Abstract

   The yield heterosis of rice is sought by farmers and strong contributes
   to food safety, but the quality of hybrid rice may be reduced.
   Therefore, developing new varieties with both high yield and good
   quality is a heavily researched topic in hybrid rice breeding. However,
   the molecular mechanism governing yield heterosis and high rice quality
   has not been elucidated to date. In this study, a comparative
   transcriptomics and genomic analysis was performed on a hybrid rice
   variety, Chuanyou6203 (CY6203), and its parents to investigate the
   molecular mechanism and gene regulation network governing the formation
   of yield and quality stages. A total of 66,319 SNPs and InDels between
   CH3203 and C106B were detected in the 5′-UTR, exon, intronic, and
   3′-UTR regions according to the reference genome annotation, which
   involved 7473 genes. A total of 436, 70, 551, 993, and 1216 common DEGs
   between CY6203 and both of its parents were identified at the same
   stage in panicles and flag leaves. Of the common DEGs, the numbers of
   upregulated DEGs between CY6203 and CH3203 were all greater than those
   of upregulated DEGs between CY6203 and C106B in panicles and flag
   leaves at the booting, flowering, and middle filling stages.
   Approximately 40.61% of mRNA editing ratios were between 0.4 and 0.6,
   and 1.68% of mRNA editing events (editing ratio ≥ 0.8) in CY6203
   favored one of its parents at three stages or a particular stage,
   suggesting that the hypothetical heterosis mechanism of CY6203 might
   involve dominance or epistasis. Also 15,934 DEGs were classified into
   19 distinct modules that were classified into three groups by the
   weighted gene coexpression network analysis. Through transcriptome
   analysis of panicles and flag leaves in the yield and quality stages,
   the DEGs in the green-yellow module primarily contributed to the
   increase in the source of CY6203 due to an in increase in
   photosynthetic efficiency and nitrogen utilization efficiency, and a
   small number of DEGs related to the grain number added spikelet number
   per panicle amplified its sink. The balanced expression of the major
   high-quality alleles of C106B and CH3203 in CY6203 contributed to the
   outstanding quality of CY6203. Our transcriptome and genome analyses
   offer a new data set that may help to elucidate the molecular mechanism
   governing the yield heterosis and high quality of a hybrid rice
   variety.

   Subject terms: Genetics, Plant sciences

Introduction

   Rice is one of the world's major crops and is the staple food of more
   than half of the world's population. Successful application of rice
   heterosis significantly increased yield. In the past hundred years, the
   heterosis of crops with distinct yield advantages has been widely used
   in corn, sorghum, rape, and other crops. Researchers have frequently
   attempted to reveal the genetic mechanism of heterosis, and have
   proposed several classical genetic models, such as dominance,
   over-dominance, epistasis, and non-additive gene expression^[46]1.
   However, the molecular mechanism governing heterosis has not been
   elucidated.

   At the DNA and mRNA levels, genomic and transcriptome analysis
   represent a useful method to study the genetic mechanism of heterosis.
   Bao et al.^[47]2 identified 595 upregulated and 25 downregulated tags
   in LYP9 using the serial analysis of gene expression technique and
   found that most of the genes were related to enhancing carbon and
   nitrogen assimilation. Wei et al.^[48]3 detected 3926 differentially
   expressed genes (DEGs) between hybrid F[1], LYP9, and its parents
   (DEG[HP]) and between the parents (DEG[PP]) and found that the DEGs in
   the categories of energy metabolism and transport were enriched in
   DEG[HP], rather than in DEG[PP]. Huang et al.^[49]4 believed that
   numerous superior alleles, such as Hd3a, OsSPL14, Gn1a, Waxy, ALK, and
   qSW5, exhibiting positive dominance, contributed to heterosis. Chen et
   al.^[50]5 concluded that the overdominant effect probably contributed
   to the grain number heterosis of WFYT025. The genome-wide gene
   expression profiles were compared to identify the genetic basis of
   yield heterosis of such species maize^[51]6,[52]7, Bombyx mori^[53]8,
   rubber tree^[54]9, and oil palm^[55]10. These studies all showed that
   the DEGs and major genes contributed to heterosis.

   The yield heterosis of rice is sought by farmers, but the quality of
   hybrid rice is reduced. The poor eating and cooking quality of most
   hybrid rice, especially indica hybrid rice, is mainly attributable to a
   high amylose content (AC), hard gel consistency (GC), and high
   gelatinization temperature (GT)^[56]11–[57]13. Recently, we released a
   new hybrid rice variety, Chuanyou6203 (CY6203), which has the same
   eating and cooking qualities as the Thai rice variety. The quality of
   CY6203 conforms to the second grade of the State Standard of the
   People's Republic of China (Rice)(GB1354-2009). After the variety
   Chuanyou6203 (CY6203) was released, it was welcomed by farmers. And its
   planted area had quickly reached 200,000 hectares in rice area of
   southwest China. In this study, a transcriptome analysis of the
   panicles and flag leaves of the hybrid rice CY6203 and its parents in
   the formation of yield and quality stages was performed to identify the
   molecular modules related to yield heterosis and high quality. The data
   presented in this report may be help to elucidate the molecular
   mechanism governing yield heterosis and the formation of high quality.

Results

Performance and heterosis of Chuanyou6203

   Figure [58]1 shows the plant type, panicle type, and grain shape of
   hybrid rice variety CY6203 and its parents. We investigated a total of
   ten traits, including five agronomic traits (SPW/D, PH, PL, ANSPP, and
   TGW) and five quality traits (HRR, L/W, AC, ASV, and GC) (Table [59]1).
   The results showed that there was a high significant difference
   (p < 0.01) between C106B and CH3203, except for ASV and GC, and a
   significant difference (p < 0.05) in ASV. Between CY6203 and its
   parents, there was a high significant difference (p < 0.01) except for
   PL, TGW, and ASV. There was a high significant difference (p < 0.01)
   between CY6203 and C106B in TGW and ASV, respectively. In PL, a high
   significant difference (p < 0.01) was found between CY6203 and CH3203,
   and there was no significant difference between CY6203 and C106B. The
   values of over-female-parent heterosis (OFPH) ranged from − 25.53%
   (L/W) to 45% (TGW) (Fig. [60]2). Of these values, OFPHs of agronomic
   traits were all positive. However, OFPHs were only detected positively
   in two quality traits, AC (6.71%) and ASV (15%), while strongly
   negative OFPHs were identified in the other three quality traits HRR
   (− 6.45%), L/W (− 25.53%), and GC (− 9.64%). The values of
   over-male-parent heterosis (OMPH) ranged from − 9.64% (GC) to 44.63%
   (AC). Positive OMPHs were observed in SPW (8.85%), PL (11.3%), ANSPP
   (27.77%), HRR (11.54%), AC (44.63%), and ASV (6.15%). In contrast, the
   intensities of OMPHs and OFPHs were remarkably different in PH, PL,
   TGW, HRR, AC, and L/W, but similar in SPW/D, ANSPP, ASV, and GC. The
   results suggest that the high yield of CY6203 should be primarily
   determined by heterosis in SPW/D and ANSPP, while its high quality
   maybe attributable to the balance among traits AC, ASV, and GC.

Figure 1.

   [61]Figure 1
   [62]Open in a new tab

   Plant phenotype, panicle and grain shape of hybrid rice variety
   Chuanyou6203 and its parents. (A), (B), (C) are Chuan106B, Chuanyou6203
   and Chenghui3203, respectively.

Table 1.

   Phenotypic analysis of hybrid rice CY6203 and its parents.
   Traits SPW/D (g/d) PH (cm) PL (cm) ANSPP TGW (g) HRR (%) L (W) AC (%)
   ASV (Grade) GC (mm)
   C106B 0.215 ± 0.002 95 ± 2.15 25.5 ± 0.54** 140 ± 5.31** 20 ± 1.12
   62 ± 1.35** 4.7 ± 0.15** 16.4 ± 0.64** 6 ± 0.1 83 ± 2.5
   CH3203 0.226 ± 0.002** 110 ± 3.71** 23 ± 1.5 130 ± 2.58 30 ± 2.45**
   52 ± 1.37 3.1 ± 0.15 12.1 ± 0.75 6.5 ± 0.1* 83 ± 1.5
   CY6203 0.246 ± 0.001**^ab 111.6 ± 1.9**^ab 25.6 ± 0.63**^b
   166.1 ± 6.12**^ab 29 ± 2.5**^a 58 ± 1.33**^ab 3.5 ± 0.05**^ab
   17.5 ± 0.48**^ab 6.9 ± 0.15**^a 75 ± 1.8**^ab
   [63]Open in a new tab

   SPW/D: single plant weight per day; PH: plant height; PL: panicle
   length; ANSPP: average number of spikelets per panicle; TGW: 1000-gains
   weight; HRR: head rice ratio (ratio of head milled rice); L/W: grain
   length/grain width; AC: amylose content; ASV: alkali spreading value;
   GC: gel consistency. ^*,**Significant difference with p < 0.05, 0.01,
   respectively. ^a,bSignificance between CY6203 and C106B or CH3203.

Figure 2.

   [64]Figure 2
   [65]Open in a new tab

   Over-male-parent heterosis (OMPH) and Over-female-parent heterosis
   (OFPH) of the main agronomic traits and quality traits. SPW/D, single
   plant weight per day; PH, plant height; PL, panicle length; ANSPP,
   average number of spikelets per panicle; TGW, 1000-gains weight; HRR,
   head rice ratio (ratio of head milled rice); L/W, grain length/grain
   width; AC, amylose content; ASV, alkali spreading value; GC, gel
   consistency.

Genomic variation between two parents of Chuanyou6203

   Using the Illumina-platform, a total of ~ 175 million and ~ 65 million
   paired-end clean reads measuring 150 bp in length were generated with
   average coverage of ~ 61 × and ~ 22 × for the genomes of CH3203 and
   C106B, respectively (Supplementary Table [66]S1). The GC contents of
   CH3203 and C106B were 43.29% and 40.73%, respectively. The high-quality
   reads of each parent were aligned to the Nipponbare reference genome
   MSU RICE GENOME ANNOTATION PROJECT RELEASE 7 (MSU 7), and their mapped
   ratios were 94.98% and 92.63%. A total of 66,319 SNPs and InDels
   between CH3203 (34,102) and C106B (32,217) were detected in the 5′-UTR,
   exon, intronic, and 3′-UTR regions according to the reference genome
   annotation (Fig. [67]3 and Supplementary Table [68]S2), which involved
   7473 genes. The average number of variations per gene was 8.87, and the
   average number of variations per 200 Kb was 35.47. Although most of the
   variations were randomly distributed in 12 chromosomes, there were
   several hot-spot regions with high-density variations (mean
   value = 385.4) on chromosomes 1, 2, 4, 8, 10 and 11. On
   3,000,001 bp-3200000 bp of chromosomes 2, the number of variations per
   200 Kb was up to 578.

Figure 3.

   [69]Figure 3
   [70]Open in a new tab

   Density of different SNPs and InDels in Chenghui3203 and Chuan106B
   genomes and their different cloned yield-, quality- and KEGG
   pathway-associated genes.

Gene expression profiling of Chuanyou6203 and its parents

   We performed RNA sequencing and obtained an average of 67,645,949
   high-quality clean reads for each of the 45 samples after the removal
   of rRNA and low-quality reads (Supplementary Table [71]S3).The mean
   ratio of the high-quality reads mapped to the reference genome was
   82.80%, ranging from 77.59 to 86.61%, and an average of 26,344.67
   mapped unique genes per sample was identified simultaneously. The
   significantly positive correlation (R^2 = 0.625, R^2 = 0.631,
   P < 0.0001) between the RNA-seq and qRT-PCR using the value of
   Log[2]FC[CY6203/CH3203 or C106B] of 18 randomly selected genes
   indicated the accuracy and reliability of the RNA-seq results
   (Fig. [72]4 and Supplementary Table [73]S7). Then, a total of 39,503
   differentially expressed genes (DEGs) were detected among all samples
   (Supplementary Table [74]S4). Based on the hierarchical clustering of
   these samples using the FPKM value (Supplementary Figs. [75]S1 and
   [76]S2), two discrete samples E6203-2 and Y6203-1 were excluded.

Figure 4.

   [77]Figure 4
   [78]Open in a new tab

   Comparison of the log[2](FC) of 17 randomly selected transcripts using
   RNA-Seq and qRT-PCR in Chuanyou6203/Chenghui3203 and
   Chuanyou6203/Chuan106B.

   There were 1164 (42 + 591 + 95 + 436), 642 (23 + 530 + 19 + 70), 4515
   (497 + 3264 + 203 + 551), 2180 (246 + 586 + 355 + 993), and 2615
   (135 + 1103 + 161 + 1216) DEGs between CY6203 and C106B, and similarly
   7617, 915, 1738, 7742 and 8049 between CY6203 and CH3203 in panicles
   and flag leaves at the booting, flowering, and middle filling stages,
   respectively (Fig. [79]5A). Among these DEGs between CY6203 and C106B,
   there were 829 (27 + 475 + 73 + 228) (829/1164 × 100% = 71.22%), 459
   (71.5%), 1,658 (36.72%), 1,488 (68.26%), and 1,453 (55.56%) upregulated
   DEGs, respectively. Among these DEGs between CY6203 and CH3203, there
   were similarly 4,582 (60.15%), 673 (73.55%), 868 (49.94%), 5,662
   (73.13%), and 5927 (73.64%) upregulated DEGs at the booting, flowering,
   and middle filling stages, respectively. At the booting stage, the
   number of DEGs between CY6203 and CH3203 was greater than that of DEGs
   between CY6203 and CY106B, but the ratio of upregulated common DEGs
   between CY6203 and C106B was almost the same (58.26% vs. 59.17% and 72%
   vs. 73.72%) as that of upregulated common DEGs between CY6203 and
   CH3203. At the middle filling stage, the phenomenon was opposite and
   the ratio of upregulated common DEGs between CY6203 and CH3203 was
   higher than (60.8% vs. 42.11% and 67.85% vs. 47.04%) that of
   upregulated common DEGs between CY6203 and C106B. The number of common
   DEGs between C106B and CH3203 in panicles and flag leaves at the
   booting, flowering, and middle filling stages was greater than that of
   common DEGs between CY6203 and its parents (Fig. [80]5B). In addition,
   a total of 436, 70, 551, 993, and 1,216 common DEGs between CY6203 and
   both of its parents were identified at the same stage in panicles and
   flag leaves. Of the common DEGs, the numbers of upregulated DEGs
   between CY6203 and CH3203 were all greater than those of upregulated
   DEGs between CY6203 and C106B in panicles and flag leaves at the
   booting, flowering, and middle filling stages.

Figure 5.

   [81]Figure 5
   [82]Open in a new tab

   DEGs in hybrid rice CY6203 and its parents. (A) Venn diagrams of DEGs
   between CY6203 and its parents in panicles and flag leaves. (B) Venn
   diagrams of DEGs at each stage. Blue font represents upregulated DEGs,
   B: C106B vs. CY6203, R: CH3203 vs. CY6203, and P: C106B vs. CH3203; S:
   Panicle samples7 days before heading. E: Panicle samples 3 days after
   heading. W: Panicle samples 15 days after heading. Y: Sward leaf
   samples 7 days before heading. H: Sward leaf samples 3 days after
   heading.

   The frequency of mRNA editing ratios of each sample of CY6203 followed
   normal distribution (Fig. [83]6 and Supplementary Table [84]S5). The
   average frequency of mRNA editing loci of 15 samples of C106B was 8.13
   per Mb, and that of CH3203 was 7.94 per Mb, and that of CY6203 was
   63.05 per Mb. The results showed that a great quantity of mRNA editing
   phenomena occurred in CY6203 due to its heterozygosity, and most mRNA
   editing ratios were approximately equal to 0.5. Though CH3203 and C106B
   were pure lines, a small amount of mRNA editing phenomena occurred. For
   example, the mRNA editing ratio of the Wx gene was an average of
   approximately 0.5119 in an SNP site (T[1768724] to C[1768724]) of exon
   9 in CY6203, while it was an average of approximately 0.8144 in C106B.

Figure 6.

   [85]Figure 6
   [86]Open in a new tab

   Distribution of frequency of mRNA editing ratios among hybrid rice
   CY6203 and its parents in leaves or panicles.

Identification of coexpressed gene network modules

   We remove no significant DEGs between F[1] (CY6203) and its parents
   (C106B and CH3203) in panicles and flag leaves at booting, flowering,
   and middle filling stages. The 17,105 DEGs were used to construct
   coexpression modules by WGCNA^[87]14. A total of 15,934 DEGs were
   classified into 19 distinct modules (R^2 = 0.85, β = 18) (Supplementary
   Fig. [88]S3), which were generated by a hierarchical clustering tree.
   Among these DEGs, discrete DEGs were classified into the grey module.
   The number of DEGs in the 18 functional modules ranged from 39 (light
   green module) to 2996 (turquoise module). When the Euclidean distance
   was 1.0, 19 modules were divided into three groups (Supplementary Fig.
   [89]S3D). Grey 60 (88 genes), yellow (1574 genes), green-yellow (298
   genes), and midnight blue (186 genes) modules were merged into one
   group, and tan (292 genes), red (1198 genes), black (706 genes), pink
   (673 genes), blue (2905 genes), and purple (337 genes) modules were
   combined into another group, and the remaining modules were classified
   into the other group.

   The expression heatmap of each module and expression histogram of
   corresponding eigengenes showed that the eigengenes were regularly
   expressed at different levels at the different stages or materials
   (Supplementary Fig. [90]S4). For example, blue and turquoise modules
   showed upregulation at the booting and heading stages in panicles. The
   brown module showed upregulation in CY6203 compared with CH3203, except
   for the middle filling stage. The green-yellow module showed that
   eigengenes were higher in CY6203 than C106B. Among the 18 modules, few
   eigengenes showed higher expression in CY6203 than its parents at the
   three stages.

GO and KEGG pathway enrichment in coexpression network modules

   A total of 219 DEGs of 9 modules were significantly enriched in 20 KEGG
   pathways (Fig. [91]7A and Supplementary Tables [92]S8, [93]S9). Out of
   these DEGs, KEGG pathways of photosynthesis-antenna proteins,
   photosynthesis, nitrogen metabolism, and carbon fixation in
   photosynthetic organisms were significantly enriched in the
   green-yellow module (Fig. [94]7B, Supplementary Table [95]S10),
   suggesting that the green-yellow module should be associated with yield
   heterosis. The GO terms of the green-yellow module were significantly
   enriched in GO:0015979 (photosynthesis; FDR = 3.6E−07), GO:0019684
   (photosynthesis, light reaction; FDR = 5.8E−03), and GO:0009765
   (photosynthesis, light-harvesting; FDR = 1.4E−04) (Supplementary Fig.
   [96]S5). The DEG expression levels of CY6203 were inhibited first and
   then increased in pathways of photosynthesis-antenna proteins and
   photosynthesis compared with its parents (Supplementary Figs. [97]S6
   and [98]S7). In contrast, the DEG expression levels of CY6203 in
   pathways of nitrogen metabolism and carbon fixation were continuously
   higher than those of its parents in photosynthetic organisms of all
   three stages, while their increase gradually decreased, especially in
   charge of carbonic an hydrase [EC:4.2.1.1] (LOC_Os01g45274,
   LOC_Os04g33660, and LOC_Os08g36680) and ribulose-bisphosphate
   carboxylase small chain [EC:4.1.1.39] (LOC_Os12g17600, LOC_Os12g19381,
   and LOC_Os12g19470) (Supplementary Fig. [99]S8).

Figure 7.

   [100]Figure 7
   [101]Open in a new tab

   KEGG pathway significantly enriched in WGCNA modules (A) and GO terms
   significantly enriched in green-yellow module (B). The size of the
   bubble indicates the number of genes in each module. The colour of the
   bubble indicates a significance level for a KEGG pathway or GO term
   group of genes.

   Interestingly, the nitrogen metabolism pathway was also identified in
   the yellow module (FDR = 0.02), including 9 enriched DEGs. Of these
   DEGs, 5 known genes, PSR1^[102]15, OsNRT2.3^[103]16,[104]17,
   OsNRT2.4^[105]18, OsGS1;2^[106]19–[107]21 and OsGS2^[108]22,[109]23,
   carried variants between CH3203 and C106B (Supplementary Table
   [110]S9). The mRNA editing ratios of these genes were an average of
   approximately 0.5 in CY6203 (Supplementary Table [111]S5). Their
   expression levels were higher in CY6203 than C106B and were
   approximately equivalent to that of CH3203 at the middle filling stage
   (Supplementary Table [112]S6).

Comparison of cloned gene expression between Chuanyou6203 and its parents
related to yield, quality and KEGG pathway

   According to the annotation of the cloned genes (Fig. [113]3), we
   listed the Log[2]FC[CY6203/CH3203 or C106B] values of cloned genes
   related to the detected KEGG pathways, yield, and quality in
   Supplementary Table [114]S11. Of these genes, the expression levels of
   some genes of CY6203 were higher than that of C106B and lower than that
   of CH3203 at three stages, such as HGW^[115]24 (Fig. [116]8), while
   some genes exhibited the opposite phenotype, such as Gn1a^[117]25. Some
   were synchronously higher than that of its parents at the booting
   stage, and then synchronously lower than that of its parents at the
   middle filling stage, such as Gnp4^[118]26. The expression level of
   most genes related to chlorophyll was higher in CY6203 than that of
   C106B, but lower than that of CH3203 at the middle filling stage, for
   example, Fd1, LHCB, and CAB2R. While the genes related to grain quality
   were opposite, such as Wx^[119]27–[120]29 and GL7^[121]30. The
   expression level of ALK^[122]31,[123]32 was higher in CY6203 than in
   its parents at the middle filling stage (Fig. [124]8), which might be
   the main reason why CY6203 showed over-parent heterosis in ASV
   (Fig. [125]2).

Figure 8.

   [126]Figure 8
   [127]Open in a new tab

   Partial cloned gene expression level of CY6203 compared with its
   parents using Log[2] fold change (Log[2]FC[CY6203/CH3203 or C106B]).
   Log(FC)_SR represents Log[2]FC[CY6203/CH3203] of panicle samples7 days
   before heading. Log(FC)_SB represents Log[2]FC[CY6203/C106B] of panicle
   samples7 days before heading. Log(FC)_ER represents
   Log[2]FC[CY6203/CH3203] of panicle samples 3 days after heading.
   Log(FC)_EB represents Log[2]FC[CY6203/C106B] of panicle samples in
   3 days after heading. Log(FC)_WR represents Log[2]FC[CY6203/CH3203] of
   panicle samples 15 days after heading. Log(FC)_WB represents
   Log[2]FC[CY6203/C106B] of panicle samples in 15 days after heading.
   Log(FC)_YR represents Log[2]FC[CY6203/CH3203] of sward leaf samples
   7 days before heading. Log(FC)_YB represents Log[2]FC[CY6203/C106B] of
   sward leaf samples7 days before heading. Log(FC)_HR represents
   Log[2]FC[CY6203/CH3203] of sward leaf samples 3 days after heading.
   Log(FC)_HB represents Log[2]FC[CY6203/C106B] of sward leaf samples3
   days after heading.

Discussion

   The use of crop heterosis has notably helped to ensure world food
   security. For over a hundred years, researchers have been exploring the
   molecular mechanism of heterosis, starting from the previously proposed
   hypotheses of dominance, over-dominance, and epistasis to later
   transcriptome analysis at gene expression levels^[128]1. It is
   generally believed that the DEGs between CY6203 and its parents can
   control the heterosis of the hybrid^[129]2,[130]3,[131]6–[132]8. Wei et
   al. identified 3,926 DEGs between LYP9 and its parents and between the
   parents, 93-11, and PA64s^[133]3. Zhang et al. indicated that
   polymorphic promoter cis-regulatory elements were closely correlated
   with DEGs with an additive, as well as over- and under-dominance, gene
   action^[134]33. Huang et al. believed that numerous superior alleles
   contributed to heterosis, but only a small number of loci with strong
   overdominance affected heterosis in hybrids; for example, Ehd1, Hd1,
   Ghd7, and OsSoc1 strongly affected heading date, and Gn1, Ghd7, and
   OsSPL14 affected grain number, and qSW5 and Wx affected chalky grain
   rate^[135]4. Chen et al. also believed that the overdominant effect was
   a major contributor to the grain number heterosis of WFYT025^[136]5.

   Therefore, heterozygosity is generally a prerequisite for gene
   expression and phenotypic variation in hybrids^[137]1. In this study,
   CY6203 with excellent quality and high yield showed visibly overparent
   heterosis in SPW/D (OMPH:8.85%; OFPH:14.42%), ANSPP (OMPH:27.77%;
   OFPH:18.64%), AC (OMPH:44.63%; OFPH:6.71%) and ASV (OMPH:6.15%;
   OFPH:15%). A total of 7473 heterozygous genes in CY6203 were identified
   due to variant sites between CH3203 and C106B (Supplementary Table
   [138]S5). In the results of mRNA editing (Supplementary Table [139]S5),
   transcripts of 14,979 genes from both parents were simultaneously
   expressed in CY6203, and approximately 40.61% of mRNA editing ratios
   were between 0.4 and 0.6 (Fig. [140]5). Interestingly, 1.68% of mRNA
   editing events (editing ratio ≥ 0.8) in CY6203 favoured one of its
   parents at three stages or a particular stage, suggesting that the
   hypothetical heterosis mechanism of CY6203 might involve dominance or
   epistasis. For example, the average 0.8376 of GL7^[141]30 type belonged
   to CH3203 type at three stages, and 0.8576 of the Edh2 type belonged to
   CH3203 type at only the booting stage. We also noticed that the DEG
   number of CY3203 vs. CY6203 (7617 DEGs, 4582 upregulated) substantially
   exceeded that of C106B vs. CY6203 (1164 DEGs, 829 upregulated) in
   panicles at the booting stage, but at the middle filling stage, the
   number of DEGs was quite the opposite (4515 and 1658 vs. 1738 and 868).
   These results indicated that CH3203 and CY106B together contributed to
   the high yield and good quality of CY6203.

   To explain the genetic mechanism of high yield and quality of CY6203,
   we summarized the following two points. First, the DEGs in the
   green-yellow module mainly contributed to the increase in the source of
   CY6203 due to an increase in photosynthetic efficiency and nitrogen
   utilization efficiency, and a small number of DEGs related to the grain
   number added spikelet number per panicle amplified its sink
   (Fig. [142]9). OMPHs for SPW/D and ANSPP were 8.85% and 27.77%,
   respectively. Similarly, OFPHs for SPW/D and ANSPP were 14.42% and
   18.64%, respectively (Figs. [143]2, [144]9). Epigenetic regulation of
   circadian-mediated changes in chlorophyll biosynthesis and starch
   metabolism is a direct reason for growth vigour in plant
   hybrids^[145]1. Further studies showed that an appropriate reduction in
   the size of the chlorophyll antenna or chlorophyll content can improve
   the photosynthetic CO[2] uptake rates and nitrogen utilization
   efficiency^[146]34–[147]37. Of the 298 DEGs in the green-yellow module,
   31 DEGs were significantly enriched in pathways of
   photosynthesis-antenna proteins, photosynthesis, nitrogen metabolism,
   carbon fixation in photosynthetic organisms, glyoxylate, and
   dicarboxylate metabolism and linoleic acid metabolism (Supplementary
   Table [148]S8). In the biological process, photosynthesis negatively
   regulated light-harvesting (Supplementary Fig. [149]S5). The pathway of
   nitrogen metabolism was also identified from the yellow module. Of the
   five cloned genes, PSR1^[150]15 encoded a nitrite reductase with nitric
   acid assimilation, OsNRT2.3^[151]16,[152]17 and OsNRT2.4^[153]18 were
   inferred to control regulation of rice high-affinity nitrate transport,
   and OsGS1;2^[154]19–[155]21 and OsGS2^[156]22,[157]23 showed
   coexpression affecting nitrogen metabolism and plant growth. In
   summary, we speculate that the photosynthetic CO[2] uptake rates and
   nitrogen utilization efficiency are improved in CY6203 due to its
   sustained better chlorophyll accumulation and suitable size of the
   photosynthesis antenna from the booting to the middle filling stage
   (Fig. [158]9). Therefore, these genes were coexpressed, which increased
   nitrogen utilization efficiency and carbon dioxide absorption from the
   booting to the middle filling stage in CY6203. Thus, the coordinated
   expression of DEGs in the green-yellow module can improve the
   photosynthetic efficiency and nitrogen utilization efficiency of
   CY6203.

Figure 9.

   [159]Figure 9
   [160]Open in a new tab

   Predicted high yield mechanism of hybrid variety Chuanyou6203. the
   Green colour represents pathway in the green-yellow module. the
   yellowish-green gradual discoloration represents the pathway of
   nitrogen metabolism in the green-yellow and yellow module. The
   green–red gradual discoloration represents the gene expression level of
   F[1], Chuanyou6203 compared with its parents. The green arrow
   represents negative regulation, and 1 and 2 represent panicle samples
   7 days before heading. 3 and 4 represent panicle samples 3 days after
   heading. 5 and 6 represent panicle samples 15 days after heading. 7 and
   8 represent sword leaf samples 7 days before heading. 9 and 10
   represent sword leaf samples 3 days after heading. SPW/D: single plant
   weight per day; ANSPP: average number of spikelets per panicle.

   Among 1079 cloned genes with different alleles between CH3203 and
   C106B, 15 genes were related to yield trait-based previous reports at
   [161]https://www.ricedata.cn/ (Fig. [162]3). Only SPP1 was inferred to
   control the number of spikelets per panicle^[163]38. The expression
   level of CY6203 was higher than that of its parents at the booting
   stage (Supplementary Table [164]S11). However, the gene (new name:
   OsJAR2) was determined to be involved in the wound- and
   pathogen-induced jasmonic acid signalling^[165]39. Although two
   well-known genes controlling the grain number per panicle, Gn1a^[166]25
   and Gnp4^[167]26,[168]40, without variants between CH3203 and C106B,
   showed middle-parent and over-parent expression levels in CY6203 at the
   booting stage, respectively (Supplementary Table [169]S11). Ashikari et
   al.^[170]25 reported that the decrease in the activity of Gn1a
   increased the spikelet number. Similarly, increasing the expression
   level of Gnp4 also increased the spikelet number^[171]26. Therefore, we
   infer that the coregulation of Gn1a and Gnp4 jointly increases the
   average number of spikelets per panicle, resulting in the amplified
   sink of CY6203.

   Second, the balanced expression of the major high-quality alleles of
   C106B and CH3203 in CY6203 contributed to the outstanding quality of
   CY6203. At the middle filling stage, the number of DEGs between CY6203
   and C106B was greater than that of DEGs between CY6203 and CH3203.
   However among coexpressed DEGs, the ratio of upregulated DEGs between
   CY6203 and CH3203 was greater than that of DEGs between CY6203 and
   C106B (Fig. [172]5A), which might indicate the balanced expression of
   some related quality genes of C106B and CH3203 in CY6203. Complex
   genetic networks that affect AC, GC, and gelatinization temperature
   (GT) determine the cooking and eating quality (ECQs) of rice^[173]41
   (Fig. [174]10). Among these genes, Wx and SSII-3 are the two essential
   genes that affect AC, GC, and GT. Different alleles of Wx lead to
   regional variation in rice AC and have affected consumer
   preferences^[175]27–[176]29,[177]41–[178]43. The substitution of bases