Fresh leaves of P. pedunculata were collected from Hohhot, Inner Mongolia, China (40.57°N, 111.93°E) and deposited in the National Medium-Term Genebank Forage Germplasm (Hohhot, China). Genomic DNA was extracted from fresh leaves using a Plant DNA Isolation Kit (Tiangen, Beijing, China) and sequenced using an Illumina MiSeq platform (Novogene Co., Ltd., Tianjing, China). Around 7.52 Gb clean data with 50.16 million reads were yield and used for mitogenome de novo assembling. The assembly of mitochondria was performed using the software GetOrganelles V 1.7.5.3 (Jin et al., 2020) with default parameters (-R 50 -k 21,45,65,85,105,115,127 -P 1000000) (Jin et al., 2020). The accuracy of the assembly results was checked using the visualization software Bandage (Wick et al., 2015) and by mapping clean reads using Bowtie2 (Langmead and Salzberg, 2012). Afterwards, the average coverage depth was assessed to be 242.9x by SAMtools (Li et al., 2009) ( Supplementary Figure S1 ). The coverage was visualized by Integrative Genomics Viewer (IGV) (Thorvaldsdottir et al., 2012). The cp genome of P. pedunculata was assembled in a similar manner. The complete mitochondrial genome sequence was deposited in GenBank (accession number: OQ 556854.1) and the complete chloroplast genome sequence was deposited with accession numbers of OR343251.

P. pedunculata mitogenomes were annotated using GeSeq (Tillich et al., 2017) with reference to previously released mitogenome data of Prunus species and then manually adjusting the data into a circular mitogenome model. The cp genome was annotated using Plastid Genome Annotator (PGA) tools (Qu et al., 2019). Subsequently, Geneious V9.0.2 was used to amend mistaken codons (Kearse et al., 2012). The genome map was visualized using the Organellar Genome Draw (OGDRAW) software (Greiner et al., 2019).

Simple sequence repeats (SSRs) of the P. pedunculata mitochondrial genome were identified using the MISA software (Beier et al., 2017) with the parameters of minimum nucleotide numbers of mono-10, di-6, tri-4, tetra-3, penta-3 and hexa-3, respectively. Tandem Repeats Finder (TRF) (Benson, 1999) was used to identify tandem repeats with default parameters, whereas dispersed repeats larger than 70 bp were identified as forward, reverse, palindromic, and complementary repeats using the online tool REPuter (Kurtz et al., 2001) with a Hamming distance of 3 and a cutoff e-value of 1e-5.

The RSCU value of PCGs and their amino acid composition were calculated by the Molecular Evolutionary Genetics Analysis software (MEGA v11.0.26) (Tamura et al., 2021), codon preferences were configured using Perl scripts.

Non-synonymous (Ka) and synonymous (Ks) substitution rates were calculated using DnaSP 6.12.0 (Rozas et al., 2017), based on a total of 20 shared PCGs (atp1, atp4, atp6, atp8, atp9, ccmB, ccmC, ccmFc, ccmFn, cob, cox2, cox3, mttB, nad2, nad3, nad4, nad5, nad6, nad9, and rps13) between the mitogenomes of P. pedunculata and 29 other Rosaceae species and Oryza sativa, Triticum aestivum.

Based on three RNA-seq datasets of P. pedunculata deposited in the SRA database (https://www.ncbi.nlm.nih.gov/sra/; accession numbers: SRR13261917, SRR13261918 and SRR13261919) (Bao et al., 2021), we identified the putative RNA editing sites in mitochondrial PCGs. Mapping the RNA-seq data onto the sequences of mitochondrial PCGs by using BWA v0.7.15 (Li and Durbin, 2010) software. Then we called single nucleotide polymorphism sites (SNPs) by using SAMtools v1.17 (Li et al., 2009) and BCFtools v1.17 (Danecek et al., 2021). To identify and annotate RNA editing sites, the SNP-calling data were processed using REDO v 1.0 (Wu et al., 2018), a specialized tool for easily identifying RNA editing sites in plant organelles. To exclude the false positive RNA editing sites, BWA v0.7.15 was used to mapping the DNA-Seq data to P. pedunculata mitogenome. The SNP-calling method was conducted using BCFtools, then eliminating the RNA editing sites detected in genomic SNPs.

The chloroplast genome data for P. pedunculata were obtained from our assemblies. Homologous sequences between the chloroplast genome and mitogenome were identified, and the transferred DNA fragments were screened using BLASTN with a cutoff value of 1e-5. Gene transfer from the chloroplasts to mitochondria was visualized using TB tools (Chen et al., 2020).

Phylogenetic analysis based on 20 shared PCGs (as mentioned in Section 2.5) derived from the complete mitogenomes of 30 selected Rosaceae species was performed, with O. sativa and T. aestivum as the outgroups. Mitogenome data of the reference accessions were downloaded from the NCBI ( Supplementary Table S1 ). The corresponding nucleotide sequences of the PCGs in the chosen genomes were concatenated, and the MAFFT program (Katoh and Standley, 2013) was used to perform multiple sequence alignment. Both the ML algorithm and Bayesian methods were used to construct the phylogenetic tree, and the best models were selected using ModelFinder (Kalyaanamoorthy et al., 2017). The ML method was conducted using RAxML (Stamatakis, 2006) with the GTRGAMMA model and bootstrap of 1000 replicates. Bayesian inferences (BI) using MrBayes v3.2.6. (Ronquist et al., 2012) were calculated to select the best-of-fit model GTR+F+I+G4.

The complete mitochondrial genome of P. pedunculata was assembled into a typical single circular molecule with a size of 405,855 bp ( Figure 1 ) and a GC content of 45.63% ( Table 1 ). The mitogenome comprised 27.06% adenine, 27.31% thymine, 22.77% guanine, and 22.85% cytosine. Although 84.82% of the mitogenome was composed of non-coding regions, the proportion of PCGs and cis-spliced-introns was 7.39% and 6.58% in the mitogenome, respectively, and that of tRNA and rRNA accounted for 0.45% and 1.27%, respectively.

The mitochondrial genome of P. pedunculata contains 14 core genes, including five ATP synthase genes (atp1, atp4, atp6, atp8, atp9), four cytochrome C biogenesis genes (ccmB, ccmC, ccmFc, ccmFn), three cytochrome c oxidase genes (cox1, cox2, cox3), one ubiquinol cytochrome c reductase gene (cob), and one maturase gene (matR) ( Table 2 ). In addition, an atp6-like pseudogene was detected. The genome also comprises 20 variable genes, three ribosomal RNAs (rrn18, rrn26, rrn5), and 19 transfer RNAs. These variable genes included nine NADH dehydrogenase genes (nad1, nad2, nad3, nad4, nad4L, nad5, nad6, nad7, nad9), two large ribosome protein subunits (rpl10, rpl5), six small ribosome protein subunits (rps1, rps12, rps13, rps3, rps4, rps7), one transport membrane protein (mttB), and two succinate dehydrogenase genes (sdh3, sdh4). In addition, the rpl16-like and a rps14-like sequences were identified as pseudogenes. As for transfer RNAs, trnS-GCT and trnM-CAT had three copies, while trnC-GCA, trnP-TGG, and trnY-GTA, had two copies. Other tRNA genes were represented by identical copies in the mitogenome. Intron sequences were identified in nine genes ( Table 2 ), among which nad1, nad2, nad5, and nad7 contained four introns; nad4 contained three introns; and ccmFc, cox2, rps3, and one copy of trnY-GTA possessed one intron region.

Repeat sequences included SSRs, tandem repeats, and dispersed repeats. A total of 50 interspersed repeats were identified in the P. pedunculata mitogenome, including 33 palindromic repeats (P, 66%) and 17 forward repeats (F, 34%); no reverse or complement repeats were detected. The lengths of the repeats were unevenly distributed. Most of the repeat sizes were between 70 bp-190 bp (78%), and 11 repeats (22%) exceeded 200bp, whereas only one repeat was longer than 1 kb, which was the largest repeat (1257 bp) ( Figure 2 ; Supplementary Table S2 ).

With regards to the SSRs, the distribution and structure of 112 SSR repeats in the P. pedunculata mitogenome were analyzed, comprised 45 single-nucleotide motifs (40.18%), five dinucleotide repeats (4.46%), 12 trinucleotide repeats (10.71%), 40 tetranucleotide repeats (35.71%), four pentanucleotide repeats (3.57%), and six hexanucleotide repeats (5.36%) ( Figure 3 ; Supplementary Table S3 ). Among all the SSRs, most repeats (82, 72.32%) were rich in A/T. Notably, 34 SSRs were entirely composed of A/T, including 26 monomer units (A/T), three dimer units (AT/TA), two trimer units (AAT/TTA, TAT/ATA), and three tetramer units (AAAT/ATTT, AATT/TTAA). The A/T richness of the 48 SSRs ranged from 50 to 80%. The majority of SSRs (95) were located in IGS region, while nine SSRs were distributed on introns, two on exons, and one on the ORF region of rps1; only one SSR repeat was positioned across the intron region of nad1 and part of the coding region of matR ( Supplementary Table S4 ). These extensive SSRs provide abundant potential molecular markers for the identification and genetic study of Prunus.

Additionally, 15 tandem repeat sequences ranging from 14 to 51 bp were evenly distributed in the mitogenome of P. pedunculata with a similarity match greater than 80%. These tandem repeats were predominantly located in the IGS, and only one repeat resided on rrn26 ( Supplementary Table S5 ). However, not all the repeat sequences were copied. Certain sequences had multiple non-integral copies. For instance, the TACATATTCGAGAA motif was repeated twice in the IGS between atp9 and nad1, whereas the GACTATGAAACAGATCGC repeat unit was present in rrn26 with a repetition number of 2.4.

The codon usage of 34 PCGs in the P. pedunculata mitogenome with a total length of 29,973 bp, encoding 9991 codons was analyzed. The results of RSCU analysis are shown in Figure 4 . Leucine (Leu) was a predominant amino acid in PCGs with a frequency of 1066 (10.67%), followed by serine (8.98%) and isoleucine (7.97%), whereas cysteine (Cys) and tryptophan (Trp) were the least adopted amino acids, which only occurred 146 and 145 times (1.46% and 1.45%, respectively). The PCGs had the highest preference for UUU (Phe), which was used 364 times in PCGs, with an RSCU value of 1.12. The TAG termination codon was the least frequently used ( Table 3 ; Supplementary Table S6 ). Alanine (Ala) had a preference for GCT, which occasionally occurred 248 times in PCGs, with a maximum RSCU value of 1.56.

Interestingly, in the PCGs, A/T bases preferentially appeared in the third codon position, rather than C/G. Codons ending in A/T had RSCU values greater than 1. In addition, almost all PCGs began with the typical start codon ATG, except for nad1 which started with ACG. Five models of termination codons were observed in the PCGs. TNAs (TAA, TGA, and N for A, T, C, or G, respectively) were the most dominant codons in the 24 PCGs. Six PCGs (atp4, mttB, matR, nad7, rps1, and ccmFn) were terminated with TAG. Meanwhile, ccmFc, atp9, and sdh4 were stopped by CGA and atp6 ended with CAA, which may be incomplete codons ( Supplementary Table S7 ).

Non-synonymous and synonymous substitution ratios (Ka/Ks) were calculated for the mitogenomes of 32 species based on 20 homologous PCGs. P. pedunculata was used as the reference. In most PCGs, the Ka/Ks values were notably less than 1 ( Figure 5 ; Supplementary Table S8 ), implied that these genes were dominated by purifying selection during evolution. Conversely, the Ka/Ks ratios of ccmC in P. anserina, ccmFn in R. chinensis and R. rugosa versus P. pedunculata were greater than 1(1.02956, 1.12327, and 1.05751, respectively), inferring positive selection. In the case of nad2 from the P. kanzakura and P.yedoensis mitogenome, the Ka/Ks ratios were close to 1 (0.99014), suggesting a tendency of neutral selection. Additionally, 68 pairwise Ka/Ks values were 0 and 120 were pairwise with non-applicable (NA) Ka/Ks values. The highest average Ka/Ks ratios were observed for cytochrome c biogenesis genes (0.62132). Meanwhile, the lowest values of average Ka/Ks ratios were noted for atp9(0.0539), nad3(0.09952) and nad9(0.06257). In particular, the maximum substitution ratios of atp9, atp1, and nad9 were as low as 0.235319, 0.325493, and 0.326979, respectively. The low Ka/Ks values suggested that these genes may have been highly conserved during the evolution of the P. pedunculata mitogenome.

RNA editing events, especially the C-to-U editing sites, are enriched in plant mitogenomes. A total of 262 RNA editing sites in 26 analyzed PCGs of P. pedunculata mitogenome were identified among which 249 sites exhibiting C-to-U RNA editing ( Figure 6 ; Supplementary Table S9 ). Most of the predicted RNA editing events occurred at the first (76, 29.01%) or second (173, 66.03%) positions of the codons, only 13 were found in the third positions (4.96%). The largest number of RNA editing sites was detected in the NADH dehydrogenase genes (129), among them nad7 (34) maintained the most editing sites in all mitochondrial genes, then followed by nad4 (28). There is only one editing site were predicted in rps1 and rps3, respectively. No potential RNA editing sites were identified in atp6, ccmB, ccmC, ccmFc, mttB, nad4L, nad9, rps7 and rps13, no RNA editing sites were found in two assumed pseudogenes (rpl16, rps14), either. The majority of RNA editing sites (247) were non-synonymous variations, synonymous editing sites were merely (15) found. A total of 31 types of amino acid conversion were identified at these RNA editing sites ( Supplementary Table S9 ), including two special sites in atp9 and nad7, which convert to termination codons. The most frequently occurred amino acid changes among all of the identified mutations were histidine (H) to leucine (L) and serine (S) to leucine (L) change, with frequency of 61 times (23.28%) and 55 times (20.99%), respectively. Besides, nad1 was initiated with ACG as its start codon ( Supplementary Table S7 ), implying an alteration caused by RNA editing event.

The mitogenome (405, 855 bp) of P. pedunculata was approximately 2.57 times larger than the chloroplast genome (157,830 bp), so the distribution of mitochondrial genes in P. pedunculata was relatively sparse compared to that of chloroplast genes ( Figure 7 ). In this study, 56 chloroplast-like fragments that might have undergone gene transfer were identified in the mitogenome, based on sequence similarity between the chloroplast and mitochondrial genomes of P. pedunculata ( Figure 7 ; Supplementary Table S10 ). These inserted fragments, ranging from 30 to 863 bp, were distributed on the mitochondrial genome, with a total length of 11,582 bp, which comprised 2.85% of the complete mitogenome. The longest sequence (863 bp) was transferred from rrn16S and rrn16S-2 in the cp genome to rrn18 in the mitogenome. These migrated sequences were mainly located in the IGS (19) or rRNA genes (20) of the mitogenome. There are 22 rrn16S and rrn23S sequences of the P. pedunculata chloroplast genome that were inserted into the mitogenome and were mostly transferred into rrn18 or rrn26, except for two fragments that were transferred to IGS regions. Among the remaining chloroplast-like sequences, nine fragments were located on tRNA genes, three on core genes (atp1, ccmC), and two on the ribosomal protein rsp12. Some transferred fragments were intact genes (atp1, rrn18, rrn26, rps12, ccmC, trnD-GTC, trnH-GTG, trnN-GTT, trnF-GAA, trnP-TGG, trnQ-TTG), whereas others were partial sequences (trnW-CCA, trnP-TGG-2, trnN-GTT, trnM-CAT-3).

Migration may occur from gene to gene, from gene to IGS, from IGS to IGS, or from gene to introns/exons. Our findings showed that 22 rRNA sequences were transferred from chloroplasts to the mitochondria, mostly maintaining the function of ribosomal RNAs; whilst only two were inserted into the IGS region and became nonfunctional. Most sequences from the encoded genes of the chloroplasts lost their function and relocated to the IGS of the mitochondria. In contrast, atpA from the chloroplasts was transformed into atp1 in the mitochondria. Five segments that immigrated from the IGS remained in the IGS regions. The transferred sequences of 14 tRNA genes were shared between the chloroplast and mitochondrial genomes, accounting for 25% of the total transferred sequences. Relocated sequences were also observed in the intron and exon regions of some genes.

Phylogenetic trees between P. pedunculata and 30 other Rosaceae species were constructed using the ML and BI methods, based on 20 shared mitochondrial PCGs and 72 shared chloroplast PCGs, respectively. O. sativa and T. aestivum functioned as outgroups. ( Figure 8 ; Supplementary Table S1 ). Both ML and BI analyses indicated that most branches of the phylogenetic tree had high support values and the topology presented high consistency. The evolution relationships of both chloroplast and mitochondrial genomes among all taxa were separated into three clades, as deduced from the phylogenetic trees. The first large clade consisted of 11 Prunus species, which were further clustered into three secondary clades. P. armeniaca, P. sibirica, P. mume, and P. salicina form the Prunus subclade. As is shown in Figure 8 , P. pedunculata was settled as a single monophyletic branch and group into the Prunus subgenus Amygdalus clade together with P. mira in both organelle genome trees. The subgenus Cerasus, including P. avium, P. conradinae, P. schneideriana, P. yedoensis, and P. kanzakura constitutes the third subclade of the Prunus genus. These results are consistent with the classification taxonomy.

Ten other ligneous species of Rosaceae, comprising species from the genera Eriobotrya, Sorbus, Torminalis, Pyrus, Photinia, and Malus, formed Clade 2, among which E. japonica and S. aucuparia formed distinct subclades, the rest of the species in the group clustered into another subclade in the mitogenome tree, whereas S. aucuparia was more related to Pyrus in the cp genome tree. Species from the genera Fragaria, Rosa and Potentilla comprised Clade 3. Most species in Clade 3 were herbaceous plants, in addition to three shrub species: R. chingii, R. rugosa, and R. chinensis. R. chingii and G. urbanum accessions diverged into two independent subclades with different genetic distance in the mitogenome tree and cp genome tree, though they are separated from the other relatives in this group.

Mitochondria have more complex genomes in plants than those of animals, owing to variation and repeated sequences. Due to the complexity of plant mitogenomes, extensive research has been focused on plastids, leaving multifarious mitogenomes to be investigated (Zardoya, 2020). Each of the 60 Rosaceae mitogenomes deposited in NCBI database was assembled into cyclic structure with remarkable variation in size, ranging from 270, 143bp (Rosa hybrid cultivar: OQ628291.1) to 535,727bp (P. mume: NC_065232.1). The genome size within the Prunus species varies slightly ( Supplementary Table S1 ), among which P. pedunculata mitogenome (405, 855bp) was the smallest. In this survey, the complete mitogenome of P. pedunculata was thoroughly characterized and assembled into a typical circular structure. The GC contents of Rosaceae mitogenomes are relatively static, ranging from 43.31% in R. chingii to 45.62% in P. avium (Sun et al., 2022). Interestingly, P. pedunculata exhibited the highest percentage of GC base (45.63%). The corresponding chloroplast genome sequence of P. pedunculata was found to be 157,830 bp long, with a GC content of 36.8%.

Plant mitogenomes contain highly conserved genes, relatively low gene densities, abundant non-coding sequences, and RNA editing occurrences (Niu et al., 2023). The high account of non-coding regions (84.82%) of the P. pedunculata mitogenome might be a consequence of sequence duplication during evolution (Ma et al., 2022). Mitochondrial DNA (mtDNA) encodes tRNAs, rRNAs, and a dynamic number of ribosomal proteins (Kubo and Newton, 2008). A total of 34 PCGs were identified in the P. pedunculata mitogenome, which is higher than that of most mitogenomes of Rosaceae species, aside from P. salicina, which has 39 PCGs in its mitogenome. The high content of PCGs indicated that fewer original mitochondrial genes had transfer into nuclear region during the evolutionary history of P. pedunculata. No gene loss events were observed in PCGs of the P. pedunculata mitogenome. Thirty-five genes were detected in most Prunus mitogenomes, except for rpl16 which was found only in the P.avium cultivar Glory, the P. avium cultivar Staccato, P. tenella, P. armeniaca, and P. salicina (Sun et al., 2022). Additionally, similar with O. sativa (Kazama and Toriyama), the P. pedunculata mitogenome contains two duplicated loci of the atp6 gene, one of them was assumed to be a pseudo-copy. Most Rosaceae species contain three ribosomal RNAs in their mitogenomes, except some species that possess four rRNAs (G. urbanum, M. domestica cultivar Gala, M. domestica cultivar Yantai fuji 8, M. sylvestris and P. serratifolia), and S. aucuparia has two rRNAs (Sun et al., 2022). In accordance with the known mitogenomes of Prunus, the mitogenome of P. pedunculata contains three rRNAs (rrn18, rrn26, and rrn5). In general, transfer RNAs in the mitogenome of angiosperm do not form a complete set (Michaud et al., 2011). Eight tRNAs (trnP-AGG, trnP-GGG, trnR-TCT, trnS-ACT, trnS-GGA, trnT-GGT, trnV-GAC, and trnI-GAT) were absent in the mitogenomes of Prunus, including P. pedunculata. Specifically, there are two copies of trnC-GCA in Prunus species and trnN-ATT appears solely in Prunus mitogenomes. Three copies of trnM-CAT, two copies of trnP-TGG, and trnY-GTA have been identified in most Prunus species. Moreover, trnI-TAT and trnL-TAA were absent from P. pedunculata, P. schneideriana, P. conradinae, and P. tenella, whereas trnK-TTT and trnG-GCC were present in these four species, differs greatly from other Prunus species (Sun et al., 2022). As for the intron-containing tRNAs (trnN-ATT, trnT-TGT and trnY-GTA) in P. pedunculata mitogenome, splicing their introns is a critical step of tRNA maturation (Hayne et al., 2022).

Repetitive sequences, especially abundant SSRs (Freitas et al., 2022), increase the mitogenome size, generation of multiple copies of genes, genome diversification, and structural variations (Fang et al., 2021). SSRs are characterized by cross-species transfer and high polymorphism, which allow them to be applied as molecular markers in phylogenetic analyses (Rossetto et al., 2002). In our study, 112 SSRs of P. pedunculata mitogenome distributed across different genomic regions: ORFs, introns, exons, or intergenic regions (IGS). The most abundant type of SSR in the mtDNA was mononucleotides (40.18%), consistent with a previous study (Kuntal and Sharma, 2011). However, the second-most common SSR in P. pedunculata mitogenome was tetranucleotide repeats (35.71%), rather than dinucleotide repeats. A near-universal A/T bias (Smith and Keeling, 2015) was also observed in P. pedunculata mitogenome, these SSRs were composed of motifs rich in A and T, which agrees with previous observations (Kuntal and Sharma, 2011), and corroborated the correlation between AT content of the complete mitogenome and SSRs (Freitas et al., 2022).

The non-synonymous and synonymous substitution ratios (Ka/Ks) have a complex relationship with biological functions of PCGs, and reflect gene selective pressures, thus contributing to our understanding of the evolutionary dynamics of PCGs among related species (Wang et al., 2010). In the current study, for the cytochrome c biogenesis genes ccmC and ccmFn that underwent positive selection (Ka/Ks>1), non-synonymous variations were favored because of functional adaptation or useful mutations. The average Ka/Ks value was the largest (0.62132) for cytochrome c biogenesis genes. Stabilizing selection (Ka/Ks>1) have been reported for other plants mitochondrial genes (Bi et al., 2020; Cheng et al., 2021; Yu et al., 2022). In contrast, genes that with Ka/Ks ratios below 1 may be highly conserved and have undergone negative selection during evolution (Greilhuber et al., 2012). Nonsynonymous variations are likely to be disadvantageous, therefore, genes with crucial functions subjected to stabilizing selection are partial to having lower Ka/Ks values. The lowest average values were observed for genes related to NADH dehydrogenase (0.18653). Because of constraint on protein function, synonymous substitutions are more common in most mitochondrial PCGs of P. pedunculata that dominated by purifying selection, which concordant with the fact that PCGs in the mitogenome are conserved across green plants (Mower et al., 2007; Cheng et al., 2021). With the Ka/Ks value approximately equal to 1, the protein function of nad2 might not constrain evolutionary changes. These results implied that the majority of genes were subjected to purifying selection. Thus, further research of the gene selection and evolution of Prunus species is still required.

Selection events for biased codon usage and recognition motifs for RNA editing sites occur in angiosperm mitogenomes (Liu and B., 2005). Deaminated cytosines in RNA transcripts become uridines at RNA editing sites in numerous mitochondrial DNA-encoded genes (Morley and Nielsen, 2017; Hao et al., 2021). According to the RSCU analysis, the most unique PCGs encoded by the P. pedunculata mitogenome contained the classical start codon ATG, in accordance with the allocation of amino acid compositions in other plant mitogenomes (Sloan et al., 2018b). Nevertheless, nad1 starts with ACG, presumably as a consequence of C-to-U editing at the second site. The A/T bias in the third codon position of PCGs with higher RSCU values (>1) indicates that A/T (U) stews across the mitochondrial genome of P. pedunculata, which may be a consequence of the A/T mutation bias common in plant mitogenomes (Smith and Keeling, 2015; Bi et al., 2020). CGA acts as a stop codon for ccmFc, atp9, and sdh4; however, analogous observations have rarely been reported. The cox1 gene of many lepidopteran species uses CGA as a start codon (Wu et al., 2020a; Jiang et al., 2021) rather than a stop codon. CAA acts as a stop codon for atp6 in the P. pedunculata mitogenome. Similar results have been reported for atp9 in wild carrots (Mandel et al., 2012).

It is believed that RNA editing events plays a pivotal role in the plant development regulation and stress resistance (Tang and Luo, 2018). In general, plant mitochondria contain more RNA editing sits than chloroplast (Small et al., 2020). We have predicted 262 RNA editing sites in P. pedunculata mitogenome, which is lower in comparison to other angiosperms (Giegé and Brennicke, 1999; Sloan et al., 2010; Richardson et al., 2013; Shan et al., 2023). Most RNA editing events in plant organelles arise from the site-specific C-to-U conversion (Small et al., 2020). Similar results were found in the current research. Particularly, 3 reverse changes (U-to-C conversion) were also been identified, which is more likely enriched in basal plants (Chen et al., 2023). In addition, G-to-A, G-to-U, A-to-G, A-to-U, U-to-A and C-to-A conversion types were also observed once and G-to-C changes were found twice among all RNA editing sites. RNA editing events may trigger the variation of amino acids and change the start or stop codons of PCGs. There are editing sites being transformed to stop codons in atp9 and nad7. ACG was used as a start codon in nad1 gene of P. pedunculata mitochondria directly, possibly due to a C-to-U conversion at the second site without editing in mitogenome (Zandueta-Criado and Bock, 2004). RNA editing events were reported existing in the initiation codon of nad1in other plants (Li et al., 2018; Fang et al., 2021; Wee et al., 2022). Consistent to most plants, there is few RNA editing events occur in rRNA, tRNA and introns (Giegé and Brennicke, 1999). In view of previous studies, RNA editing sites barely exist at the third codon position (Bi et al., 2020; Ma et al., 2022), largely because of the limitations of the methodology. Therefore, further research on the precisely predictive methods is still required.

Intergenomic gene transfer between organellar genomes has been an important phenomenon throughout the long-term evolution of higher plants (Sloan and Wu, 2014). As an important type of intracellular gene transfer (IGT) (Wang et al., 2018d) in the mitochondria, the mechanism of transfer events remains challenging because most transferred DNA sequences are located in non-coding regions. Plant mitogenomes maintain massive mitochondrial plastid fragments (MTPTs) (Lai et al., 2022). The proportion of plastid-derived fragments in mitogenomes varies from 0.44% (P. serratifolia) to 16.34% (R. chingii) in the family Rosaceae (Sun et al., 2022). The integration rate of chloroplast-derived mitogenomic sequences in P. pedunculata was 2.85%, which was much higher than that in other Prunus species (less than 1.16%). Similarly, the total length of the transferred sequences in the P. pedunculata mitogenome was 11,582 bp, which was the longest among the Prunus species. The species with the highest total length of plastid-derived sequences in the Rosaceae family is R. chingii (77,163 bp), which also comprises the highest proportion (16.34%) (Sun et al., 2022). Mostly, chloroplast-derived sequences are non-functional (Zhang et al., 2023). Remarkably, fragments from the exons of two genes (rps12, rps12-2) in the chloroplast genome of P. pedunculata changed to functional rps12 after transfer. Conversely, as a consequence of gene transfer events, the original functional gene sequence of ndhF was relocated to the exon region of nad5. The rrn16-rrn23 regions were used to construct transformation vectors, indicating that the rrn16-rrn23 chloroplast region might affect transformation efficiency (Lopez-Ochoa et al., 2015). The lost tRNAs in mitogenomes can be replaced by tRNAs inserted into other organelles (Sloan et al., 2010). The transfer of tRNA gene sequences from chloroplasts to mitochondrial genomes are common in plants (Bergthorsson et al., 2003). Inserted tRNAs accounted for 25% of the total transferred sequences in the P. pedunculata mitogenome, which contradicts the hypothesis that most transferred genes are tRNA genes (Chang et al., 2013).

Polyploidy of P. pedunculata leads to difficulties in the evolutionary analysis of genome sequences and phenotypes (Wang et al., 2020a). Phylogenetic inference of P. pedunculata has been controversial. According to previous publications, P. pedunculata was placed outside the monophyletic subgenus Amygdalus based on molecular data. In terms of morphological classification, it is clustered as a sister clade to peach (Yazbek and Oh, 2013). Previous studies have suggested that P. pedunculata is closely associated with P. tomentosa and P. triloba (Wang et al., 2018a; Duan et al., 2020; Wang et al., 2020a; Wang et al., 2021a). Nevertheless, there were no complete mitochondrial genome data for these two species available for phylogenetic analysis In this study, within the Prunus clade, P. pedunculata was divided into an independent branch, suggesting that the mitogenome divergence process produced distinctive maternal lines of P. pedunculata. The analysis provided strong evidence of the phylogenetic affinities of P. pedunculata from the perspective of 20 conserved PCGs in the mitogenome and provided a better understanding of the phylogenetic relationships among Rosaceae species. Nevertheless, additional investigations of more accessions of Prunus species and more molecular data are required to comprehensively understand the phylogeny of Prunus (Pervaiz et al., 2015).