A P‐type pentatricopeptide repeat protein ZmRF5 promotes 5′ region partial cleavages of atp6c transcripts to restore the fertility of CMS‐C maize by recruiting a splicing factor

Summary

A fast evolution within mitochondria genome(s) often generates discords between nuclear and mitochondria, which is manifested as cytoplasmic male sterility (CMS) and fertility restoration (Rf) system. The maize CMS‐C trait is regulated by the chimeric mitochondrial gene, atp6c, and can be recovered by the restorer gene ZmRf5. Through positional cloning in this study, we identified the nuclear restorer gene, ZmRf5, which encodes a P‐type pentatricopeptide repeat (PPR) family protein. The over‐expression of ZmRf5 brought back the fertility to CMS‐C plants, whereas its genomic editing by CRISPR/Cas9 induced abortive pollens in the restorer line. ZmRF5 is sorted to mitochondria, and recruited RS31A, a splicing factor, through MORF8 to form a cleaving/restoring complex, which promoted the cleaving of the CMS‐associated transcripts atp6c by shifting the major cleavage site from 480th nt to 344 th nt for fast degradation, and preserved just right amount of atp6c RNA for protein translation, providing adequate ATP6C to assembly complex V, thus restoring male fertility. Interestingly, ATP6C in the sterile line CMo17A, with similar cytology and physiology changes to YU87‐1A, was accumulated much less than it in NMo17B, exhibiting a contrary trend in the YU87‐1 nuclear genome previously reported, and was restored to normal level in the presence of ZmRF5. Collectively these findings unveil a new molecular mechanism underlying fertility restoration by which ZmRF5 cooperates with MORF8 and RS31A to restore CMS‐C fertility in maize, complemented and perfected the sterility mechanism, and enrich the perspectives on communications between nucleus and mitochondria.

Introduction

Cytoplasmic male sterility (CMS) is a phenomenon genetically controlled by the factors in the cytoplasm in plant and exhibits normal development except for male organs. CMS was first reported in maize (Rhoades, 1931) and is found in more than 200 plant species (Hu et al., 2012). Molecular evidence shows that, in all instances till now, CMS is associated with abnormal open reading frames (ORFs) located in mitochondrial genomes (Chase, 2007; Chen and Liu, 2014). These ORFs can be processed by specific nuclear genes (Rfs), which rescue the male sterile phenotype (Kim and Zhang, 2018). Therefore, CMS and the corresponding Rf(s) build an ideal model system for understanding the mechanism of co‐evolution and cross‐talk between the nucleus and mitochondria in plants. The CMS/Rf systems have been widely employed for hybrid seed production in many crops, such as maize, rice, rapeseed, sorghum, radish, and Chinese cabbage (Kim and Zhang, 2018).

Recently, a bunch of Rf genes have been cloned and identified in various crops and other plants (Kim and Zhang, 2018). Rf genes annihilate the detrimental effects associated with CMS at various levels. At the genomic level, Fr gene restores fertility of CMS‐Sprite common bean through altering the 25‐kb PVS (Phaseolus vulgaris sterility)‐associated mitochondrial genomic sequence (Janska et al., 1998). At the post‐transcriptional level, the CMS‐associated transcripts are processed by the restorer either through RNA editing and degrading (Gagliardi, 1999; Qin et al., 2021; Tang et al., 1999), or through cleaving and degrading (Jiang et al., 2022; Kazama et al., 2008; Kennell and Pring, 1989; L'Homme et al., 1997; Menassa et al., 1999; Wang et al., 2006) or through direct degrading (Luo et al., 2013). Rf genes also function by translational or post‐translational mechanism to suppress CMS (Dewey et al., 1991; Luo et al., 2013; Sarria et al., 1998; Uyttewaal et al., 2008; Wang et al., 2021). The Rf2 gene in maize CMS‐T may functions at the metabolic level because it encodes a detoxification enzyme and has no effects on urf13‐orf221 transcripts and URF13 protein (Liu et al., 2001). In previous studies, more than half of the cloned Rf genes belong to the pentatricopeptide repeat (PPR) family, most of which perform their best at the post‐transcriptional or translation/post‐translational levels (Chen and Liu, 2014; Kim and Zhang, 2018).

The PPR family has been greatly expanded with more than 400 members in different plant genomes. They are subdivided into two major subfamilies (P and PLS) according to the structures of the three basic PPR motif (P, L and S; Barkan and Small, 2014). Each basic motif of PPR gene is predicted to recognize one nucleotide, which confers sequence‐specific RNA binding ability on PPR proteins (Prikryl et al., 2011; Yan et al., 2019). Most plant PPR proteins identified so far are primarily targeted to either plastids or mitochondria, where these proteins were generally involved in various post‐transcriptional RNA regulation (Barkan and Small, 2014; Dahan and Mireau, 2013; Schmitzlinneweber and Small, 2008). Mutations in PPR genes lead to metabolic disorders of particular organellar RNAs, exhibiting a number of plant developmental flaws, such as slow or aborted embryo and endosperm (Cai et al., 2017; Sosso et al., 2012; Yang et al., 2019), retarded leaf, postponed flowering (Zhu et al., 2012), abnormal organelle biogenesis (Chateigner‐Boutin et al., 2008) and defected kernel (Cai et al., 2017; Qi et al., 2019). Nearly all RF‐related PPRs are members of the P subclass except for PPR13 in sorghum, however, in most cases, lack or mutation in these PPR genes have no influences on growth in normal cytoplasm. Therefore, it is considered as a gain‐of‐function mutations for Rf‐related genes (Gabay‐Laughnan et al., 2018). P‐family RF proteins are usually devoid of RNA binding sites, thus they often recruit additional components to form a restoration complex (Gillman et al., 2007; Hu et al., 2012; Huang et al., 2015).

RF‐like PPRs or P‐class PPRs have multiple role in RNA metabolism in mitochondria. Some P‐class PPRs are involved in 5′ processing of mitochondria RNA, which brings 5′ end polymorphisms of mitochondrial transcripts. RNA processing factor 1 (RPF1), RPF2, RPF3, and RPF5 are required for processing 5′ end of nad4, cox3/nad9, ccmC, nad6/rrn26/atp9, respectively, and these RPFs are RF‐like PPR proteins (Binder et al., 2013). In maize, P‐class PPR MPPR6 assists 5′ end maturation of the rps3 gene and facilitates its translation initiation (Manavski et al., 2012). Cleaving is a common economical strategy that many restorer gene adopt to eliminate deleterious transcripts of sterility gene. How much transcripts to be cleaved, where to cleave, directly or indirectly, or alone or with other partners depend on the inherent property of both the restorer gene and the sterility genes. P‐class PPR Rfs so far restore fertility by cleaving CMS transcripts at the middle or 3′ region (Jiang et al., 2022; Qin et al., 2021; Tang et al., 2014), or at the inter‐genic region of transcribed dicistronic RNA (Huang et al., 2015; Kazama et al., 2008; Wang et al., 2006).

CMS‐C is one of the three major groups of CMS in maize in the world and has been introduced in maize hybrid seed production for its fertility stability and resistance to epidemic pathogens (Hu et al., 1999). CMS‐C belongs to a sporophytic CMS group with pollen abortion at the mid‐to‐late mononucleate stage. The mitochondrial gene atp6c is associated with CMS‐C through interfering with complex V in the respiratory chain (Yang et al., 2022a). Transcriptome analysis shows that the energy metabolism‐related pathways have been greatly affected in the sterile lines (Xue et al., 2019). Pollen abortion can be rescued by two duplication fertility restorer genes, ZmRf4 and ZmRf5, and in the present study, we reported the cloning of the dominant CMS‐C restorer gene, ZmRf5, which encodes a typical P‐type PPR protein. The findings demonstrated that ZmRF5 was directed towards mitochondria and, by interacting with MORF8 (Multiple organellar RNA editing factor 8) and RS31A (Ser/Arg‐rich splicing factor 31a) to create the cleaving/restoring complex, enabled the cleavage of too much transcripts of atp6c. Additionally, it was demonstrated that a deficit of ATP6C also caused male sterility. Our results perfected the sterility mechanism of CMS‐C maize and suggested a new mechanism of restoration for PPR‐like restorers.

Results

Map‐based cloning of the restorer gene ZmRf5

To identify the location of the CMS‐C restorer gene, ZmRf5, 10 sterile plants and 10 fertile plants from the BC₁F₁ population derived from CMo17A × (CMo17A × C6233R) were used for gene mapping. The markers phi328189, bnlg1267, umc1536, bnlg1721, umc1126, bnlg1940, bnlg1520, umc1736, and bnlg1893 were polymorphic between the two parents and two bulks; all of these markers were positioned on the chromosomal bin2.09. DNA from additional 1364 individuals (fertile:sterile = 706:658, χ² = 1.68 < 3.84 (${\mathrm{\chi }}_{0.05}^2$)) of the backcross population was interrogated with these SSR markers and the ZmRf5 gene was mapped to a 5.1 Mb region between umc1736 and bnlg1520. The population was tested with additional seven polymorphic SSR markers, and the locus was delimited to a 1.6 Mb region between SSR64 and SSR285 (Figure 1a). To further narrow down the localization of ZmRf5, the two nearest flanking linkage markers were used to screen a BC₁F₁ population with 32 000 individuals for recombinants. DNA from 92 recombinants was evaluated with 7 new polymorphic markers within the 1.6 Mb region, and the ZmRf5 locus was delimited to a 382 kb region (Figure 1a). In this region, there are 6 annotated genes based on the maize reference genome (B73 version 4), and one of these genes belonged to the PPR family—PPR814 (Zm00001d007531). To obtain the genomic sequences of these genes, gene‐specific primers were used to amplify the corresponding sequences of the six genes in CMo17A and C6233R. No sequence differences were identified between the two lines except for Zm00001d007531, in which the gene in CMo17A had a 23‐bp and a 111‐bp deletions in the coding sequence compared to it in C6233R (Figure S1a). Zm00001d007531 in C6233R encodes a canonical P‐type PPR protein containing 19 PPR motifs (Figure 1b) and the 23‐bp deletion introduced a stop codon leading to a premature protein with 37 amino acids (Figure 1b; Figure S1b).

Figure 1.

Functional and expression analysis of the ZmRf5 gene. (a) Schematic representation of map‐based positional cloning of the ZmRf5 gene on chromosome 2. The ZmRf5 locus was initially mapped to chromosome 2 between the SSR markers, SSR64 and SSR285, and subsequently narrowed down to a 382‐kb interval between the markers SSR36 and Indel61. (b) Illustration of the protein domain of the ZmRf5 gene. MTS, mitochondrial targeting sequence; P, P‐class PPR motif. (c) Functional complementation of the restorer gene in the C‐type cytoplasm. c1‐c4, tassels, bars = 5 cm; c5‐c8, pollen grains stained with I₂‐KI, bars = 200 μm; c9‐c12, spikelets, bars = 1 mm. C01, transgenic receptor inbred line ZZC01; OE, the ZmRf5 transgenic over‐expression plants. pZmRf5::ZmRf5, the ZmRf5 transgenic plants driven by the ZmRf5 promoter. (d) Segregation analysis of the F₂ progeny from test‐crossing the C01 line with the ZmRf5 transgenic line (χ² < 3.84, P = 0.05 for 3:1). (e) Gene editing of ZmRf5 by CRISPR/Cas9 resulting in male sterility in the restorer line. e1‐e3, tassels, bars = 5 cm; e4‐e6, pollen grains stained with I₂‐KI, bars = 200 μm; e7‐e9, spikelets, bars = 1 mm. C01, transgenic receptor inbred line ZZC01; Crispr, transgenic maize lines containing the CRISPR/Cas9 system targeting to the ZmRf5 gene. (f) Semi‐quantitative RT‐PCR analysis of ZmRf5 expression levels in different maize tissues in C6233R. GAPDH was used as an internal control. (g) Comparison of ZmRf5 expression levels between C6233R and CMo17A by real‐time quantitative PCR. S6, development of microspore mother cells surrounded by four‐layered anther walls; S7, meiocytes start meiotic division; S8b, after meiosis II, tetrads enclosed by the callose wall are formed; S9, early mononucleate microspore; GAPDH was used as an internal control. (h) Subcellular location analysis of ZmRF5. eGFP, enhanced green fluorescent protein; Mito Tracker, a mitochondrial‐specific permeable dye (MitoTracker Red). Scale bars, 10 μm.

The CDS sequence of Zm00001d007531 from C6233R was cloned into the pCUB vector driven by the ubiquitin promoter (OE) or the ZmRf5's own promoter (pZmRf5::ZmRf5), and introduced into the inbred line ZZC01. The test‐cross of ZZC01 with CMo17A showed that no male fertility were observed, indicating no restorer genes (ZmRf4, ZmRf5, or other partial restorers) existed in the transgenic receptor line (Figure 1c). The positive transgenic lines from 20 transgenic events of each vector were crossed to the CMS line CMo17A for complementation test. The positive F₁ plants from nine transgenic events of OE and from seven events of pZmRf5::ZmRf5 showed normal fertile phenotype (Figure 1c). The fertile plants from three independent F₁ test‐crosses were selfed, and a 3:1 genetic segregation ratio was observed in the progeny of each selfing (Figure 1d). These results indicated that both pZmRf5::ZmRf5 and OE could restore fertility of the CMS lines. To further confirm the function of ZmRf5, CRISPR/Cas9 was used to generate loss‐of‐function lines within the restorer line C6233R. Four identified mutants, zmrf5‐1, zmrf5‐2, zmrf5‐3, and zmrf5‐4, that carried null mutations (Figure S2a) in the ZmRf5 gene showed male sterility as the CMS lines (Figure 1e). These sterile mutants were crossed to C6233R, and the heterozygous plants (+/−) showed full male fertility. The resultant F₂ plants from these heterozygous plants exhibited a fertility:sterility ratio of 3:1 (Figure S2b). These results further supported Zm00001d007531 was the restorer gene.

The expression pattern and subcellular localization of the ZmRf5 gene

ZmRf5 expression was quantified in the root, stem, leaf, and anthers at various developmental stages using RT‐PCR. The expression of ZmRf5 was detectable in all tested tissues, but at very high levels in the S6 anthers and the stems in the restorer lines (Figure 1f). qPCR analysis between C6233R and CMo17A showed that ZmRf5 was highly expressed at S6 and gradually decreased from S8a to S9 in C6233R. Whereas in CMo17A, the expression of the ZmRf5 gene was barely detectable (Figure 1g).

The restorer gene for CMS usually functions in mitochondria, and the domain prediction for ZmRF5 by SignalP (version 5.0) showed a mitochondria target signal at the N‐terminal of the protein. To test sub‐cell localization, the full‐length CDS of ZmRf5 was fused to the N‐terminal of the enhanced green fluorescent protein (eGFP) driven by the 35S promoter, and transiently expressed in maize leaf protoplasts. The fusion protein was observed in the mitochondria, which confirmed the location for the putative involved function for the restorer gene (Figure 1h).

ZmRF5 does not bind the atp6c transcripts, but promote multi‐site cleavages at the 5′ region

The interaction between over‐accumulated ATP6C and ATP8/ATP9 in mitochondria hampers the assembly of the F₁F_o‐ATP synthase and causes the sterility of CMS‐C maize (Yang et al., 2022a). The restorer genes usually cleave the aberrant transcript into unharmful short fragments or degrade the protein products of the aberrant transcript (Chen and Liu, 2014). We therefore asked how the sterility gene was affected under the presence of the restorer gene. Sequence identification revealed the sequence diversity at 5′‐region of atp6c transcripts in anthers from C6233R and CMo17, however, the major transcripts between two lines were different (Figure 2a,b). Transcripts with 5′ end at around the 344th nt were more accumulated under the presence of ZmRf5 in the restorer line, whereas transcripts with 5′ termini at around the 480th nt were enriched more without ZmRf5 in CMo17A. Next, we designed specific primers covered the unique and latter common region to examine the expression changes of the sterility gene (Figure 2c). The expression level of total transcripts (sum of most transcript species) was higher in C6233R than that in CMo17A (Figure 2d), while the expression level of the full‐length transcript was significantly lower in the restorer line (Figure 2e). The decrease in the long transcript of atp6c was likely caused by the cleaving function provided by the restorer gene. We thus predicted ZmRF5 binding sites on the atp6c transcript using the PPR code tool (Yan et al., 2019). Five possible binding sites selected from PPR Code and one manually selected probe sequence, which covered the two cleaving positions, were used to test the binding possibility between ZmRF5 and the atp6c transcript. No binding signal was detected in RNA‐EMSA assays for all six probes, and further testing of four binding sites by Y1H showed the same results (Figure S3a–e). The above experiments suggested that ZmRF5 does not bind, but cleaves the atp6c transcripts to decrease the accumulation of the full‐length transcripts. We noticed that the full‐length transcripts of atp6c were not completely cleaved, and atp6c is a replacement for atp6. Thus, the accumulation of ATP6C was interrogated with anti‐ATP6 antibodies. Western blot results showed that the accumulation of ATP6C in CMo17A was significantly lower than the accumulation of ATP6 in NMo17B and also lower than it in OE lines and restoration lines with homo or hetero ZmRf5 genes, that is, the restorer gene recovered the protein level of ATP6C in CMS‐C maize (Figure 2f,g; Figure S4a). In the sterile lines with different nuclear genome, the accumulation regarding sterile lines over maintainer lines showed the contrary trend, indicating the impact of nuclear genome on the expression of ATP6C (Figure 2g). NAD7 from Complex I and SDH1 from Complex II were also decreased in CMo17A concurrently with the decline of Complex V, but they were later recovered to normal levels of accumulation (Figure S4b). These results indicated that the restorer gene encouraged the cleaving of the full‐length atp6c transcripts by shifting the major cleavage site from 480th nt to 344th nt without direct interaction and facilitate the translation of atp6c to compensate the shortage of ATP6C.

Figure 2.

Cleavage analysis of atp6c transcripts mediated by ZmRF5. (a) Strategy for ligation‐mediated amplification of the 5′ ends of cleaved atp6c mRNA for analysis of the cleavage site. Solid line, mitochondrial RNA; dotted line, cDNA; orange box, 5′‐race adapter; green box, reverse transcription primer; and yellow boxes, PCR primers. (b) Agarose gel electrophoresis analysis of the 5′ RACE PCR products from C6233R and CMo17A. Adapters can be ligated to matured and cleaved mitochondrial RNA at 5′ terminus, except for nascent mitochondrial RNA. Red arrows, cleaved fragments of C6233R (251 bp) and CMo17A (113 bp). (c) The physical location of primers used in the qRT‐PCR experiment. (d, e) qRT‐PCR analysis of the expression of atp6c in anther during meiosis of C6233R and CMo17A using primer atp6c‐1F/R (514–908) and atp6c‐2F/R (26–236). Each experiment was replicated three times. Values are the means with SEs (*P < 0.05, **P < 0.01, student's t‐test). (f, g) The accumulation of ATP6C and ATP6 in the different lines, respectively. COX2 was used as an internal control. NMo17B, the maintainer line, N(zmrf4zmrf4zmrf5zmrf5); CMo17A, the sterile line, C(zmrf4zmrf4zmrf5zmrf5); C6233R, the restorer line, C(zmrf4zmrf4ZmRf5ZmRf5); CF₁, crossing CMo17A with C6233R, C(zmrf4zmrf4ZmRf5zmrf5); YU87‐1A, the sterile line, C(zmrf4zmrf4zmrf5zmrf5); YU87‐1B, the maintainer line, N(zmrf4zmrf4zmrf5zmrf5).

ZmRF5 cleaves the atp6c transcripts through interacting with a splicing factor and MORF8

No direct contact between ZmRF5 and atp6c means other partners may participate in the 5′ region processing of the sterility gene. Therefore, we designed a probe specifically matched with atp6c transcripts to pull down endogenous binding proteins on atp6c in CMo17A for searching ZmRF5 interaction partners. LC–MS analysis had identified 299 peptides and 148 proteins from the pulled mixture, and 35 proteins of them showed differential accumulations compared to the negative control (Table S1), which were potential interactors with ZmRF5 and atp6c. Annotation of the resulting protein candidates showed that protein A0A1D6J4J8 is a Ser/arg‐Rich (SR) splicing factor (RS31A) that is involved in RNA processing, and protein K7WGQ3 is a multiple organellar RNA editing factor 8 (MORF8). Previous researches have implied that the SRs and MORFs are essential for plant development (Jin, 2022; Yang et al., 2022b). Thus, we selected these two proteins for further study.

The electrophoretic mobility shift assay (EMSA) was used to validate the bindings of RS31A or MORF8 with the atp6c transcripts. It was shown that atp6c shifted with RS31A when they were incubated together (Figure 3a). While no bindings were detected for MORF8 and atp6c transcripts (Figure S5a,b). After validation, we asked whether ZmRF5 interacted with the two proteins. Three constructs, ZmRF5‐N, MORF8‐C, and RS31A‐C, were created and ZmRF5‐N together with MORF8‐C or RS31A‐C was co‐infiltrated into tobacco leaves. LUC activities detected by imaging method showed that luminescence signals were detected when ZmRF5‐N and MORF8‐C, or MORF8‐C and RS31A‐C, were developed in the same area, whereas no signals were emitted from areas harbouring controlled constructs (Figure 3b), which meant MORF8, as a linker, interacted both with RS31A and ZmRF5. The interaction between ZmRF5 and MORF8, or between MORF8 and RS1A, was further confirmed by the bimolecular fluorescence complementation (BiFC) assay (Figure 3c; Figure S6). ZmRF5 and MORF8 was also interacted in the Y2H system, whereas the interaction between MORF8 and RS31A could not be repeated in Y2H, probably because the Y2H is not suitable for these two genes, or the interaction is not strong enough to be detected by Y2H (Figure 3d). No interaction signals were recorded between RS31A and ZmRF5 (Figure 3b–d). Besides, RS31A also formed a dimer in the cleaving complex (Figure S5c–e). The sub‐cell localization of MORF8 and RS31A gained attention for their putative function. MORF8 was localized to mitochondria as expected (Figure S7a), whereas RS31A was targeted to nucleus, although in accordance with the presumed place in the cleaving process (Shen et al., 2004), but did not satisfy the desired spot in present case (Figure S7b). The expression of these two genes were higher in C6233R than in CMo17A (Figure S7c,d). We suspected RS31A might get a lift with another factor to enter mitochondria. Mitrotracker staining on tobacco leaves co‐infiltrated with RS31A and MORF8 also showed that the interaction took place in mitochondria (Figure 3c), which meant that RS31A was localized to mitochondria in the presence of MORF8. Taken together, these results provide persuasive evidence that MORF8 interacts with ZmRF5 or RS31A in mitochondria.

Figure 3.

Interactions among ZmRF5, MORF8 and RS31A. (a) REMSA of GST‐RS31A and GST using a fixed concentration of the atp6c RNA probe P4 (diagrammed in Figure S3a) with a range of protein concentrations and GST‐tag as a control. (b) BiLUC activity assays in tobacco leaves infiltrated with different combinations of effector and reporter constructs to detect interactions between ZmRF5/RS31A, ZmRF5/MORF8 and RS31A /MORF8. (c) BiFC assay tests of the interaction between ZmRF5, MORF8 and RS31A in tobacco leaves. YFP, yellow fluorescent protein. Bars = 20 μm. (d) Y2H assay tests of the interaction between ZmRF5, MORF8 and RS31A.

MORF8 often participates in RNA editing in mitochondrial editosomes (Yang et al., 2022b). The side effect of the interaction of MORF8 with ZmRF5 may have some impacts on the editing levels of mitochondrial RNA, especially on atp6c transcripts. Thus, the atp6c RNA was interrogated for the editing status of the eighteen edited sites (Yang et al., 2022a). The cDNA sequencing results showed that the editing on four sites, 572nd nt, 779th nt, 1010th nt, and 1232nd nt, were suppressed by one third for the former two, and a half for the latters, of the total transcripts (Figure S8a). Meanwhile, the four unedited (supressed) sites did not overlap with the cleavage sites (Figure S8b), which likely indicated that the editing and the cleaving were the two independent processes.

ZmRF5 recovers complex V, normal PCD, and physiological indexes

Previous study showed that the over‐accumulation of ATP6C plus interaction with other factors affects the assembly process of complex V (Yang et al., 2022a). The cleavage of the full‐length atp6c transcript would relieve the stress and restore the native state of the related complex. Thus, blue native (BN) PAGE was applied to evaluate the recovery of complex V after the cleavage. The complex V abundance of the restorer line in Coomassie blue staining gel was recovered compared with it in the CMS‐C lines (Figure 4a). And activity of complex V, detected through in‐gel activity assay, also returned to normal (Figure 4b). Western blot of second‐dimension SDS‐PAGE showed that ATP6 and ATPβ at the position of complex V accumulated more under the presence of ZmRF5 (Figure 4c).

Figure 4.

ZmRF5 recovers complex V assembly, PCD, and physiological indexes. (a) Accumulation (left) and ATP hydrolyse activity (right) of complex V on equal loaded total mitochondrial proteins. CV, ATP synthase; F₁, the F₁ component of ATP synthase; ${\mathrm{F}}_1^{\prime }$, transition complex of F₁. (b) Accumulation of complex V by western blot using anti‐ATP6 (left) and anti‐ATPβ (right) antibodies. (c) Immunodetection analysis with antibodies against ATP6 (up) and ATPβ (bottom) after second‐dimension SDS‐PAGE. (d) Detection of DNA fragmentation in anthers by TUNEL assays. The red fluorescence results from staining of anthers with propidium iodide (PI), green fluorescence indicates TUNEL‐positive nuclei of PCD cells. S6, development of microspore mother cells surrounded by four‐layered anther walls; S7, meiocytes at the start of meiotic division; S8, meiocytes at meiotic division; S9, early mononucleate microspore; S10, late mononucleate microspore; S11, binucleate microspore. Ta, tapetum. Scale bars, 50 μm. (e) ATP content (left) and ATPase activity (right) among C6233R, CMo17A, and CF₁ (ZmRf5zmrf5). (f–i) Measurements of physiological indicators among C6233R, CMo17A, and CF₁ (ZmRf5zmrf5): H₂O₂ content (f), POD activity (g), MDA content (h), and CAT activity (i). Values are the means and SEs. Each experiment was replicated three times.

PCD in tapetal cells starts early at S6, and continues till S10 in the sterile lines (Yang et al., 2022a), and the PCD process was similar in the sterile lines with the different nuclear background (Figure 4d). In the restorer lines, the ZmRf5 gene had restored the complex V assembly and activity, it likely also brought PCD back to normal. The tunnel assay showed that PCD under the presence of ZmRF5 started and ended normally (Figure 4d), which is likely ensured by the compensated accumulation of complex V and the resultant sufficient ATP level (Figure 4e, left), although the activity of complex V per unit was lower in the restorer lines (Figure 4e, right). The fertility of the sterile lines was restored in the restorer lines and the OE lines, but the H₂O₂ content, instead, was higher than it in the sterile and CRISPR/Cas9 knockout lines (Figure 4f). The high H₂O₂ level is correlated with the opposite low level of peroxidase in the two kinds of restored lines (Figure 4g), which caused more stress in the tapetal cells of the restored lines (Figure 4h). However, the stress was relieved by high activities of catalase (Figure 4i) to ensure the functionality of the tapetal cells because the fertility was back to normal under the presence of the restorer gene.

Discussion

The analysis of CMS and its restoration in various plant species have proposed four‐level restoration mechanism with most Rf genes functioned at the RNA and protein level (Chen and Liu, 2014). The cognate CMS transcripts were cleaved by RF proteins or Rf complex, leading to decreased accumulation of CMS transcripts and the corresponding proteins and hence to fertility restoration (Kennell and Pring, 1989; Luo et al., 2013; Menassa et al., 1999; Wang et al., 2006; Xiao et al., 2006; Yi et al., 2002). In the cases that the transcript amount and size do not alter in fertility‐restored plants, the buildup of CMS proteins was suppressed either by translation blocking (Uyttewaal et al., 2008; Wang et al., 2021) or by other degradation manners (Dewey et al., 1991; Itabashi et al., 2011; Sarria et al., 1998). In the present study, the two newfound results have refreshed our knowledge about the CMS/Rf system. The first one is that the CMS gene atp6c in the Mo17 background employed quite a different manner to cause pollen abortion by less accumulation of CMS proteins, contrary to it in the Yu87‐1 background (Figure 2; Yang et al., 2022a). This indicates that in CMS‐C maize, the nuclear genome will affect the sterility‐inducing way, which is not present in all other CMS types (Kim and Zhang, 2018). Another is that in the restorer lines, many but not all full‐length atp6c transcripts are cleaved at the 5′ region to lower the superfluously accumulated transcripts, and the major cleavage site of the transcripts is shifted from 480th nt to 344th nt likely for easy degradation and less interfering with translation, which recovers the protein level of ATP6C (Figure 2). These results have led us to propose the following CMS/Rf model that the shortage of ATP6C could cause defects in pollen development, and ZmRF5 performed the cleavage shifting to decrease amount of atp6c transcripts through recruiting a splicing factor RS31A and MORF8 to form a restoring/cleaving complex, which recovered the normal assembly of complex V, thereby, brought the aberrant PCD back to normal and rescued the fertility of the sterile lines (Figure 5). This molecular mechanism is somewhat similar to the mechanisms of other restorer genes with cutting function. However, there are two prominent different points: first, the quantity of the full‐length atp6c transcripts has to be elaborately controlled. It cannot be fully degraded or be decreased to a very low amount like other sterility genes; second, several major cutting sites are all located in the 5′ region of the atp6c transcript, mimicking 5′ region maturation consequence of mitochondrial RNA. The reason for these special features in sterility‐inducing and restoring actions is likely determined by the inherent property of the sterility gene in CMS‐C maize (Yang et al., 2022a), and is also a result of co‐evolution of the two genes. Two strategies are adopted by new mutated or evolved genes to gain new molecular or cell function, one is to integrate into the existing gene network, and the other is to rebuild a new interaction network (Chen et al., 2013; Zhang et al., 2015). Most studied Rf/CMS systems choose the second one because the sterility genes in these systems are independent of all other conserved functional mitochondrial genes, whereas the ZmRf5/atp6c system we presented here fulfils its restoring function by integrating into the old networks because of the replacement of atp6c for atp6 (Yang et al., 2022a), which perfected the previously proposed sterility mechanism and unfolded a new mechanism of restoration for plant CMS.

Figure 5.

Schematic representation of proposed restoration mechanism for ZmRF5. The shortage of ATP6C could cause defects in pollen development, and ZmRF5 performed the cleavage through recruiting a splicing factor RS31A and MORF8 to form a restoring complex, which recovered the normal assembly of complex V, thereby, brought the aberrant PCD back to normal and rescued the fertility of the sterile lines.

P‐type PPR proteins are often involved in RNA splicing of mitochondrial group II introns, 5′ end processing, and translation (Manavski et al., 2012; Wang et al., 2021; Yang et al., 2022c; Zhang et al., 2017). The restorer gene ZmRf5 identified in this study encodes a canonical P‐type PPR protein without direct contacting with atp6c (Figure S3). Thus, additional players such as RS31A and MORF8 were required for performing cleavage and were included with preliminary evidences in the proposed restoration model (Figure 5). MORFs were usually involved in abundant editing events in mitochondria or plastids alone or by interacting with each other or other PPR factors and mutations in MORF proteins reduce RNA editing efficiency in many target sites (Yan et al., 2018; Zhao et al., 2019). The present results showed that MORF8 participate in the cleaving at the 5′ region of the atp6c transcripts, which suggests MORFs may have novel role in cleaving, restoration complex, or 5′ region processing complex besides in editosome. To our knowledge, this novel location for MORFs has not recorded so far by other reports, but the related studies have indicated the similar acting manner that MORF proteins can act as “YueLao/月老”—the god who unites persons in marriage, bringing PPR proteins or other factors together (Gualberto and Newton, 2017; Yang et al., 2022b). MORF8 was identified by RNA pull‐down assay in the sterile lines, in which the atp6c transcripts are fully edited (Yang et al., 2022a). However, the editing in the four sites of atp6c are inhibited in partial transcripts in the restorer lines due to the interaction with ZmRF5 (Figure S8a). Thus, the editing status in the restorer line and the concomitant consequence on restoration are worthy of further investigations. In this case, MORF8 gathered ZmRF5 and RS31A to assist ZmRF5 to fulfil the restoration function.

RS31A is a Ser/Arg‐rich protein, which usually forms the spliceosome by interacting with other proteins and contacting the 5′ splicing site of RNA (Jin, 2022; Shen et al., 2004; Zhang et al., 2020). As a lifelong RNA chaperone accompanying from transcription, translation to degradation, many RS proteins usually localized to nuclear speckle, nucleoplasm, nucleolus, and cytoplasm (Zhang et al., 2020), by which RS31A abided when sub‐localization was tested alone (Figure S7a). Interestingly, some RS31A were also localized to mitochondria when it was co‐expressed with MORF8 (Figure 3c), which is in accordance with its expected roles in mitochondria. In restoration complex, many factors lack mitochondrial signal peptides, but were imported to mitochondria by interaction with other mitochondria‐targeted proteins as GRP162 in HL‐CMS rice does (Chen and Liu, 2014; Hu et al., 2012). It is reasonable to introduce an extra splicing factor into mitochondria to provide the required cleaving function in the restoring complex, because the cleaving activity in PPR proteins is usually too high for partial cleavage of atp6c (Kim and Zhang, 2018; Qin et al., 2021). As a P‐type PPR restorer, ZmRF5 may have the endonucleolytic activity, however, the function of cleaving is also likely provided by RS31A or other endonucleases, more experiments should be done to test enzymatic activities of ZmRF5 and RS31A or to find the endonuclease and their role in restoration and 5′ region cleaving or maturation in the future work.

The application of CMS in seed production is determined by the stable sterility phenotype and the presence of a reliable restorer gene. In most widely applied CMS, the corresponding Rf genes usually cleave the sterility‐induced genes (Chen and Liu, 2014; Kim and Zhang, 2018), which ensures zero chance of reversion to the original sterility gene and complete fertility restoration. The ZmRf5 gene can fully restore the fertility of maize C‐type CMS in all tested situations, the only factor that lags the application of ZmRf5 is the fewer germplasm resources in natural population and breeding inbred lines, especially in paternal lines, compared with another restorer ZmRf4 for CMS‐C maize (Jaqueth et al., 2019). However, the cloning and elucidating restoration mechanism of this gene in the present study facilitate the gene screening and introgression among different lines, which will boost the confidence and accelerate the application of CMS‐C in maize. Besides, co‐introgression of ZmRf5 with ZmRf4 into the same maize line, used as mutual backup of fertility recovery, could safeguard the fertility of hybrid plants under various extreme stress conditions.

Experimental procedures

Plant materials and growth conditions

CMo17A is a CMS inbred line containing C cytoplasm of the CI subgroup and obtained by successive backcrossing using Mo17 (NMo17B, C denotes a C cytoplasm, N indicates normal cytoplasm and B means the maintainer line) as the male donor for more than six generations. ZmRf4, ZmRf5, and other weak restorer alleles are absent from the two lines. The restorer line C6233R (R means the restorer line) was developed by marker selection for non‐Rf4 plants from the F₂ population obtained after crossing CMo17A with Fengke1 (ZmRf4ZmRf4ZmRf5ZmRf5). The genotype of C6233R is identified through allelic test in a F₂ population derived from the cross between C6233R and A619 (Hu et al., 2006). C6233R was crossed to CMo17A to produce F₁ generation plants. The F₁ plants were backcrossed to NMo17B to generate BC₁F₁ seeds for fine mapping of the ZmRf5 locus. ZZC01 is a transformation receptor line developed by co‐cultivation at the Life Science and Technology Center of China National Seed Group Co. Ltd, Wuhan, China. All non‐transgenic plants were grown in the standard growing conditions in Zhengzhou Farm (113.8E, 34.8N), Henan, or in Jiusuo Farm (108.9E, 18.4N), Ledong, Hainan. All transgenic lines were planted in the special isolation locations in Hainan or Zhengzhou following standard operation protocols for transgenic plants. Anthers from the S6 to S9 developmental stages were collected from different tassels and immediately frozen in liquid nitrogen for future use. For pollen staining, anthers were incubated with I₂‐KI solution to judge fertile and sterile grains under a microscope (Motic, China).

Mapping of ZmRf5

The degree of fertility was evaluated according to six‐grade anther exsertion and degree of anther dehiscence (Kheyr‐Pour et al., 1981; Li et al., 1963). In the BC₁F₁ population, plants with ZmRf5zmrf5 and zmrf5zmrf5 genotypes produce approximately 100% and 0% fertile pollen, respectively, and the ratio between the two genotypes is 1:1. The fertility of 1500 (small population) and 32 000 (enlarged population) individual plants, respectively, was evaluated. The rough location of ZmRf5 was first detected by bulk‐segregation analysis and quantitative trait loci mapping using the small population. The linkage markers, umc1736 and bnlg1520, from the primary mapping, were used to screen for the recombinants in the enlarged population. The markers for fine‐mapping were developed based on the genome sequences of B73‐ref and Mo17. The sequences of the primers used for mapping are listed in Table S2.

Transgene constructs

For testing the restoration function, the full‐length ZmRf5 coding sequences from a C6233R inbred line were cloned into pCUB vector between the Ubiquitin1 promoter and the nopaline synthase terminator or between the ZmRf5 promoter and the nopaline synthase terminator to generate the constructs UBI1::ZmRf5 and pZmRf5::ZmRf5, respectively. To knockout ZmRf5, 20 nt upstream of the NGG sites were picked as target sequences based on the design principles of the target sequences in the CRISPR/Cas9 system. Double‐stranded target sequences were cloned into the pBUE411 vector linearized by Bbe I. Transformation of the constructs into the calli of ZZC01 by Agrobacterium‐mediated transformation, and cultivation of regenerated plantlets were performed at the Life Science and Technology Center of China National Seed Group (Wuhan, China). The expressions of these introduced genes in the subsequent progeny were confirmed by herbicide screening (Basta, NO.A614229, Sangon, China) and PCR detection. Two target sequences in the PPR motif and primer sequences for bar detection are listed in Table S2.

RNA extraction and qRT‐PCR

Total RNA isolation was extracted using FastPure® Plant Total RNA Isolation Kit (Vazyme Biotech Co., Ltd, Nanjing, China). For qRT‐PCR, the first‐strand cDNA synthesis was performed using All‐in‐One Script RTpremix (with ds DNase; Kermey Biotech, Zhengzhou, China), 2 × SYBR Green qPCR Premix (Universal; Kermey Biotech, Zhengzhou, China) was used to amplification according to the manufacturer's instructions. Each experiment was replicated three times. The primer sequences used in PCR analysis are listed in Table S2.

Subcellular localization

The stop codon‐deleted CDS of ZmRf5 was fused to the N‐terminus of enhanced GFP (eGFP) reporter gene and cloned into pPRTL2 vector (35S::ZmRf5‐eGFP) using the Hieff Clone® Universal One Step Cloning Kit (Cat No.10922; Yeasen, Shanghai, China). The required E. coli strain DH5α competent cells of this experiment were all prepared by the Ultra‐Competent Cell Preps Kit (Sangon Biotech, Shanghai, China). The construction and the pPRTL2 vector were transformed into maize mesophyll protoplasts using a polyethylene glycol‐mediated transient transformation system. The expression of empty pPRTL2 vector in protoplasts was used as blank eGFP control. eGFP signal was observed and imaged by a confocal laser scanning microscope (LSM710; Zeiss, Oberkochen, Germany). MitoTracker® Red CMXRos (Invitrogen, California, USA) was used as organelle markers for mitochondria.

Protein expression and RNA EMSA

The ZmRf5 sequences and terminator codon were cloned into the protein expression vector pMal‐c5x with MBP label right after ZmRf5. The construct was transformed into the E. coli strain Rosetta. Protein expression was induced by addition of 0.1 mm IPTG at 16 °C for 18 h. For RS31A, a full‐length cDNA was joined to the protein expression vector pGEX‐4T‐1 with GST label and transformed into the E. coli strain Rosetta with 0.3 mm IPTG at 28 °C for 16 h. For MORF8, a MORF ORRM1 like domain was cloned into the vector pMal‐c5x and transformed into the E. coli strain Rosetta with 1 mm IPTG at 16 °C for 16 h. The fused protein ZmRF5‐MBP and MORF8‐MBP were extracted and purified using MBPSep Dextrin Agarose Resin 6FF following the manufacture's directions, whereas RS31A‐GST was purified with GSTSep Glutathione Agarose Resin. RNA probes (Table S2) specific to atp6c were synthesized in Generalbiol (Chuzhou, China) and labelled with biotin at the 3′ end. RNA probes were dissolved in DEPC H₂O to a final concentration of 10 μM. Serial dilutions of purified proteins were incubated with biotin‐labelled RNA probes in a 20 μL mixture consisting of 10 mm HEPES pH 7.3, 20 mm KCl, 1 mm MgCl₂, 1 mm DTT, 5% glycerol, 2 μg tRNA, 0.5 μM Biotin–IRE Control RNA, 4 μg purified proteins at 25 °C for 30 min. The mixture was separated in 0.5× TBE buffer by 6% native PAGE and then transferred to a nylon membrane (Roche, Basel, Switzerland). The signals were detected using the Chemiluminescent Imaging System (Tanon 5200, Shanghai, China).

Immunoblotting

Mitochondrial protein extraction and Western blotting were done as previously reported (Yang et al., 2022a). The abundance of ATP6 and ATPβ proteins was detected using anti‐ATP6 (Biogot technology, China)/anti‐ATPβ (AS05085; Agrisera, Uppsala, Sweden) antiserum diluted to 1:1000. COXII (1:2000, AS04053A; Agrisera, Uppsala, Sweden) was used as protein loading control. Anti‐ATP6 antibody can recognize both ATP6 and ATP6C proteins because the immunogen sequence used in antibody generation is in a shared region of the two proteins. Other mitochondrial antibodies we used included anti‐Nad7 (1:2000, PHY05138S; PHYTOAB, California, USA), anti‐SDH1 (1:2000, PHY0558S; PHYTOAB, California, USA), anti‐CYC1 (1:2000, PHY0566S; PHYTOAB, California, USA).

5′‐RACE

The 5′ RACE assay was performed using FirstChoice® RLM‐RACE Kit (AM17500; Thermo, Waltham, USA) following the instruction of the manufacturer. Mitochondrion RNA isolation was extracted from the anthers of C6233R and CMo17A were used as template. Adapter ligation was carried out in an RNase‐free microfuge tube consisting of 10 μL mixture (1 μg mtRNA, 1 μL 5′ RACE Adapter, 1 μL 10× RNA Ligase Buffer, 5 U T4 RNA Ligase) at 37 °C for 1 h. First‐strand cDNA was synthesized using M‐MLV reverse transcriptase with random decamer primers, according to the manufacturer's instructions. The resulting cDNA was amplified using the 5′ RACE inner primer and gene‐specific primer 595R. The PCR products were separated in a 2% agarose gel, stained with goldview and visualized under UV light. All of the amplicons were cloned into Zero Background pTOPO Simple Cloning Kit (LANY, Beijing, China) and sequenced by ABI PRISM 3700 Genetic Analyzer (Sangon Biotech, Shanghai, China). The primers used are listed in Table S2.

Yeast two‐hybrid assay

The CDS of ZmRf5, Morf8, and Rs31a, with/without MTS, was cloned into the pGBKT7 and pGAD‐T7 vectors. The bait and prey constructs were co‐transferred into yeast strain Y2HGold. The transformed yeast cells were inoculated on the synthetic defined (SD) medium without leucine and tryptophan (SD/‐Leu/‐Trp) and incubated at 30 °C for 3 days. The cells were then diluted and inoculated on the defective SD medium (SD/‐Leu/‐Trp/‐His/‐Ade) and the SD/‐Leu/‐Trp medium (as a loading control) and cultured for 2–3 days at 30 °C to observe the growth of the strain.

BiFC assay

The full‐length cDNA without stop codons of ZmRf5, Morf8, and Rs31a were cloned into the vectors pCAMBIA1300S‐nYFP and pCAMBIA1300S‐cYFP, respectively. The BiFC assay was performed by penetrating A. tumefaciens strains GV3101 carrying the constructs into the leaves of 4‐week‐old N. benthamiana plants. After penetration, the plants were cultivated in the dark for 48–72 h, and the fluorescence signals were observed using a confocal microscope (LSM880; Zeiss). The primers used are listed in Table S2.

LCI assay

The putative interacting proteins ZmRF5, MORF8 and RS31A, were fused to either the N terminus (nLUC) and the C terminus of firefly LUC (cLUC), respectively. Two different vectors were co‐transfected into 4‐week‐old tobacco leaves epidermal cells by agroinfiltration using A. tumefaciens strain GV3101. After infiltration at 28 °C for 72 h, D‐luciferin (Promega, Wisconsin, USA) was injected into leaves at a final concentration of 1 mm. LUC signals were detected and imaged using the Chemiluminescent Imaging System (Tanon 5200, Shanghai, China). The primers used for the LCI assays are given in Table S2.

TUNEL

The TUNEL assay was performed using TUNEL BrightGreen Apoptosis Detection Kit (Vazyme Biotech Co., Ltd) according to the manufacturer's instructions. Images were taken using a fluorescent inverted microscope (Scope A1; Zeiss) with green fluorescence at 520 ± 20 nm under a standard fluorescence filtration device and red fluorescence of propidium iodide at 620 nm.

Y1H

The sequences of different fragments (Figure S4d) from atp6c were cloned into the pAbAi vector using homologous recombination, and integrated into the yeast strain Y1H Gold (Clontech, California, USA) to generate the yeast bait strains. The constructed recombinant vector AD‐ZmRF5 was transferred into the positively confirmed Y1H Gold and the strain was incubated on SD/‐Leu‐Ura medium at 30 °C for 3 days. The cells were diluted and inoculated onto SD/‐Leu‐Ura‐AbA medium, and cultured at 30 °C for 3–5 days to observe the growth of the strain. The primers used are listed in Table S2.

Quantification of ATP, enzyme activities, and ROS levels

ATP content in anthers at the S8b stage was measured with a luciferin‐luciferase ATP assay kit (S0027, Beyotime, Jiangsu, China). ATPase activity was measured in anthers at the S8b stage using CheKineTM Na+/K+‐ATPase Activity Colorimetric Assay Kit (KTB1800; Abbkine, Wuhan, China).

The hydrogen peroxide (H₂O₂) content of anthers at the S8b stage was measured using the CheKineTM Hydrogen Peroxide Assay Kit (KTB1041; Abbkine). Peroxidase (POD) activity in meiotic anthers at the S8b stage was measured using the CheKineTM Peroxidase Activity Colorimetric Assay Kit (KTB1150; Abbkine). The malondialdehyde (MDA) content of anthers at the S8b stage was measured using the CheKineTM Lipid Peroxidation Assay kit (KTB1050; Abbkine). Catalase (CAT) activity in anthers at the S8b stage was measured using the CheKineTM Catalase Activity Assay Kit (KTB1040; Abbkine). All experiments were repeated at least three times.

Blue native PAGE, mitochondrial complex V activity, and 2D‐SDS‐PAGE

Anther mitochondria from different stages were extracted using a Plant mitochondrial Extraction kit (BJBALB; BeiJing, China) and BN‐PAGE, 2D‐SDS‐PAGE, and mitochondrial complex V activity assays were performed based on the previously described method (Yang et al., 2022a).

RNA pull‐down assay

RNA probes labelled with biotin at 3′ end were synthesized by Generalbiol (Chuzhou, China). The pulldown assay was done using Magnetic RNA‐Protein Pull‐Down Kit (20164; Thermo) with minor modifications. Briefly, the Streptavidin Magnetic Beads were washed twice with a 2× volume of 0.1 M NaOH, 50 mm NaCl, and once with 100 mm NaCl, then re‐suspended in 1× Binding and Washing Buffer (5 mm Tris–HCl (pH 7.5), 0.5 mm EDTA, 1 M NaCl, 1 U/μL RNase Inhibitor). 50 pmol of labelled RNA was added to the beads and incubated for 2 h in room temperature. The supernatant was discarded using a magnetic stand, and the beads were washed with 20 mm Tris (pH 7.5) twice. 100 μL of 1× Protein‐RNA Binding Buffer (20 mm Tris (pH 7.5), 50 mm NaCl, 2 mm MgCl₂, 0.1% TweenTM‐20 Detergent) was added to the beads and mixed well. After discarding supernatant, 500 μL Master Mix of RNA‐Protein Binding Reaction including 50 μL 10× Protein‐RNA Binding Buffer, 150 μL 50% glycerol, 50 U RNase Inhibitor, 5 μL 100× Cocktail, 300 μg CMo17A anther mitochondrial protein were added to the beads and incubated at 4 °C overnight or even longer with agitation or rotation. Magnetic beads were collected on the magnetic stand to remove the supernatant and washed with 100ul of pre‐cooled 1× Wash buffer (20 mm Tris (pH7.4), 10 mm NaCl, 0.1% Tween‐20) twice. Finally, 20 μL protein loading buffer was added and proteins were heated at 100 °C for 10 min. The proteins captured by Streptavidin Magnetic Beads were subjected to LC–MS analysis.

Conflicts of interest

The authors declared that no competing interests exist.

Author contributions

XY, LZ and TJ designed and supervised the research; CH mapped the gene; LY, YH, LH, LX and LB performed the experiments; GY, LY, and LB analysed data; CY, LH, YH, and LB participated in field work and sample collection; HY and XY acquired the funding; XY, DD and TJ wrote the manuscript. All authors have read and agreed to the published version of the manuscript.

Funding

This work was supported by the National Natural Science Foundation of China (32272165 and 31571745).

Supporting information

Figure S1 Amino acid sequence alignment of ZmRF5 between C6233R and CMo17A.

Figure S2 Analysis in transgenic progeny generate by CRIPR/Cas9 in C6233R.

Figure S3 ZmRF5 does not directly bind the atp6c transcripts.

Figure S4 Immunoblot detection of mitochondrial proteins.

Figure S5 RS31A forms homo‐dimers.

Figure S6 Negative control of BIFC for MORF8 (a) and RS31A (b).

Figure S7 Subcellular localization and expression of MORF8 and RS31A.

Figure S8 Changes of atp6c RNA editing between C6233R and CMo17A.

Table S1 Protein list in CMo17A from RNA pull‐down analysis.

Table S2 The primers used in the experiment.