heterosis in rice, although the interaction effects were influenced by different genetic backgrounds. Of these heterotic loci for MPH and OSH, 32 also revealed major effects on the trait performances in the CSSLs (Fig. 3a). Under comparison of the QTLs detected in the CSSLs with those in previous reports, at least 81 loci co-localized in the same or overlapping regions harboring the genes associated with yield-related traits (Additional file 1: Table S2). For example, qPH1.4 for PH detected in the CSSLs was located near sd1, a gibberellin synthesis gene (OsGA20ox2) regulating plant height (Sasaki et al. 2002). Five loci (qGN7.2/qHD7.1/qPH7.2/qSN7.2/ qYD7.1) were localized in the region that contains the known yield gene Ghd7 (Xue et al. 2008). Five QTLs (qGN7.6/qHD7.4/qPH7.5/qSN7.6/qYD7.5) were mapped in the same region where Ghd7.1 was reported to affect grain number, plant height, and heading date (Yan et al. 2013). Six loci (qGN8/qPB8/qPH8.1/qPW8.1/qSB8/ qSN8.1) were detected in the region encompassing Ghd8, a yield-related gene with a pleiotropic effect on grain number, plant height, and heading date (Yan et al. 2011).
Notably, many studies reported that Ghd8 could be an important candidate gene that affects heterosis for yieldrelated traits Huang et al. 2015Huang et al. , 2016Chen et al. 2019;Lin et al. 2020). However, it lacked transgenic validation. In the present study, we identified four HL MP (qBHD8.1/qBPL8.2/qBSB8.1/qBSN8.1) and one HL OS (qTHD8.1) (Additional file 1: Table S3-S4) that were commonly detected in the Ghd8 region and validated them as the major heterotic locus for yield and spikelet number using transgenic experiments (Fig. 4). Moreover, the interaction effects arose from some hetero-allelic combinations of Ghd8 caused different levels of YD and SN heterosis. Three functional alleles (Ghd8 NIP , Ghd8 9311 , and Ghd8 ACC10 ), when interacting with non-functional allele Ghd8 ZS97 , significantly increased heterosis for YD and SN compared with the combination of two non-functional alleles (Ghd8 ZS97 and Ghd8 MH63 ) (Fig. 5). The heteroallelic combinations, such as Ghd8 ACC10 /Ghd8 MH63 and Ghd8 9311 /Ghd8 MH63 , also exhibited much significantly higher MPH for YD and SN than the check combination Ghd8 ZS97 /Ghd8 MH63 . These data suggest that the allelic interaction effect arose from the functional and nonfunctional alleles at Ghd8 could produce more spikelet (See figure on next page.) Fig. 5 Yield mid-parent heterosis (MPH) of 10 allelic combinations at Ghd8 in heterozygotes within the background of ZS97. a Schematic gene mode showing nucleotide variations in the coding region of Ghd8 among five parents (NIP, ZS97, 9311, ACC10 and MH63). NIP (Nipponbare) is used as a reference. Polymorphic nucleotides are indicated by different color lines. "F" and "N" denote functional and non-functional alleles, respectively. b Plant image of five NILs (NIL-Ghd8 ZS97 , NIL-Ghd8 NIP , NIL-Ghd8 9311 , NIL-Ghd8 ACC10 , and NIL-Ghd8 MH63 ). The images were taken at the maturity of NIL-Ghd8 ZS97 ; scale bar, 10 cm. c, d MPH of Ghd8 for spikelet number (SN) and yield per plant (YD) in ten hybrids with different allelic combinations. The error bar represents the mean ± SE (n = 3). NIL-Ghd8 NIP , NIL-Ghd8 9311 , NIL-Ghd8 ACC10 , and NIL-Ghd8 MH63 represent those NILs carrying corresponding Ghd8 alleles from different donors within the same ZS97 background, respectively. The different letters denote significant differences by LSD test at P < 0.05. The error bar represents the mean ± SE (n = 3) number and grain yield in heterozygotes than in the corresponding homozygotes or the check combination. However, the molecular mechanisms for the varied heterosis levels led by allelic combinations require further investigation. It has been reported that various allelic interactions may lead to novel hybrid expression patterns (He et al. 2010;Groszmann et al. 2015;Shao et al. 2019), protein metabolism (Goff, 2011;Chen, 2013), and epigenetic changes such as small RNAs and histone modification (Springer and Stupar, 2007;Lauss et al. 2019). In this regard, the transcriptional or post-transcriptional regulations, or polymer/dimer products from the allelic interaction may be causes of the heterosis variation. The developed NILs each contains a particular allele at the heterotic locus will provide an excellent stock to dissect the underlying mechanisms.
In addition, the elite rice hybrid SY63 has been successfully used for commercial hybrid production with the largest cultivated area in China during the past three decades (Xie and Zhang, 2018). In the present study, we found that the original hetero-allelic combination Ghd8 ZS97 /Ghd8 MH63 in SY63 did not contribute to yield heterosis. However, the hetero-allelic combinations between Ghd8 ZS97 (or Ghd8 MH63 ) and any other alleles like Ghd8 NIP , Ghd8 9311 , or Ghd8 ACC10 in heterozygotes could produce a significant positive yield heterosis compared with the combination between Ghd8 ZS97 and Ghd8 MH63 . Thus, the replacement of either allele Ghd8 ZS97 or Ghd8 MH63 of SY63 with those identified desirable Ghd8 alleles with a marker-assisted selection approach can be used to improve the yield potential of hybrid cultivars.

Conclusion
A large number of HLs for yield-related traits were identified using three rice CSSL interconnected populations. Of these loci, Ghd8 was validated as a major HL for spikelet number and grain yield by transgenic experiments. Moreover, the investigation of 10 hetero-allelic combinations at Ghd8 exhibited several desirable allelic interactions in heterozygotes that can enhance yield heterosis. These data provide new insights into understanding the genetic basis of heterosis and will be exploited for increasing yield potential in hybrid rice breeding programs to meet the demand of growing population.