Induce male sterility by CRISPR/Cas9-mediated mitochondrial genome editing in tobacco

Genome editing has become more and more popular in animal and plant systems following the emergence of CRISPR/Cas9 technology. However, target sequence modification by CRISPR/Cas9 has not been reported in the plant mitochondrial genome, mtDNA. In plants, a type of male sterility known as cytoplasmic male sterility (CMS) has been associated with certain mitochondrial genes, but few genes have been confirmed by direct mitochondrial gene-targeted modifications. Here, the CMS-associated gene (mtatp9) in tobacco was cleaved using mitoCRISPR/Cas9 with a mitochondrial localization signal. The male-sterile mutant, with aborted stamens, exhibited only 70% of the mtDNA copy number of the wild type and exhibited an altered percentage of heteroplasmic mtatp9 alleles; otherwise, the seed setting rate of the mutant flowers was zero. Transcriptomic analyses showed that glycolysis, tricarboxylic acid cycle metabolism and the oxidative phosphorylation pathway, which are all related to aerobic respiration, were inhibited in stamens of the male-sterile gene-edited mutant. In addition, overexpression of the synonymous mutations dsmtatp9 could restore fertility to the male-sterile mutant. Our results strongly suggest that mutation of mtatp9 causes CMS and that mitoCRISPR/Cas9 can be used to modify the mitochondrial genome of plants.


Introduction
Mitochondria are double-membraned cellular organelles of bacterial origin that play fundamental roles in many cellular processes, including energy production, calcium homeostasis, cellular signaling, and apoptosis (Dyall et al. 2004). Mitochondria possess their own genome, mtDNA, which is maternally inherited and encodes partial subunits of respiratory chain complexes, rRNAs, and mitochondrial tRNAs. mtDNA is present in multiple copies per cell, ranging from approximately 1000 copies in somatic cells to several hundred thousands of copies in oocytes, and has a very high mutation rate (Shoubridge and Wai 2007). Plant mtDNA is usually within the size range 200-700 kb, being larger than those of animal cells (about 16.5 kb), and contains an abundance of repeat sequences of different sizes and number. Large repeats (>500 bp) can be involved in frequent homologous recombination, while small repeats (<50 bp) can promote illegitimate microhomology-mediated recombination (Gualberto and Newton 2017). Due to the fundamental role of mitochondria in energy production, damage to mtDNA, such as by point mutations or deletions, contributes to or predisposes individuals to a range of dysplasias in animals and plants, associated with degeneration of tissues and organs with high energy demands (Filosto et al. 2011;Li et al. 2014). In most situations, mtDNA mutations are frequently mixed at a low percentage with normal wild type mtDNAs within the cell, a state known as heteroplasmy (Duan et al. 2018;Grady et al. 2018;Jackson et al. 2020; Yanzi Chang and Baolong Liu contributed equally to this work. Pereira et al. 2021). During both mitotic and meiotic cell division, the percentage of heteroplasmic alleles can change, leading to a potentially continuous array of bioenergetic defects (Wallace 2013).
In order to both study mitochondrial functions and try to cure mtDNA-based dysfunctions, it is essential to be able to manipulate mtDNA directly, and this goal remains an important challenge. Recent advances with sequence-specific nucleases, such as TALENS and the CRISPR/Cas9 system, have made it possible to easily modify the nuclear genome (Bacman et al. 2014a;Reddy et al. 2015). TALENs designed to edit the mitochondrial genome (mito-TALENs) have been created by adding a mitochondrial localization signal to the N-terminus. The mito-TALENs can reduce the levels of specific mtDNA mutations in cultured mammalian cells in attempts to cure mitochondrial diseases (Bacman et al. 2013;Hashimoto et al. 2015). In plants, the double-strand breaks (DSB) induced by mito-TALENs are repaired by homologous recombination of large repeats, leading to knock out of CMS-associated genes orf79 and orf125 of cytoplasmic male-sterile varieties of rice and rapeseed, respectively, to reverse the male sterility (Kazama et al. 2019). Unlike the TALENs, which involve only the protein, the CRISPR/Cas9 system involves both RNA and protein. The mito-CRISPR/ Cas9 system has been used to reduce mtDNA copy number in both human cells and zebrafish and knock an exogenous, single-stranded DNA with a short homologous arm accurately into the target locus (Bian et al. 2019). These experiments demonstrate the potential of genome editing as a tool for studying mtDNA and for future therapy of mitochondrial disorders. However, prior to the current study, it was unknown as to whether the mito-CRISPR/Cas9 system would work in plants.
Mitochondrial ATP synthase is an enzyme complex which is used to convert ADP to ATP in the final step of oxidative phosphorylation (Jonckheere et al. 2012). ATP synthase subunit 9, encoded by the mtDNA-specific mtatp9 gene, has been speculated to be responsible for cytoplasmic male sterility (CMS) in many plants (Bégu et al. 1990;Salazar et al. 1991;Hernould et al. 1993;Laser et al. 1995;Zabaleta et al. 1996;Dieterich et al. 2003;Szklarczyk et al. 2014;Wang et al. 2022;Sun et al. 2023). However, this speculation still needed to be confirmed by direct mitochondrial gene-targeted modification. In the work described in the current paper, the CRISPR/Cas9 system was modified by adding a mitochondrial localization signal to modify the mtatp9 gene in tobacco mitochondria with the aim of inducing male sterility. The mtatp9 of mtDNA was cleaved by the mitoCRISPR/Cas9 system in tobacco. The decreased number of mtDNA copies and the percentage change of heteroplasmic mtatp9 alleles could cause energy supply imbalance and influence the normal development of stamens, resulting in male sterility. Our results provide a case study for the use of mitoCRISPR/Cas9 system to successfully modify the plant mitochondrial genome.

Male sterility caused by an optimized mito-CRISPR/ Cas9 system
To construct the mito-Cas9 vector, the Cas9 sequence, containing a mitochondrial localization signal at the N-terminus and an sgRNA expression cassette carrying the gRNA spacer sequence for targeting mtatp9 (Fig. 1a), was synthesized and ligated to the backbone of the pCambia1300 binary vector. The mitochondrial localization signal was the cytochrome C oxidase subunit IV (COX IV) presequence from Saccharomyces cerevisiae (Kohler et al. 1997), which was expected to guide transport of the Cas9 protein into the mitochondria of plant cells (Fig. 1b). The mito-Cas9 vector was introduced into the leaves of "Samsun" tobacco (Nicotiana tabacum) using Agrobacterium transformation, and 27 Cas9-positive plants was obtained (Supplemental Fig. 1). The observation of floral morphology found that 8 Cas9-positive plants produced mutant flowers. The mito-Cas9 protein was determined by western blotting on extracted mitochondria. The result showed that the mito-Cas9 protein could be transported into the mitochondria (Supplemental Fig. 2).
The mutant flowers could be divided into two types, based on the phenotype: the petaloid stamen type (PST), with petaloid stamens and abnormal anthers, and the abortive stamen type (AST), with shortened filaments and abnormal anther dehiscence (Fig. 1c). Compared with the wild type flowers, the pollen of PST and AST flowers exhibited lower quantity and zero viability as detected by staining with TCC and subsequent microscopic observation (Fig. 1d). Of the eight transgenic plants with mutant flowers, four exhibited normal and PST flowers, two exhibited normal and AST flowers, all six of which were chimeric mutants, whereas two exhibited only abortive type stamens. After pollination, the capsules of normal flowers of mutant plants developed as usual, and seed setting rate was normal, while PST and AST flowers did not develop capsules and seed setting rate was zero in each case.

Heteroplasmic allele shifts
Previous research had demonstrated that cleaved mtDNA is degraded quickly in vivo (Papeta et al. 2010;Peeva et al. 2018) and that the nucleases involved can be used for specific elimination of mtDNA and induction of heteroplasmic shift in mammalian cells (Bacman et al. 2013(Bacman et al. , 2014bReddy et al. 2015). To analyze the effect of the mito-CRISPR/ Cas9 system on stamens, we first detected the change in the relative copy number of mtDNA by the qPCR with the primers for mtatp1, mtatp9, and mtorfX of mtDNA ( Fig. 1a and Supplemental Tab. 1). The results showed that the mtDNA copy number in PST stamens was similar to that of wild type stamens, whereas the copy number in AST stamens decreased by about 24% compared with normal stamens. In addition, the copy number of mtatp9 was similar with that of mtatp1 and mtorfX in AST stamens. This result may hint that knocking out of large fragments induced by homologous recombination at the mtatp9 locus was not triggered (Fig. 2a).
In plant cells, mtDNAs consist of normal wild type mtD-NAs and a small amount of mutant mtDNAs. For calculating the percentage of heteroplasmic mtatp9 alleles in the stamens, the high-throughput tracking of mutations (Hi-TOM) was performed, and more than 40,000 reads of mtatp9 were obtained for each sample. Based on these sequencing results, a mess of mtatp9 alleles was detected between wild type and mutant stamens. There were not novel insertions or deletions variants at the target site of the mtatp9 in either PST or AST stamens (Supplemental the excel of Hi-TOM analysis of mtatp9). In the wild type stamens, the proportion of the typical mtatp9 allele encoding the functional mtATP9 subunit was 80.75%; while in PST and AST stamens, it had decreased to 54.57% and 42.17% respectively (Fig. 2b). The other atypical mtatp9 alleles, which were present at very lower proportions in wild type stamens, showed increasing proportion in PST and AST stamens, whereas the number of mtatp9 alleles with a percentage greater than 0.1% also increased in mutant stamens (Fig. 2c). In wild type stamens, Fig. 1 Induction of male sterility by mitochondrial genome editing of tobacco plants with the mito-CRISPR/Cas9 system. a Schematic diagram of tobacco mitochondrial DNA (mtDNA). Blue triangle indicates site targeted by guide RNAs (gRNAs) against mtatp9, and yellow arrows indicate ATP synthase subunits encoded by mtDNA. b Schematic diagram of mito-CRISPR/Cas9 system. MtLP, COX IV presequence; PCas9, Cas9 gene without nuclear localization sequence; AtU3, small nuclear RNA (snRNA) gene promoter from Arabidopsis thaliana; gRNA, guide RNA; T, target sequence. c Representative flowers and stamens of male-sterile mutant. Red arrows indicate anthers; bar = 5 mm. d Pollen of the male-sterile mutant following staining with 2, 3, 5-triphenyl tetrazolium chloride. Bar = 20 μm. Stained pollen grains are viable, and colorless pollen grains are non-viable. WT, wild type; PST, petaloid stamen type mutant; AST, abortive stamen type mutant the number of atypical mtatp9 alleles accounting for over 1% was only five. However, in PST and AST stamens, this number increased to eight and seventeen respectively (Supplemental the excel of Hi-TOM analysis of mtatp9). This was particularly the case with respect to the mtatp9 alleles containing the substitution 79G->A in the protospacer adjacent motif (PAM) region, a short, specific sequence following the target DNA sequence, that is essential for cleavage by the Cas nuclease of the targeted editing sequence, the percentage frequency of which was less than 0.1% in wild type stamens, but rose to 3.38% in PST and 8.02% in AST stamens, respectively ( Fig. 2d and Supplemental Fig. 3). The results demonstrated that the mitoCRISPR/Cas9 system could specifically cleave the mtatp9 alleles, resulting in targeted elimination of mtDNAs in PST and AST stamens. This elimination was followed by reproduction of the remaining mtDNAs, resulting to the heteroplasmic shift of mtatp9 alleles.

Stamen transcriptomic analysis of wild type tobacco and its male-sterile mutant
To reveal the molecular mechanism of stamen abnormality, the mutant exhibiting only AST flowers was used for further investigation. RNA-Seq was carried out with three replicates for each sample to analyze the variation in level of nuclear gene expression between wild type and AST stamens. More than 8 Gb of data were obtained from each sample. A total of 341,710,510 high-quality reads were obtained as a result of quality control and data assembly. Over 91.88% of the total clean reads were mapped to genes, and the total length was 55,756,576,500 bp, with a Q30 quality score of 94.42%.
Use principal component analysis to show that samples are correctly grouped (Supplemental Fig. 4). A heatmap of gene expression between wild type and AST stamens showed that samples from the same source were clustered together (Supplemental Fig. 5). The results indicated that the transcriptomic data were reliable and of high quality, so it could be used for further analysis.
A total of 6440 differentially expressed genes (DEGs) between wild type and AST stamens were screened using | log2 fold change (FC) | ≥2 and p value < 0.01 as standards. Expression levels of a total of 458 genes were up-regulated, and 5982 genes were down-regulated in AST relative to wild type stamens (Supplemental Fig. 6). We performed GO classification and KEGG pathway analysis on the DEGs. Galacturonan metabolic process, pectin metabolic process, pectin catabolic process, and carbohydrate metabolic process were highly enriched GO items in the genes downregulated in AST stamens. KEGG metabolic pathway analysis showed that DEGs were significantly enriched (corrected p value < 0.05) in amino sugar and nucleotide sugar metabolism, glycolysis/gluconeogenesis, nucleotide sugar metabolism, pentose and glucuronate interconversions, starch and sucrose metabolism, and pyruvate metabolism (Supplemental Fig. 7). All of these metabolic pathways were related to carbohydrate metabolism, and almost all DEGs exhibited down-regulated expression in AST stamens relative to wild type stamens (Supplemental Tab. 2).
Respiration involves the glycolysis pathway, the tricarboxylic acid cycle, and the oxidative phosphorylation pathway. In the glycolysis pathway, expression of a total of 27 DEGs encoding ten enzymes was significantly down-regulated in Fig. 2 Relative quantification of mtDNA copy number and the ratio of mtatp alleles in stamens. a The relative quantification of mtDNA copy number using real-time quantitative PCR of mtatp1, mtatp9, and mtorfX. b-d The percentage of typical mtatp9 alleles (b), atypical mtatp9 alleles (c), and atypical mtatp9 alleles with substitution at nt 79G->A (d) by Hi-TOM detection. Error bars represent ± SD, **p < 0.01, ***p < 0.001, ****p < 0.0001 following ANOVA and Tukey's post hoc test. WT, wild type; PST, petaloid stamen type mutant; AST, abortive stamen type mutant AST stamens, with log2 FC from −2.00 to −5.52, whereas expression levels of six DEGs encoding fructose-1,6-bisphosphatase I (FBP) and aldolase (ALDO) were significantly upregulated in AST stamens, with log2 FC 2.01-3.09. During the tricarboxylic acid cycle, expression levels of a total of 14 DEGs encoding seven enzymes were significantly downregulated in AST stamens, with log2 FC −2.32 to −3.72, whereas expression levels of three DEGs encoding malate dehydrogenase 1 (MDH1) were significantly upregulated, with log2 FC of 2.05-2.09. In addition, the phosphoenolpyruvate-carboxykinase (pckA) and acetyl-CoA synthetase (ACS), which link the glycolysis pathway and the tricarboxylic acid cycle, also showed down-regulated expression in AST stamens. Although FBP, ALDO, and MDH1 were upregulated, these increases may not promote the glycolysis pathway and the tricarboxylic acid cycle because of their reverse reaction functions. The results suggested that the glycolysis pathway and the tricarboxylic acid cycle were constrained in AST stamens (Fig. 3a). Furthermore, levels of expression of the DEGs related to the oxidative phosphorylation pathway were also downregulated in AST stamens (Supplemental Fig. 8).
The method of RNA-Seq used is to fish polyA and build library. However, the mitochondrial genes usually do not have PolyA, and the library cannot be built and sequenced. Because our RNA-Seq work obtained transcriptomic profiles of only nuclear genes, the expression levels of ATP synthase subunits encoded by mtDNA were detected by qPCR. Compared with the wild type, the relative expression levels of mtatp1, mtatp6, mtatp8, and mtatp9 in AST decreased by 45.82%, 32.97%, 29.28%, and 53.69%, respectively, although the expression of mtatp4 in AST stamens was similar to that of the wild type (Fig. 3b). These results indicated that the heteroplasmic shift of mtatp9 alleles may limit expression of all of the pathways related to respiration and hence influence the energy supply to AST stamens.
In addition, potential off-target sites for non-destination mutations in the AST stamens were analyzed using transcriptomic data. Based on wild type transcriptomic data, five potential off-target sites were predicted, and no mutations were detected at these sites in the AST transcriptomic data (Supplemental Tab. 3). The potential off-target sites in the nuclear genome were not edited as mito-Cas9 was not transported into the nucleus.

Overexpression of dsatp9 can restore fertility
To restore the fertility of the AST flowers, complementation studies were carried out in AST plants. The dsatp9 gene contains a synonymous sequence at the mtatp9 target site to prevent cleavage by mitoCRISPR/CAS9 and contains a mitochondrial localization signal at the N-terminus. The dsatp9 gene was ligated to the backbone of the overexpression vector pCambia2300 (Fig. 4a), and the constructed vector was introduced into the leaves of AST mutant Nmu81 by Agrobacterium transformation. The independent dsatp9-postive plants showed normal stamen, pollen, and capsules, whereas the null-transgenic AST plants regenerated from AST leaves without dsatp9 transformation were still male sterile (Fig. 4b). In addition, the height of the fertile transgenic plants was similar to that of wild type plants; conversely, the height of AST plants was significantly lower than that of wild type plants (Supplemental Fig. 9). Restoring the fertility of the AST mutant via overexpression of the dsatp9 gene proved that dysfunction of the mtATP9 subunit was responsible for the abnormal development of the stamens.

Discussion
Recently, the CRISPR-Cas9 system is particularly popular in nuclear genome editing because of its simple operation and high efficiency (Li et al. 2021). The advent of genome editing has stimulated attempts to manipulate mtDNA, because of the important role of mtDNA in different areas of biology and medicine. However, whether the nucleic acid can be imported into mitochondria, a prerequisite for mitochondrial CRISPR/Cas9 gene editing remains controversial (Gammage et al. 2018). The RNA mitochondrial import pathway may be a unique natural mechanism, although the relatively low abundance of mtRNA makes it difficult to confirm this as an endogenous RNA mitochondrial import pathway. Several important features of this mechanism have been discovered, and mitochondrial import of nuclear-encoded tRNAs has been described in yeast, plants, and protozoans (Kolesnikova et al. 2000(Kolesnikova et al. , 2004Karicheva et al. 2011). Interestingly, gRNA molecules were detected in purified mitochondria, suggesting that gRNA can import into mitochondria independently, without extra import determinants (Loutre et al. 2018). In the present study, successful transport of mito-Cas9 protein into mitochondria was observed, consistent with previous findings (Bian et al., 2019). Selective elimination of mtDNAs demonstrated that the mito-CRISPR/Cas9 system effectively cleaved target sites within mtDNA. This result indirectly confirmed the importation of both mito-Cas9 protein and sgRNA into mitochondria.
Non-homologous end joining and homologous recombination, two primary self-repair mechanisms of DBS in the nucleus, may be absent or inefficient in mammalian mitochondria (Gammage et al. 2018). However, an abundance of repeat sequences in plant mtDNAs is involved in frequent homologous recombination of large repeats (Gualberto and Newton 2017). In rice and rapeseed, CMS-associated genes were knocked out by homologous recombination of large repeats when mito-TALENs induced DSB (Kazama Fig. 3 The expression levels of genes in stamens. a Structural genes in glycolysis pathway and tricarboxylic acid cycle metabolism with differential expression levels between the wild type (WT) and abortive stamen type (AST) mutant stamens using RNA-Seq. Up-regulated genes are marked in red and down-regulated genes in green in AST. PGM, phosphoglucomutase; GPI, glucose-6-phosphate isomerase; PFK, 6-phosphofructokinase 1; PFP, diphosphate-dependent phosphofructokinase; FBP, fructose-1,6-bisphosphatase 1; ALDO, fructose-bisphosphate aldolase, class I; GAPDH, glyceraldehyde 3-phosphate dehydrogenase; PGK, phosphoglycerate kinase; PGAM, 2,3-bisphosphoglycerate-dependent phosphoglycerate mutase; GPMI, 2,3-bisphosphoglycerate-independent phosphoglycerate mutase; ENO, enolase; pckA: phosphoenolpyruvate carboxy kinase; LDH, L-lactate dehydrogenase; ACS, acetyl-CoA synthetase; CS, citrate synthase; ACLY, ATP citrate (pro-S)-lyase; ACO, aconitate hydratase; IDH3, isocitrate dehydrogenase (NAD+); SDHA, succinate dehydrogenase (ubiquinone) flavoprotein subunit; SDHB, succinate dehydrogenase (ubiquinone) iron-sulfur subunit; FH, fumarate hydratase, class II; MDH1, malate dehydrogenase 1. b The expression levels in stamens of ATP synthase subunits encoded by mtDNA. Gene expression was measured by real-time quantitative PCR. Error bars represent ± SD, *p < 0.05 following Student's t-test. WT, wild type; AST, abortive stamen type mutant et al. 2019). The presence of large repetitive fragments in the vicinity of the DBS locus is a prerequisite for successful homologous recombination. However, no large repeat fragment was discovered in the vicinity of mtatp9 locus in mtDNA when performing BLAST comparison between the flanking sequence of mtatp9 and tobacco mtDNA. This suggests that homologous recombination at the mtatp9 locus would not be triggered. The similarities of the mtatp9, mtatp1, and mtorfX copies confirmed this speculation. So, the broken mtDNA cleaved by mito-CRISPR/Cas9 system would be quickly degraded in mitochondria, and reproduction of the remaining mtDNA resulted in percentage heteroplasmic allelic shifts.
Furthermore, pathogenic mtDNA mutations are always mixed with normal mtDNA. Once the number of copies of mutant mtDNA exceed a threshold level, a mitochondrial disease phenotype may appear. Szklarczyk et al. (2014) revealed the heteroplasmy of mtatp9 alleles in different carrot plants. In petaloid male-sterile plants, the atp9-1 version dominates over atp9-3, while in male-fertile plants, this proportion is reversed. Here, the percentage of the typical mtatp9 gene was 80.75% in wild type stamens but decreased to 54.57% and 42.17%, respectively, in PST and AST stamens. This results once again demonstrated that the heteroplasmy of mtatp9 alleles was one of the important factors affecting stamen development. The decrease in proportion of typical mtatp9 genes may induce dysfunction of the mtATP9 subunit and energy supply imbalance. This conjecture was confirmed by stamen transcriptomic analysis and qPCR. Compared with wild type stamens, expression of the glycolysis pathway, tricarboxylic acid cycle, and oxidative phosphorylation pathway, which are three stages of respiration, was restricted in AST stamens. The result also provided a molecular explanation for the male sterility of AST mutants. The dysfunction of the mtATP9 suppressed the respiration of AST stamens. Stamen development is the high-energy differentiation. Insufficient energy supply caused stamens to undergo abnormal development. Finally, we reversed the male sterility of the AST mutant via the overexpression of the dsatp9 gene, further confirming that mtatp9 dysfunction is the cause of male sterility in AST tobacco mutants.

Plant materials
Tobacco (Nicotiana tabacum) "Samsun" plants were grown in a greenhouse under controlled conditions with a 16 h photoperiod and a constant temperature of 25 °C. "Samsun" tobacco material and plasmids were provided by the Key Laboratory of Crop Molecular Breeding, Northwest Institute of Plateau Biology, Chinese Academy of Sciences, Xining, Qinghai Province, China. All primers used in this study and listed in Supplemental Tab. 1 were synthesized by Sangon Bioengineering Co., Ltd (Shanghai, China).

Construction of the CRISPR/mtCas9 vector
The SpCas9 gene from the Streptococcus pyogenes type II CRISPR/Cas system was optimized according to the codon usage bias of the plant, and the mitochondrial targeting sequence (cytochrome C oxidase subunit IV presequence from Saccharomyces cerevisiae, COX IV) was added. The designed target sequence (GCT TCA GCG GGA GCT GCT AT) was selected at the 5′-coding region of the tobacco typical mtatp9 gene by online software CRISPR-GE (https:// skl. scau. edu. cn/ targe tdesi gn/). A complete single guide RNA (sgRNA) expression frame contained the 3′ region of the AtU3d and the target sgRNA sequence. The total sequence of the mtCas9 gene and the sgRNA expression frame was then synthesized by Genewiz Biotechnology Co., Ltd (Suzhou, China). The mito-Cas9 vector was constructed by ligating the synthesized sequence with linearized pCambia 1300 binary vector digested by Hind III and Kpn I.

Tobacco transformation
The vector was transferred into Agrobacterium tumefaciens strain LBA4401 competent cells (Biomed Gene Technology Co., Ltd, Beijing, China) by the freeze-thaw method (Holsters et al. 1978). The positive clone identified by screening with PCR using the specific primers Cas9F and Cas9R (Supplemental Tab. 1) was cultured in liquid MG medium containing 100 mg/L kanamycin and 100 mg/L rifampicin at 28 °C and 200 rpm for 48 h. When the OD600 value reached 0.8, genetic transformation of small, young tobacco leaves was carried out using the leaf disk transformation/ regeneration method (Horsch et al. 1985). Firstly, the leaves were placed in the co-culture medium in the dark for 2 days, transferred to solid differentiation medium for 7 days, and then transferred to solid differentiation medium containing 30 mg/L hygromycin for 14 days. Shoots were transferred to rooting medium until they rooted, after which the plantlets were acclimated and transplanted into soil.
To identify transgenic tobacco plantlets, genomic DNA was extracted from leaves from regenerated tobacco plantlets using Genomic DNA Extraction Kit (Takara, Beijing, China), and PCR was performed using specific primers Cas9F and Cas9R (Supplemental Tab. 1). The PCR reaction volume was 20 μL, in which DNA template was 1 μL, 2×SanTaq PCR Mix 10 μL (Sangon, Shanghai, China), primers were each 1 μL, and ddH2O 7μL. PCR reaction conditions were pre-denaturation at 94 °C for 5 min, denaturation at 94 °C for 30 s, annealing at 58 °C for 30 s, extension at 72 °C for 1 min for 32 cycles, and then a final extension at 72 °C for 10 min. An aliquot (3 μL) of the products was separated by electrophoresis on 1.0% agarose gel.

Mitochondrial extraction and western blotting
Mitochondria were extracted from leaves using plant mitochondrial extraction kit (Shyuanye, Nanjing, China). Mitochondrial protein was separated on 8% sodium dodecyl sulfate (SDS)-polyacrylamide gel, transferred from the gel to PVDF membranes by an electric transfer system, and incubated with antibody against Cas9 (GenScript, Nanjing, China) overnight at 4 °C. The membranes were incubated with the secondary antibody coupled to HRP (ProteinTech, Chicago, USA) at 24 °C for 1 h. Finally, the membranes were assessed by western chemiluminescent HRP Substrate on Tanon 5200 (Tanon, Shanghai, China).

Staining of pollen activity by the TTC method
The viability of different tobacco pollens was determined by the 2, 3, 5-triphenyl tetrazolium chloride (TTC) viability assay. Mature pollens were placed onto glass slides and stained with 10 μL 2% (w/v) TTC solution (Solarbio, Beijing, China) at 35 °C for 15 min. The phenotypes of the pollen grains (stained = viable, unstained = non-viable) were observed, and images were captured using an Olympus BX53 light microscope (Olympus Corp, Tokyo, Japan). The experiment was performed in three biological replicates, with five fields-of-view observed in each replicate.

Relative quantification of the mtDNA copy number in stamens
Total DNA were extracted from stamens using Genomic DNA Extraction Kit (Takara, Beijing, China) for real-time quantitative PCR (qPCR), and three biological replicates were carried out. The quality of the genomic DNA was determined spectrophotometrically (NanoDrop 2000c;Thermo, Shanghai, China). The total DNAs of samples were diluted to 40 ng/μL, respectively. The mtatp9, mtatp1, and mtorfX (transport membrane protein) of mtDNA were selected for monitoring the copy number of mtDNA ( Fig. 1a and supplemental Tab. 1). The housekeeping gene α-tubulin was co-amplified as a control for normalizing DNA templates (Supplemental Tab. 1).
qPCR was performed using the 7500 Fast Real-Time PCR System (ABI, Carlsbad, USA). The reaction mixture (20 μl) contained 2 μl of DNA, 0.8 μl each of forward and reverse primers (from 10 μM stock), 10 μl TB green, 0.4 μl ROX (TB Green Premix Ex Taq II; Takara, Tokyo, Japan), and 6 μl nuclease-free water. Each sample was analyzed with three technical replicates. The PCR program was set up with an initial denaturation step at 95 °C for 3 min, followed by 40 cycles at 95 °C for 20 s and 60°C for 1 min. The dissociation curve for checking the specificity of PCR production was acquired by adding a step at 95 °C for 15 s. The relative copy number of target genes was calculated by the 2−ΔΔCt method (Livak and Schmittgen 2001). We defined the change of mtatp9 in mutiples in wild type as 1 for normalizing the relative copy number of other mitochondrial genes.

High-throughput tracking of mutations
The proportion of mtatp9 alleles in stamens of wild type and mutant tobacco was explored using the high-throughput tracking of mutations (Hi-TOM) method (Liu et al. 2019). Genomic DNA of stamens of wild type and mutant tobacco was extracted and used as template for the first round of PCR amplification, using mtatp9 specific primers, TOMATP-F and TOMATP-R (carrying the bridging sequence; Supplemental Table 1), and three biological replicates were carried out. Then, the second round of PCR amplification was carried out with Hi-TOM Mix (Novogene, Tianjin, China). The second round PCR reaction system (20 μL) consisted of Hi-TOM Mix 10 μL, first-round PCR product 1.0 μL, and ddH2O 9.0 μL. The PCR reaction procedure involved pre-denaturation at 94 °C for 5 min, denaturation at 94 °C for 30 s, annealing at 58 °C for 30 s, extension at 72 °C for 1 min for 32 cycles, and then a final extension at 72 °C for 10 min. An aliquot (3 μL) of the products was separated by electrophoresis on 1.0% agarose gel. High-throughput sequencing of the second round of PCR amplification products was carried out by Novogene Co., Ltd (Tianjin, China). The results of sequencing were analyzed using the Hi-TOM online tool (https:// www. hi-tom. net/ hi-tom/ index-CH. php), and the sequence of typical mtatp9 allele was identified as WT in the Hi-TOM analysis.

RNA-sequencing
The stamens of mutant and wild type tobacco were selected to construct cDNA libraries. Six libraries, each with three biological replicates, were prepared and sequenced using an Illumina HiSeq 2000 system (Novogene, Tianjin, China). The reference genome GCF_000715135.1_Ntab-TN90 was used for RNA-seq analysis. Before data assembly, the original sequencing results were preprocessed using CUTADAPT (4.3) to remove low-quality sequences and obtain high-quality sequences. High-quality sequencing data were mapped using HISAT2 (2.2.1) a short-read assembly program, to obtain reliable transcripts.
The Gene Ontology (GO) and Kyoto Encyclopedia of Gene and Genome (KEGG) databases were used for annotating the gene function. Gene expression levels were calculated using fragments per kb per million read (FPKM) values. The absolute value of |log 2 ratio| ≥ 1 was used as the threshold to determine the significance of differential expression levels of genes. Significantly enriched GO and KEGG terms were obtained using the R platform (3.6.0).

Expression levels of genes encoded by mtDNA
For cDNA synthesis, total RNA of stamens was extracted using RNAprep Pure Plant Kit (Tiangen, Beijing, China), and three biological replicates were carried out. The quality of the RNA was determined spectrophotometric methods (NanoDrop 2000c;Thermo, Shanghai, China), and 800 ng RNA was reverse-transcribed using PrimeScript RT reagent Kit with gDNA Eraser (Takara, Tokyo, Japan). The ten-fold dilution of cDNAs was used as template for real-time quantitative PCR applications. Using the appropriate primers (Supplemental Tab. 1), mitochondrion-encoded genes of interest were amplified with three technical replicates. The housekeeping gene α-tubulin was co-amplified as a control for normalizing cDNA templates, and qPCR was performed as described earlier. The expression levels of target genes were calculated by the 2−ΔΔCt method (Livak and Schmittgen 2001).

Effect of overexpression of atp9 on fertility recovery
The target region of mtatp9 gene was codon modified by overlapping extended PCR to generate synonymous mutations (dsatp9) so that the CRISPR/Cas9 ribonucleoprotein complexes could not cleave the dsatp9 gene sequence of the overexpressed vector. Moreover, the COX IV presequence which targeted mitochondria was added to the 5′ end of the dsatp9 gene sequence to ensure that the dsATP9 protein could enter mitochondria smoothly. The overexpression vector of dsatp9 was constructed by linking the linearized vector pCambia2301-KY, digested by restriction endonucleases BamHI and SacI, with the fragment containing the COX IV presequence and dsatp9 by homologous recombination using ClonExpress II One Step Cloning Kit (Vazyme Biotech Co., Ltd., Nanjing, China). The overexpression vector of dsatp9 was introduced into abortive stamen type tobacco leaves via the leaf disk transformation/regeneration method (Horsch et al. 1985). Regenerated tobacco plantlets were obtained using kanamycin as the selection agent. AST leaves were transformed by Agrobacterium tumefaciens without overexpression vector of dsatp9, and the regenerated plants were null-transgenic AST regenerated plants.

Statistical analysis
The statistical analysis was performed with Tukey's post hoc ANOVA for multiple comparisons and Student's -test for comparison of two groups. Statistical comparisons were made with PASW Statistics 18 (IBM SPSS, Chicago, USA). A value of p < 0.05 was considered to be statistically significant, and all values from a sample were expressed as the mean ± standard deviation. Each experiment was replicated three times independently.

Conclusion
In this study, the mito-CRISPR/Cas9 system could selectively eliminate mtDNA and induce heteroplasmic mtatp9 allelic shifts. The dysfunction of the mtATP9 caused lower energy supply and triggered abnormal stamen development. Our work has confirmed that the mtatp9 is a key gene for CMS and proved the potential of the mito-CRISPR/Cas9 system as a tool for mtDNA research and for future therapy of mitochondrial disorders in plants.