Copy number variation of the restorer Rf4 underlies human selection … – Nature.com

Identification of variants and haplotypes of the Rf4 locus in rice cultivars

Our and others previous studies identified five variants at the Rf4 locus, namely, Rf4M and Rf4I (in indica fertile restorer lines), rf4aus (in circum-aus rice, see below), rf4i (in indica CMS-WA/maintainer lines), and rf4j (in japonica rice)15,17,18. To further reveal the origin and evolution of the Rf4 locus, we re-analyzed the genome sequences of the Rf4 locus using publicly available rice genome information for ZS97B (Zhenshan97B, a maintainer line with rf4i) (http://rice.hzau.edu.cn/rice_rs3/)23, Nip (Nipponbare, a japonica line with rf4j) (https://rapdb.dna.affrc.go.jp/)38,39, and MH6323 and SH498 (https://www.mbkbase.org/rice)40 (Shuhui498, fertility restorer line carrying functional Rf4). We focused on part of a PPR cluster that includes the Rf4 locus based on the reference genome of the restorer line MH63 (Fig.1a).

a Homologous gene relationship (microsynteny block) at the Rf4 locus in O. sativa ssp. indica and O. sativa ssp. japonica. b Different haplotypes of the Rf4 complex locus region involving in the Rf4a-copy (Copy-a) and Rf4b-copy (Copy-b) identified in Asian cultivated rice. Gray background shows other PPR genes, white background shows non-functional rf4 variants, black background shows functional Rf4 variants. j: japonica, aus:circum- aus, i: indica, I: IR24, M: MH63, Nip: Nipponbare (a japonica variety). IR24 and MH63 are indica restorer lines.

At the Rf4 locus region, we first examined two lines that lack the ability to restore fertility and identified three PPR genes (PPR7 [Os10g0495400], rf4j [Os10g0495200], and PPR10 [Os10g0495100]) in Nip (Fig.1a) and three PPR genes (PPR7, rf4i, PPR10) in ZS97B (Fig.1a). The rf4i variant was pseudogenized and non-functional due to the presence of a premature termination codon (Fig.1a).

We next examined two lines that have the ability to restore fertility. The Rf4 locus regions of MH63 and SH498 possess seven PPR genes, in addition to previously known functional Rf4 variant (here defined as Rf4a, of functional Rf4 in Copy-a site of Rf4 locus), PPR7, PPR8, and three copies of PPR10 genes, which included another functional Rf4 variant Rf4b identified in the Copy-b site of Rf4 locus (Fig.1a). The interval between Rf4a and Rf4b is 74.8kb, and the Rf4a and Rf4b coding sequences show 100% amino acid identity. The regions 7.5kb upstream of the start codon and 1.5kb downstream of the stop codon of Rf4a and Rf4b showed 98.4% and 99.7% similarity, respectively (Supplementary Data1). These findings reveal that different rice varieties show SV and CNV at the Rf4 locus.

Then, we further investigated SV and CNV at the Rf4 locus region in 311 rice cultivars by PCR amplification and sequencing. To this end, we designed site-specific PCR primers based on single-nucleotide polymorphisms (SNPs, Supplementary Table1). The Rf4a and Rf4b genes were amplified using a common primer F1 combined with the site-specific reverse primers a-R and b-R, respectively; the rf4i fragment was amplified using a primer pair F2 and i-R (Fig.1a). In total, there were seven variants and eight haplotypes (H1H8) based on the combination of the Copy-a and Copy-b variants at the Rf4 locus in the modern rice cultivars (Fig.1b and Supplementary Data2). Sequence analyses demonstrated that the restorer lines contain five haplotypes with one- or two-copy Rf4 variants, including Rf4aI (H2), Rf4bM (H6), Rf4aM-Rf4bM (H1), Rf4aI-rf4b (H7), and rf4a-Rf4bM (H8), while all current CMS-WA lines (and their maintainer lines) carry the rf4i (H4) variant; the rf4j (H5) variant is present mainly in japonica cultivars. Among these variants, rf4aus was previously named as H318 (Fig.1b).

To trace the evolutionary history of Rf4 in the Oryza genus, we further investigated the CNV and sequence polymorphisms of Rf4 and rf4 variants among wild rice and landrace rice accessions. In addition to the seven Rf4 and rf4 variants identified in the cultivars, 61 variants were identified in the Copy-a and/or Copy-b sites of the Rf4 locus (from GG to AA-genome species) (Supplementary Data25), pointing to the prevalence of SV and CNV in the Rf4 locus.

We then performed a BLAST search for putative orthologs and homologs of Rf4 and rf4 in Poaceae genomes in the GenBank database (https://www.ncbi.nlm.nih.gov/). The putative homologs in Aegilops tauschii, Setaria italica, Setaria viridis, Sorghum bicolor, Triticum aestivum, Triticum dicoccoides, and Zea mays were similar to the rice Rf4 and rf4 variants (Supplementary Fig.1). The homologous gene LOC117839145 of S. viridis shared the highest nucleotide sequence similarity of 72.2% with the Rf4M variants in the rice cultivars (Supplementary Data6).

We next used the nucleotide sequences to construct a phylogenetic tree of Rf4 and rf4 in the Oryza genus using LOC117839145 (S. viridis) as the outgroup. The variants of different species were divided into six homologous lineages (Supplementary Fig.2). Among these, the rf4i, rf4a, and rf4b clades were closely related to each other, and Rf4M and Rf4I were in the same clade, whereas the lineages containing rf4aus and rf4j were closely related (Supplementary Fig.2). During the evolution of wild rice, a considerable number of rf4a, rf4b, and Rf4 variants were developed, but only two variants each of rf4j and rf4i were generated (Supplementary Fig.2). Among Rf4 variants in the different rice cultivars, rf4a, rf4b, rf4aus, rf4i, and Rf4 mainly occur in the indica group, rf4j mainly occurs in the japonica group, and rf4i was only found in bred maintainer lines and CMS-WA lines (Supplementary Data2).

In the wild rice O. meyeriana (GG-genome), only a putative one-copy Rf4 variant appeared at the Copy-a site, but not the Copy-b site, thus, this Copy-a variant sequence likely represents the primitive sequence of ancestral Rf4, which we named Anc-Rf4 (Fig.2 and Supplementary Data3). Along with sequence variation and genetic recombination, various haplotypes (H13-H68) consisting of Rf4-like and/or rf4-like variants were generated in the genomes of EE (O. australiensis), CCDD (O. alta, O. grandiglumis, and O. latifolia), CC (O. eichingeri, O. officinalis, and O. rhizomatis), BBCC (O. minuta), BB (O. punctata), and AA genomes (O. meridionalis, O. glumaepatula, O. rufipogon, and O. nivara) of these wild species (Fig.2 and Supplementary Data3); including one-copy haplotypes Rf4a/rf4a-likes (a group of Rf4a-like or rf4a-like variants with SNPs) at the Copy-a site, and Rf4b/rf4b-likes (a group of Rf4b-like or rf4b-like variants with SNPs) at the Copy-b site (H13-H27) (Fig.2 and Supplementary Data3). Further, various two-copy haplotypes containing Rf4-like and/or rf4-like variants were formed, including Rf4aM-rf4b-likes, rf4a-likes-Rf4bM, rf4a-likes-Rf4b-likes, Rf4a-likes-Rf4b-likes, and rf4a-likes-rf4b-likes (H28-H68) (Fig.2 and Supplementary Data3). However, no Rf4 and rf4 variants were detected in the tested O. longistaminata, O. barthii, and O. glaberrima accessions at the Copy-a or Copy-b sites (Fig.2 and Supplementary Data2 and 3). The variants such as rf4i, Rf4aI, and rf4b were first identified in O. australiensis (EE-genome), while Rf4M and rf4j were first identified in O. officinalis (CC-genome), and rf4aus and rf4a appeared only in indica in O. sativa (Fig.2 and Supplementary Data24).

Current Rf4 haplotypes may have originated from an ancestral type of Rf4 (Anc-Rf4, H69), which first emerged at Copy-a in the oldest wild rice O. meyeriana (GG-genome). Sequence variation, gene duplication, and recombination events resulted in new one-copy (H12-H27) or two-copy (H28-H68) haplotypes of R/rf4a-likes (a group of Rf4a-like or rf4a-like variants with SNPs) and/or R/rf4b-likes (a group of Rf4b-like or rf4b-like variants with SNPs) in wild rice. During evolution, along with natural and human selections, the nascent one-copy and two-copy Rf4 and/or rf4 haplotypes gradually migrated into the lineages of O. rufipogon, O. nivara, and O. sativa. Rf4 haplotypes were not detected in tested accessions of O. longistaminata, O. barthii, and O. glaberrima. Among the eight Rf4 haplotypes (H1H8) in modern cultivars, three two-copy haplotypes (indicated by asterisks *, the percentages on the left showed frequency of specific haplotype in the tested accessions) are predominant in restorer lines. H1 (Rf4aM-Rf4bM) and H7 (Rf4aI-rf4b) haplotypes first appeared in O. nivara and O. australiensis (EE-genome), respectively, while H8 (rf4a-Rf4bM) is present only in landraces and modern cultivars of O. sativa. The one-copy variant Rf4aI in H2 of modern cultivars was only found in H7, suggesting that H2 may be derived from loss of rf4b in H7 from O. rufipogon. Other one-copy haplotypes such as H3 (rf4aus), H4 (rf4i), H5 (rf4j) and H6 (Rf4bM) first emerged in O. sativa, O. australiensis, O. officinalis (CC-genome), and O. glumaepatula (AA-genome), respectively. The number in the brackets next to each species represents frequency of the haplotypes detected in the species. MYA: million years ago (divergence time).

To confirm the biological functions of the rf4 haplotypes in the rice cultivars, we analyzed amino acid sequence variation among six proteins: rf4a, rf4b, rf4j, rf4aus, Rf4M, and rf4i. Whereas rf4i is a truncated product containing only eight PPR motifs, the other five proteins contain 18 PPR motifs and share high sequence similarity (93.595.3%) (Supplementary Data7). The rf4j, rf4b, rf4aus, and Rf4M proteins contain 782 amino acids, while rf4a has 798 amino acids (Supplementary Fig.3). In contrast to rf4j vs. Rf4M, which harbor 37 amino acids differences15,18, we detected 68 amino acid differences between rf4a and Rf4M, including two amino acid insertions at the N-terminal regions and 14 amino acid insertions at the C-terminal regions. In addition, rf4b has 38 amino acid differences, and rf4aus harbors 51 amino acid differences to Rf4M (Supplementary Fig.3). Compared to Rf4M, all rf4 proteins contain the common 14 amino acid substitutions at the PPR13, PPR14, and PPR15 motifs (Fig.3a and Supplementary Fig.3), suggesting that amino acid substitutions at these PPRs motifs are important for the fertility restoration of CMS-WA.

a Amino acid differences used to identify the functional Rf4 and non-functional rf4 variants in three PPR motifs (PPR13, PPR14, and PPR15). b Male fertility of transgenic (T0) lines of Jin23A (a CMS-WA line) carrying different transgenes (t) in the hemizygous state. For each complementary construct, at least 10 independent transgenic lines with similar phenotype were obtained. The pollen phenotype of three independent lines was shown in Fig.3b and Supplementary Fig.4. Viable pollen stain black; inviable pollen stain light brown. Scale bar: 50 m.

To verify our hypothesis that these rf4 variants may be non-functional for fertility restoration, binary vectors containing rf4a, rf4b, rf4j, rf4aus and Rf4M, all driven by the native promoter from Rf4, were constructed and introduced into the CMS-WA line Jin23A (WA352c/H4H4) via transformation. The male fertility of transgenic T0 plants with the transgene containing Rf4M was partially restored (Fig.3b and Supplementary Fig.4). Nevertheless, T0 transgenic plants carrying rf4a, rf4b, rf4j, or rf4aus remained completely sterile (Fig.3b and Supplementary Fig.4). These results confirm the notion that all the rf4 haplotypes are non-functional for fertility restoration of CMS-WA.

We demonstrated that the Rf4 locus has undergone CNV in different rice cultivars, and assumed that CNV at the Rf4 locus may be associated with the effect on fertility restoration of CMS-WA. To verify the CNV-mediated gene dosage effect of Rf4, we produced various lines with different copy numbers of Rf4 by crossing Jin23A (WA352c/H4H4) with two near-isogenic lines of Rf4: ZSRf4I (WA352c/H7H7) and ZSRf4M (WA352c/H1H1), knocking out Rf4 in ZSRf4M lines, and transforming Jin23A with functional Rf4, respectively (Figs.4, 5 and Supplementary Figs.5, 6). Firstly, the pollen viability (assessed by staining with I2KI) of the F1 plants derived from Jin23AZSRf4I (WA352c/H4H7) and Jin23AZSRf4M (WA352c/H4H1) were ~71% and ~88%, respectively, while the seed setting rates of these F1 plants were ~34% and ~52%, respectively (Fig.4a and Supplementary Figs.5a, 7a, b), showing the male fertility of the two-copy Rf4-carrying plants was higher than those of the one-copy Rf4-carrying plants (Fig.4 and Supplementary Figs.5, 7). Moreover, the fertile anthers were pollen-filled and yellowish, while the sterile anthers appeared thin and whitish (Figs.4, 5 and Supplementary Figs.5, 6).

a, b Pollen viability based on staining (upper panels), anther phenotype (middle panels), and seed setting rate (lower panels) of Jin23AZSRf4I (WA352c/H4H7), Jin23AZSRf4M (WA352c/H4H1) and rf4am-rf4bm-mF1/rf4am-rf4bm mutant lines (by CRISPR/Cas9 editing) in the ZSRf4M background. Red indicates the non-functional rf4a/bm after knockout of Rf4a/bM. Scale bars: 50 m in the upper panels, 1cm in the middle panels, and 5cm in the lower panels. ce Transcript levels of Rf4 (c) and WA352c (d, e) in different lines. UFC1 (UFM1-Conjugating Enzyme 1) and atp6 (a mitochondrial gene) served as internal references for Rf4 and WA352c expression, respectively. Data are shown as meanSD, n=3 biological replicates. Significant differences between two samples were determined by two-tailed Students t-test (**P<0.01, ***P<0.001, ****P<0.0001). f Sequencing of the Rf4a/bM-knockout plants derived from CRISPR/Cas9 editing. The underlined bases show protospacer adjacent motifs (PAMs). The positions highlighted in red indicate the targeted mutations. Source data are provided as a Source data file.

a Pollen staining (upper panels) to reveal viable pollen (dark), anther phenotype (middle panels) and seed setting rate (lower panels) of Jin23A/Rf4t- (WA352c/H4Rf4t-), Jin23A/Rf4tRf4t (WA352c/H4H4Rf4tRf4t), and Jin23A/Rf4tRf4tZSRf4I (WA352c/H4H7Rf4t-). Scale bars: 50 m in the upper panels, 1cm in the middle panels, and 5cm in the lower panels. b, c Transcript levels of Rf4 (b) and WA352c (c) in different lines. Rf4t indicates the Rf4 transgene, Rf4t- indicates transgenic hemizygotes. Data are shown as meanSD, n=3 biological replicates. Significant differences between two samples were determined by two-tailed Students t-test (**P<0.01, ***P<0.001, and NS represents No Significance). Source data are provided as a Source data file.

Furthermore, we knocked out both Rf4aM and Rf4bM in ZSRf4M line by CRISPR/Cas9 editing and obtained several rf4am and rf4bm mutants, which showed full abortion of pollen and spikelet (Fig.4b, Supplementary Fig.5b, and Supplementary Table2). Then, we selected three rf4am and rf4bm loss-of-function mutant lines (carrying different editing patterns) and crossed them with wild type ZSRf4M to test the dosage effect of Rf4 in the resultant mutant F1 (mF1) (Fig.4b and Supplementary Fig.5b). As expected, the pollen and spikelets of mF1 plants carrying two copies of Rf4 (WA352c/H1h1) also showed lower pollen viability (~85%) and seed setting rate (~48%), compared to those of wild type ZSRf4M (~92% and ~72%, respectively), which carries four copies of Rf4 (Fig.4b and Supplementary Figs.5b, 7c, d).

To clarify the connection between the Rf4 CNV-mediated gene dosage effect and the WA352c repression in fertility restoration, we performed qRT-PCR analysis of Rf4 and WA352c expression in anthers of different plants at the microspore mother cell stage. The expression level of Rf4 was twice as high in Jin23AZSRf4M (WA352c/ H4H1) vs. Jin23AZSRf4I (WA352c/H4H7) (Fig.4c and Supplementary Fig.5c), whereas the expression pattern of WA352c was opposite to that of Rf4 (Fig.4d and Supplementary Fig.5d). The level of WA352c transcripts was higher in the rf4am-rf4bm-mF1 and rf4am-rf4bm lines compared to ZSRf4M (Fig.4e and Supplementary Fig.5e).

In addition, we generated Rf4t-transgenic lines using CMS line Jin23A as recipient and selected a homozygous Jin23A/Rf4tRf4t (WA352c/H4H4Rf4tRf4t) plant from the T1 population to cross with ZSRf4I generating F1 plants Jin23A/Rf4tRf4tZSRf4I (WA352c/H4H7Rf4t-). We then acquired a series of materials harboring different copy numbers of the Rf4t transgene in the F2 population. As expected, lines with two-copy Rf4, including Jin23A/Rf4tRf4t (WA352c/H4H4Rf4tRf4t) and Jin23A/Rf4tRf4tZSRf4I (WA352c/H4H7Rf4t-), exhibited higher pollen viability (~87%, ~90%) and spikelet fertility (~52%, ~53%) than J23A/Rf4t- lines harboring a single copy of Rf4 (WA352c/H4H4Rf4t-), which showed ~73% pollen fertility and ~36% spikelet fertility (Fig.5a and Supplementary Figs.6a, 7e, f).

The level of Rf4 transcripts in Jin23A/Rf4tRf4t (WA352c/H4H4Rf4tRf4t) and Jin23A/Rf4tRf4tZSRf4I (WA352c/H4H7Rf4t-) was about twice as high as that in Jin23A/Rf4t- (WA352c/H4H4Rf4t-) in qRT-PCR assays (Fig.5b and Supplementary Fig.6). Consistent with this, the pattern of WA352c transcript levels was opposite to that of Rf4 (Fig.5c and Supplementary Fig.6c).

Together, these results supported the hypothesis that the dosage effect caused by different copy number of functional Rf4 plays an important role in fertility restoration of CMS-WA.

The ability to restore fertility of CMS lines restricts hybrid rice production; therefore, genetic resources with two-copy Rf4 might be beneficial for breeding stronger restorer lines for CMS-WA in hybrid rice production. To investigate the relationship between haplotypes of the Rf4 locus and the application of major restorer lines in China, we obtained data about the planting areas of hybrid rice varieties and crossing combinations of restorer lines for CMS-WA from the China Rice Data Center (https://www.ricedata.cn/). With regards to planting area of three-line hybrid rice cultivars, we selected the top sixteen related restorer lines for analysis: six lines (MH63, Ce64-7, CDR22, FuHui838, MH86, and ChengHui727) carried H1 (Rf4aM-Rf4bM), eight lines (MiYang46, Gui99, IR24, R402, Shuhui527, Huazhan, Guanghui998, and Minhui3301) harbored H7 (Rf4aI-rf4b), XianHui207 contained H8 (rf4a-Rf4bM), and MianHui725 possessed H6 (Rf4bM) (Supplementary Table3). As expected, hybrid rice varieties using six restorer lines carrying two-copy Rf4 (H1) had a larger total planting area (135,998,667 hectares) than the hybrid rice varieties using ten restorer lines having the one-copy Rf4 (78,015,332 hectares) (Fig.6). Notably, the two most widely planted hybrid rice varieties were bred from two elite restorer lines, MH63 and Ce64-7, both carrying the two-copy Rf4 (Supplementary Table3).

Total planting areas and relative percentage of hybrid rice varieties in China using 6 restorer lines carrying the two-copy Rf4 (H1) and 10 restorer lines with one-copy Rf4 (H6, H7, H8). The data information is given in Supplementary Table3.

The rice mitochondrial CMS-WA gene WA352c was generated in O. rufipogon via multiple rounds of recombination/protogene formation/functionalization, and WA352c has been widely utilized in hybrid rice breeding1. However, how WA352c co-evolved with Rf4 remains to be uncovered. To explore the evolutionary relationship of WA352c with Rf4, we analyzed their sequence structures in different O. rufipogon species and rice cultivars. The functional WA352c gene only coexisted with three haplotypes (H7, H14, and H28) of the Rf4 locus in O. rufipogon populations (Table1). This finding suggests that the first CMS-WA germplasm with abortive pollen discovered from an O. rufipogon population, called Wild Abortive, carrying a rf4a-like variant in addition to WA352c (Fig.7, Table1).

An Oryza rufipogon population (Wild Abortive) with the mitochondrial sterility gene WA352c and non-functional rf4a-like showing pollen abortion was found and used as the female parent for breeding CMS-WA lines by backcrossing with indica maintainer lines containing rf4i. A hybrid rice variety was bred by crossing the CMS-WA line with a restorer line (carrying one-copy or two-copy Rf4).

During the process of CMS-WA line breeding, the rf4a-like variant was replaced by rf4i through backcrossing with indica maintainer lines that harbor the rf4i variant, resulting in the current CMS-WA lines (Fig.7, Table1). CMS-WA (WA352c/rf4irf4i), maintainer (rf4i), and restorer lines (Rf4) made up the CMS-WA/Rf system for three-line hybrid rice production (Fig.7). Based on the above results and the previous finding that WA352c originates in O. rufipogon1, it appears that Rf4 and rf4 (except for rf4a and rf4aus) originated earlier than WA352c (Fig.7, Table1, and Supplementary Data3) and that the replacement of rf4i derived from indica type maintainer lines occurred during the creation of modern CMS-WA lines.

To facilitate the identification of Rf4 (rf4) haplotypes in hybrid rice breeding, we selected and optimized a set of eight Rf4 variant-specific PCR-based markers based on the SNPs at the Rf4 locus (Supplementary Table1). To confirm the utility of this set of markers, these primers were used to investigate the genotypes of 304 Asian cultivated rice germplasms. PCR products of these lines were first divided into three types, Copy-a, Copy-b, and Copy-a/-b, using two PCR markers (Copy-a-332 bp-F/R and Copy-b-282 bp-F/R), which generated 332-bp- and 282-bp PCR products from Copy-a and Copy-b, respectively (Fig.8 and Supplementary Table1). Then the variants and haplotypes of Rf4 and rf4 were determined using six specific primer sets. The fragments amplified from one-copy Rf4 from the Copy-a or Copy-b in rice varieties with the Rf4aI haplotype (such as Jalmagna and GH102) or Rf4bM haplotype (such as MH725 and R60) were 262bp long (Fig.8a). No products of functional Rf4 were amplified from japonica, indica, and aus varieties carrying haplotypes of rf4j, rf4i, and rf4aus, but a 372-bp, 358-bp, and 351-bp products of rf4 were obtained from the Copy-a of these varieties, respectively (Fig.8b).

Eight pairs of variant-specific primers were used for PCR to determine the Rf4 and rf4 haplotypes of different rice lines. a One-copy Rf4, b one-copy rf4, c two-copy Rf4. White dashed lines separate four different patterns of one-copy Rf4. GH993 (Guanghui993), XH207 (Xianhui207), ZSRf4I (Zhenshan Rf4I), IR8, Jalmagna, GH102 (Guanghui 102), MH725 (Mianhui725), R60, J23 (Jin23), ZS97 (Zhenshan97), IR64, 9311, MH63 (Minghui63), SH498 (Shuhui498), FH838 (Fuhui838), and IR30 are indica cultivars. Nip, T65, ZH11, and 9522 are japonica cultivars. CISOKAN, BaXiang, AR (Albania Rice), and ZCD13 (Zacaodao13) arecircum-aus cultivars. Actin1 was used as the PCR control. The PCR experiments for each sample were independently repeated at least three times with similar results. Source data are provided as a Source data file.

Varieties with the rf4a-Rf4bM haplotype (such as GH993 and XH207) or the Rf4aI-rf4b haplotype (such as ZSRf4I and IR8) all carried one-copy Rf4 (Fig.8a). They shared four common PCR products: Copy-a (332bp), Copy-b (282bp), Rf4 (262bp), and rf4a/b (197bp). Varieties with the rf4a-Rf4bM haplotype also generated another product: rf4a (446bp). Varieties carrying two-copy Rf4 with the Rf4aM-Rf4bM haplotype (such as MH63, SH498, FH838, and IR30) generated three PCR products: Rf4 (262bp), Copy-a (332bp), and Copy-b (282bp) (Fig.8c). Taken together, these results demonstrated that these primer sets are useful PCR markers for the rapid genotyping to accelerate the screening of strong restorer lines with the two-copy Rf4.

Go here to see the original:
Copy number variation of the restorer Rf4 underlies human selection ... - Nature.com