Introduction

Rice is a major cereal crop, and hybrid rice has significantly contributed to yield improvement in China. Genes responsible for heterosis of yield have been identified in the widely grown, two-line hybrid variety Liangyoupei9 (LYP9, cross PA64s/9311) (Li et al. 2016).

A group of heterosis-related genes was identified by correlating the concurrence of differentially expressed genes and yield-related QTL (quantitative trait loci) (Wei et al. 2009). The yield-related QTL qSN8 was confirmed as DTH8 (days to heading 8) by a complementation test (Gao et al. 2013). This major QTL for yield heterosis was also found to have pleiomorphic effects and was designated as DTH8/Ghd8/LHD1 as a result of integrating genetics and omics analyses (Li et al. 2016). By positional cloning, pms1 (photoperiod-sensitive genic male sterility) was isolated, and it was found to encode phasiRNAs that are involved in rice development (Fan et al. 2016). However, it is unclear if the heterozygous alleles of these genes affect heterosis with dominance effects.

A total of 66 chromosome segment substitution lines (CSSLs) and their corresponding F1 hybrids to the recurrent parent were genotyped by 137 SSR markers (Wang et al. 2012) to identify three major stable QTLs for heterosis in six environments. The QTL qFCC7L controlling the chlorophyll and photosynthetic rate was found to underlie low nitrogen stress tolerance in rice. This was based on 132 recombinant inbred lines (RILs) and fine mapping of the BC4F2 population derived from a cross between PA64s and 9311 (Ye et al. 2017). Forty-six QTLs associated with yield were detected using a set of 156 chromosome segment substitution lines (CSSLs) with overlapping segments from PA64s in a genetic background of 9311 (Liu et al. 2016). Neither population size nor the number of molecular markers in that study was sufficient for gene isolation and the analysis of gene effects.

To understand the mechanisms underlying heterosis in rice, we used 156 CSSLs previously developed by Liu et al. (2016). The CSSLs represented a population of backcross-derived lines of high-yielding variety 9311 with the elite variety Pei’ai64s (PA64s) as the donor. These varieties were also the parents of hybrid variety Liangyoupei9 (LYP9, cross PA64s/9311). We identified 10,768 SNPs in the CSSL population through genotyping-by-sequencing (GBS), and we developed two sets of hybrid populations to study heterotic effects. The B-homo hybrid set involved crosses of the CSSLs to the recurrent parent and had predominantly heterozygous substituted segments and a homozygous 9311 background. The B-heter hybrid set, produced using crosses to the donor parent, was homozygous for the substituted fragments and heterozygous for the genetic background. Heterosis of eight yield-related traits was evaluated over 2 years by comparing F1 sets with recurrent parent or Liangyoupei9 (LYP9).

We report here that qSS7 and qHD8 play key roles in the rate of seed setting and the heading date in a heterozygous background, respectively. Distinguished from other heterosis QTLs, qSS7 and qHD8 have pleiotropic effects that confer stable and strong heterosis in hybrid rice.

Materials and methods

CSSL and derived F1 populations

We developed 156 chromosome segment substitution lines (CSSLs) from a cross between high-yield cultivar 9311 as the recurrent parent and an elite cultivar Pei’ai64s (PA64s) as the donor. To understand the genetic basis of yield-related traits of LYP9, we developed two F1 populations: a B-homo F1 set derived from CSSLs/9311 with a heterozygous segment in a homozygous background and a B-heter F1 set derived from PA64s/CSSLs with a homozygous segment in a heterozygous background, respectively. Two F1 sets corresponding to CSSLs, background homozygous with segment heterozygous, abbreviated as B-homo F1, were produced from CSSLs/9311, and background heterozygous with segment homozygous, abbreviated as B-heter F1, were produced from PA64s/CSSLs.

The two parents, CSSLs, B-homo F1 (homozygous background with heterozygous segment) and B-heter F1 (heterozygous background with homozygous segment), were grown at Nanjing in 2016 and 2017, and the trials were designated as Nanjing environments E1 and E2, respectively. Each entry was a plot of two rows, containing 10 individual plants in a randomized block design with two replications, but with each parental CSSL and its corresponding F1 hybrid being planted side by side. Individual plant and row spacing within plots was 16.5 cm, and plot spacing was 23.5 cm. Fertilizer levels and disease control followed local recommendations.

Plant phenotyping

Heading date (HD) were recorded, and six plants per line were harvested from each plot at maturity. Plant height (PH), effective panicle number (EPN), grain number per panicle (GN), rate of seed setting (SS), 1000-grain weight (TGW), and panicle length (PL) were determined on three plants from each plot. Yield per plant (YPP) was measured and calculated as the average weight per plant of bulked grain obtained from the six plants. The measuring procedures were described in Xiao et al. (1998).

SNP genotyping-by-sequencing

Genotyping-by-sequencing (GBS) (Poland et al. 2012) of each entry was used for SNP calling. DNA for the GBS library was extracted from 14-d-old seedlings. PstI and MspI (NEB) were used for digestion and T4 ligase (NEB) for ligation. All of the samples were pooled for purification and PCR-amplification. The library was sequenced using the Illumina Platform, and the data were analyzed by Tassel software (Glaubitz et al. 2014).

Data analysis and QTL mapping

Estimates of marker distances, chromosome lengths, substituted segments, and overall genome size were based on the linkage map (Moncada et al. 2001). Construction of graphical genotypes and calculation of percentage of donor genome in each CSSL were performed using Tassel software. If two neighboring loci had alleles from the donor parent, then the interval between them was considered to be the length of the segment. If one locus had an allele from the recurrent parent and the other locus had the allele from the donor parent, then half of the interval between them was considered to represent the length of the substitution.

QTL in CSSLs and their F1 hybrid sets were detected by QTL IciMapping software (Li et al. 2007; Wang et al. 2007) based on the combined analysis of their marker genotypic and phenotypic data. Based on the map (Fig. 1) and three sets of phenotypic data (a set of CSS lines and corresponding hybrid sets in two environments), the effects of chromosome segments from the donor parent in CSSL and heterozygous genotypes in F1 sets were analyzed. The LOD threshold, a measure of significance, was set at 5.0 (Wang et al. 2006). QTL nomenclature followed McCouch et al. (1997). The estimated additive and dominance effects were used to calculate |d/a| and thereby classify the QTL as additive (A) (|d/a| < 0.2), partially dominant (PD) (0.2 ≤ |d/a| < 0.8), completely dominant (CD) (0.8 ≤ |d/a| < 1.2), or overdominant (OD) (|d/a| ≥ 1.2) (Stuber et al. 1992). Heterosis in the B-homo set was expressed as the mid-parent heterosis value: B-homo F1 - (CSSL mean + 9311)/2.

Fig. 1
figure 1

SNP map of the CSSL populations. SNPs were detected by genotyping-by-sequencing and analyzed using Tassel software. The green background of the SNP map is from 9311, the red substitution segments are from PA64s, and the gray segments are heterozygous or undetected

Full size image

Results

Yield-related traits in CSSLs and F1 hybrids

We constructed two F1 sets, B-homo F1 and B-heter F1, based on a set of 156 CSSLs (Fig. 1). Heterosis of the eight yield-related traits was evaluated by comparing the F1 lines with two parents or Liangyoupei9 (LYP9) in two environments (Fig. S1).

The data for parental cultivar 9311, CSSLs, and corresponding F1 sets are given in Table S1. PL, PH, EPN, SS, TGW, and YPP showed positive heterosis (measured as MPH, mid-parent heterosis) in the B-homo F1 set across both environments, but only EPN showed positive heterosis in the B-heter F1 set. HD, PL, SS, and YPP showed negative heterosis (measured as ES, effects of substituted segment) in the B-heter F1 set in both environments. Correlation analysis of two sets of F1 traits and CSSL traits, based on the mean values of eight yield-related traits of F1 hybrid sets across environments, showed that yield per plant (YPP) of the B-homo set was not significantly related to the eight traits of CSSLs. This indicated that the mid-parent heterosis of yield per plant (YPP) mainly came from the heterozygous effect of QTL for yield traits in a homozygous background. However, YPP in the B-heter set was positively related to the HD and YPP of the CSSLs and significantly related to PH, SS, and TGW of CSSLs. This indicated that the yield level of the hybrid was due to higher YPP, PH, SS, and TGW and the prolonged growth period relative to the parental lines (Table S2). Combined ANOVA of yield across environments showed that the yield component traits were influenced by significant genotypic and environmental effects, as well as genotype–environment interactions (Table S3).

SNP genotyping-by-sequencing and QTL mapping

We identified 10,768 SNPs in CSSLs subjected to the GBS, which we also applied in our previous studies with minor modifications (Tang et al. 2016, 2019; Yu et al. 2020). The SNPs were evenly distributed across the 12 chromosomes (Fig. S2). We detected QTLs underlying the yield of hybrids by association of SNPs with trait phenotypes in QTL analyses. QTLs were detected using QTL IciMapping software (Meng et al. 2015), which is available on (http://www.isbreeding.net/software/). We used the ICIM (Inclusive Composite Interval Mapping) method for QTL identification (Wang 2009). We chose the LOD 5.0 as the threshold for QTL analysis. The QTLs were listed when they were detected in both environments or two F1 sets (Table 1). The identified QTLs were named according to McCouch et al. (1997).

Table 1 Five QTLs were identified in the B-homo and the B-heter F1 sets
Full size table

QTL analysis for yield-related traits in LYP9

In a homozygous background, 14 QTLs with dominance effects were identified at the threshold of LOD 5.0 (Table S4). Five showed overdominance, six were completely dominant, and three were partially dominant. No QTL had an additive effect in a predominantly homozygous background, indicating that dominance effects play an important role in heterosis.

A total of 28 QTLs were detected in a heterozygous background, indicating that dominance affects contributed to heterosis (Table S5). QTLs qHD8 and qSS7 were identified as pleiotropic traits in both F1 sets (Table 1). The qHD8 overlapped the position of gene DTH8 on chromosome 8, whereas the qSS7 association with SS, PH, and YPP was novel.

Heading date (HD) is an important agronomic trait in hybrid rice, and it affects the desired yield level in rice breeding. QTL qHD8, controlling heading date, was a highly significant locus explaining about 60% of the phenotypic variation. PH is a target trait for hybrid rice improvement. In this study, qPH5, qPH6, qPH7, and qPH9 and qPH6 explained 14.10 and 19.00% of the phenotypic variation in each environment. SS is also an important trait in heterosis, and qSS3.1, qSS3.2, qSS4, qSS6, qSS7.1, qSS7.2, qSS9, and qSS10 were detected in both environments, explaining 5.98–6.55%, and 7.41–8.73% of the phenotypic variation. For GN, YPP, EPN, and TGW, we detected qGN8, qYPP3, qYPP7 (qSS7), qYPP9, qEPN4, and qTGW8 (qHD8). Most of the QTLs controlled a single trait, but qSS7.1 and qHD8 were exceptions in having pleiotropic effects on PH, SS, and YPP, and HD, PH, GN, and TGW, respectively.

Heterosis-related loci and candidate genes

We focused on a detailed study of QTLs qSS7 and qHD8 (Fig. 2, Fig. S3). qHD8 was previously reported to affect heterosis based on SN (spikelet number per panicle) (Gao et al. 2013; Li et al. 2016), heading date, and plant height (Li et al. 2016).

Fig. 2
figure 2

QTLs detected on 12 chromosomes. A scale at the left side represents physical position (Mb). QTLs detected in the environment E1 were marked in orange, and QTLs in E2 were marked in green. Among these QTLs, qSS7 and qHD8 were identified repeatedly in the two hybrid sets in a homozygous background (a) and heterozygous background (b), respectively

Full size image

We analyzed the dominance effects of the two QTLs in heterozygous backgrounds, and both showed positive heterosis potential of heterozygous segments (Fig. 3). The dominance effects of qSS7 on SS, YPP, and PH were 19.0, 9.4, and − 5.1, respectively. Therefore, heterozygosity of qSS7 conferred heterosis by prolonging the growth period and reducing plant height (Fig. 3a). Similarly, the dominance effects of qHD8 on HD, PH, GN, and TGW were 8.3, 1.5, 34.2, and 2.2, respectively (Fig. 3b).

Fig. 3
figure 3

Dominance effects of the two QTLs qSS7 and qHD8 in a heterozygous background. a QTL qSS7 shows positive values in terms of SS-2, YPP-1 and YPP-2 traits, and negative values in PH-2. b QTL qHD8 shows positive values in terms of HD-1, HD-2, PH-1, GN-1 and TGW-2 traits

Full size image

Heterosis utilization potentials

We found that qHD89311 increased TGW (Fig. 4b). The homozygous fragment of qHD8PA64s in a heterozygous background conferred lower TGW as shown in lines L54, L53, L55, and L90 (Fig. 4a, b). To determine why these lines had decreased TGW, we measured the seed length, width, length width ratio, and thickness, and we found that PA64s/L90 (homozygous qHD8PA64s) was not significantly different from LYP9 (Fig. 4c–f). However, a higher chalkiness rate was found in PA64s/L90 compared to LYP9, and it probably decreased TGW (Fig. 4g, h).

Fig. 4
figure 4

Comparisons of F1 lines derived from PA64s/L54, PA64s/L53, PA64s/L55, PA64s/L90, and LYP9 lines. L54, L53, L55, and L90 were four CSSLs. a The physical map of the 9311 variety and the CSSLs harboring qHD8PA64s. b Comparison of 1000-grain weight of the four F1 lines and LYP9. c–g The seed length, width, length, width ratio, and level of chalkiness in LYP9 and PA64s/L90. h Image of the seeds of LYP9 and PA64s/L90

Full size image

The QTL qSS7 had effects on PH, SS, and YPP with LOD values ranging from 7.83 to 42.51 and PVE from 7.90 to 49.65%. A phasiRNA was reported to control the rate of seed setting under long-day conditions (Fan et al. 2016). The SS of F1 with qSS7PA64s/9311 in a heterozygous background was not significantly different from 9311 or the male parent (Fig. 5a). However, the SS of F1 with qSS7PA64s/PA64s in a heterozygous background was significantly lower than the male parent (Fig. 5b), whereas the SS of F1 with qSS7PA64s/PA64s in homozygous background was not significantly different from the parents (Fig. 5c). Interestingly, the QTL was also detected to underlie PH in the B-heter F1 set, although phasiRNAs have not been previously reported to control PH.

Fig. 5
figure 5

Rates of seed setting of the lines with qSS7PA64s/9311 and qSS7PA64s/PA64s in different backgrounds. L19, L112, L45, L57, and L155 were the five CSSLs from the QTL mapping population. qSS7PA64s/9311 represents the heterozygous segment of the qSS7, and qSS7PA64s/PA64s represents the homozygous segment of the qSS7. a Rates of seed setting of F1 harboring qSS7PA64s/9311 in a heterozygous background and parents. Left: comparison of PA64s, PA64s/9311 and 9311; middle: comparison of PA64s, PA64s/L19 and L19; right: comparison of PA64s, PA64s/L112 and L112. b Rates of seed setting of F1 harboring qSS7PA64s/PA64s in a heterozygous background and parents. Left: comparison of PA64s, PA64s/L45 and L45; middle: comparison of PA64s, PA64s/L57 and L57; right: comparison of PA64s, PA64s/L155 and L155. c Rates of seed setting of F1 harboring qSS7PA64s/9311 in a homozygous background and parents. Left: comparison of L45, L45/9311 and 9311; middle: comparison of L57, L57/9311 and 9311; right: comparison of L155, L155/9311 and 9311

Full size image

Heterozygous segments lead to superior performance in B-heter set

The GBS analyzed SNPs that identified qSS7 and qHD8 in the CSSLs. An A/G SNP is located near qSS7, whereas adjacent SNPs CC/TA were close to qHD8. Using these SNPs as markers, we classified the CSSLs into four groups, qSS79311/qHD89311 (A/CC), qSS79311/qHD8PA64s (A/TA), qSS7PA64s/qHD89311 (G/CC), and qSS7PA64s/qHD8PA64s (G/TA), and we analyzed the phenotypic data of two loci in both F1 sets. The heterozygous qHD8 segment in a homozygous background significantly increased SS, but the homozygous qHD8 did not (Fig. 6a). Heterozygous qSS7 in a heterozygous background (equal to LYP9) significantly increased GN-1, SS-2, and YPP-2, but the homozygous qSS7 did not (Fig. 6b). Similarly, the heterozygous qHD8 in a heterozygous background increased HD-1, HD-2, PH-1, PH-2, GN-1, YPP-1, and YPP-2 of hybrid set, but the homozygous qHD8 did not (Fig. 6c). These results indicated that heterozygous segments of the two QTLs conferred superior performances, especially in B-heter set.

Fig. 6
figure 6

Comparisons of phenotypic variances for the two QTLs in the F1 sets. a Comparison of the SS-1 trait of the B-homo F1 sets with the homozygous and heterozygous qHD8 segments, respectively. b Comparisons of the GN-1, SS-2 and YPP-2 traits of the B-heter F1 sets with the homozygous and heterozygous qSS7 segments, respectively. c Comparisons of the HD-1, HD-2, PH-1, PH-2, GN-1, YPP-1 and YPP-2 traits of the B-heter F1 sets with the homozygous and heterozygous qHD8 segments, respectively. No significant differences were observed when comparing traits of the B-homo F1 sets with the homozygous and heterozygous qSS7 segments. PH plant height, SS rate of seed setting, GN grain number per panicle, YPP yield per plant, HD heading date. *, **P < 0.05 and 0.01, respectively

Full size image

Interaction between qSS7 and qHD8

Analysis of variance (Table S6) showed significant interactions between dominance effects of the two QTLs. There were significant differences among all of the combinations of the two QTLs showing that dominance–dominance interactions contributed to the superior performance of qSS7PA64s/9311/qHD8PA64s/9311 heterozygotes in the B-heter F1 set that were labelled as DD with the highest performances in the eleven traits (Fig. 7).

Fig. 7
figure 7

Multiple comparisons of interactions between the two QTLs qSS7 and qHD8 in the B-heter F1 sets. DD represents dominance–dominance interaction between qSS7PA64s/9311/qHD8PA64s/9311; DA represents dominance–additive interaction between qSS7PA64s/9311/qHD8PA64s/PA64s; AD represents additive–dominance interaction between qSS7PA64s/PA64s/qHD8PA64s/9311. HD heading date, PH plant height, PL panicle length, GN grain number per panicle, SS rate of seed setting, TGW 1000-grain weight, YPP yield per plant. Different uppercase letters indicate significance at P < 0.01

Full size image

Discussion

Heterosis is defined as the superior performance of a hybrid over the parental mean. Understanding the genetics of heterosis should allow production of hybrid varieties by molecular design breeding. To analyze the mechanism of heterosis at the genetic level, four conditions are required: (1) genetic materials with strong heterosis as demonstrated by field trials; (2) specific agronomic traits defined; (3) a suitable experimental population to analyze the genetic components of heterosis, including dominance, overdominance, and unbiased estimation of epistasis; and (4) a high-density genetic linkage map for QTL analysis. In this study, we identified heterosis-associated loci using 156 CSSLs derived from the parents of a super hybrid rice variety crossed to their corresponding parents. This enabled the study of specific introgressions, in homozygous and heterozygous conditions, against the genetic backgrounds that were, in turn, heterozygous or homozygous. Phenotypic variation from a field trial repeated in 2 years was then regressed on a genetic map consisting of 10,768 SNP markers. Two major QTLs were chosen for detailed study. These were qSS7, conferring a high rate of seed setting and low plant height, and qHD8, with pleiotropic effects on HD, PH, GN, and TGW.

Heterosis includes additive effects, dominance effects, and epistatic effects from the interaction effects of non-alleles between loci (Hua et al. 2003). The study of epistatic effects is challenging for the following reasons. To detect the potential interaction effects of n loci, there are n(n-1)/2 combinations of loci of which there are four types of interaction modes. These include additive–additive interaction, additive–dominance interaction, dominance–additive interaction, and dominance–dominance interaction. Interaction effects are also susceptible to genetic background interference and environmental effects. Therefore, the study of epistatic effects requires specific genetic cross designs and an appropriate genetic background of test materials. Although the hybrid F1 of the CSSL-based NCII design combined with BCF1 or TCF1 populations can eliminate genetic background interference and facilitate statistical analysis, the workload is substantial. The materials of B-heter F1 set generated in this study enabled the evaluation of four types of interaction modes, i.e., AA (additive–additive interaction), AD (additive–dominance interaction), DA (dominance–additive interaction) and DD (dominance–dominance interaction). We focused on analyzing the effects of substituted segments, deletion of the dominance effects of the fragment, and the dominance–dominance interaction between the fragments. To achieve the genetic response for heterosis utilization and make full use of advantageous allelic mutations directly, the targeted fragments with disadvantageous heterotic loci were replaced.

We collected the phenotypic data of a set of CSSLs and corresponding F1 hybrid sets in two environments. Based on a SNP map and phenotypic data of eight traits, namely, PH, GN, SS, YPP, HD, EPN, TGW, and PL, five QTLs were identified. GBS, a economical alternative to other whole-genome genotyping platforms, allowed us to calculate the substituted fragment of parental inbred lines accurately. We directly evaluated the agronomic traits of the F1 population derived from 156 CSSLs in the field. The qHD8, with a 214.46 kb region, contains the previously reported gene RH8/DTH8 (Gao et al. 2013; Li et al. 2016).

The China Rice Data Center database (http://www.ricedata.cn/) placed pms1 at approximately 6.69 Mb on chromosome 7, ~ 2.19 Mb from qSS7 (4.27–4.50 Mb). The QTL qSS7 could be a unique gene rather than pms1. If this is true, its isolation could clarify regulation of the rate of seed setting, a key yield component in hybrid rice.

Male sterility of the parent PA64s is conferred by photoperiod and thermo-sensitive male sterile genes. LOC_Os07g12130 is the thermo-sensitive candidate gene for pms1 (Zhou et al. 2011a, b). The quantitative traits are influenced by multiple genes or QTLs (Li et al. 2003). Gene expression could be greatly affected by environmental factors (Cao et al. 2001). In this study, the distance between qSS7 and pms1 could be due to environmental effects on the mapping of qSS7. Although the pms1 gene was previously reported to control pollen male sterility, we found that qSS7 increased the rate of seed setting and reduced plant height.

The pms1 gene encodes a long-non-coding RNA PMS1T that was reported to be preferentially transcribed in young panicles. PMS1T was shown to be a long non-coding RNA (lncRNA), targeted by miR2118 to produce 21-nt phasiRNAs. Under long-day conditions, these phasiRNAs preferentially accumulate in photoperiod-sensitive male sterile rice genotypes, and the higher accumulation of phasiRNAs is the cause of male sterility (Fan et al. 2016). Non-coding RNAs have been found to function in a wide range of plant physiological mechanisms. For example, phasiRNAs preferentially accumulate in maize reproductive tissues (Zhai et al. 2015). It will be intriguing to determine whether pms1 underlying male sterility is the candidate gene in qSS7 that controls the rate of seed setting and plant height. As the heterozygous segments harboring the heterosis genes affect yield in hybrid rice, the CSSLs harboring these genes can be used as donor parental lines for heterosis utilization.