In June 2023, live E. stephanianum leaves were collected from Moyun Mountain, Jinan City, Shandong Province (36°20′31.0308″N, 117°54′43.4772″E) (Gao et al., 2024). E. stephanianum was not an endangered or protected species and specific permission for the collection of E. stephanianum was not required.

All samples were thoroughly rinsed, cleaned using DEPC water, and subsequently stored at −80°C. Total DNA was extracted using TIANamp Genomic DNA Kit (Tiangen, Beijing, China). To obtain comprehensive data, we sequenced the mitogenome of E. stephanianum on both Nanopore GridION sequencing platform (Pugh, 2023) (Oxford Nanopore Technology, Oxford Science Park) and Illumina Novaseq 6,000 platform (Illumina, San Diego, United States), which enabling the construction of libraries and the generation of raw data (Nanopore raw data: 30.86 Gb, Illumina raw data: 24 Gb). The data reported in this paper have been deposited in the GenBank of NCBI (Sayers et al., 2025), under accession number PV575339 that were publicly accessible at https://www.ncbi.nlm.nih.gov/genbank/.

The E. stephanianum mitogenome was assembled using a comprehensive approach combining Illumina and Nanopore sequencing technologies. Initially, we employed Flye (Freire et al., 2022) software to conduct de novo assembly of long reads derived from E. stephanianum obtained through Oxford Nanopore sequencing. Subsequently, the BLASTn (Chen et al., 2015) was utilized to identify the draft mitogenome of E. stephanianum by comparing the assembled contigs. To facilitate this process, we created a database for the assembled sequences using makeblastdb and chose conserved mitochondrial genes from Arabidopsis thaliana (L.) Heynh. as our query sequence to pinpoint contigs that contain these conserved mitochondrial genes. The commonly parameters used for this assembly included “-evalue 1e-5 -outfmt 6 -max_hsps 10 -word_size 7 -task blastn-short”. Additionally, we conducted a hybrid assembly using Unicycler, intergrating both Illumina short reads and Nanopore long reads (Schafer et al., 2024). For the annotation of protein-coding genes (PCGs) in the mitogenome, we selected A. thaliana (NC_037304) and Liriodendron tulipifera (NC_021152.1) as reference genomes, using Geseq for the annotation process (Tillich et al., 2017). Annotation of tRNA and rRNA within the mitogenome was accomplished using tRNAscan-SE (Chan et al., 2021) and BLASTn (Chen et al., 2015), respectively. Manual correction of annotation errors in the mitogenome was performed using Apollo (Dunn et al., 2019).

Protein-coding gene (PCG) sequences were extracted from the genome using Phylosuite (Xiang et al., 2023), and the codon usage in mitochondrial PCGs was analyzed using Mega 7.0 (Kumar et al., 2016) and relative synonymous codon usage (RSCU) values were also calculated. To identify repeated sequences, including simple sequence repeats (SSRs), tandem repeats, and interspersed repeats, MISA (Beier et al., 2017), TRF (Benson, 1999), and REPuter (Kurtz et al., 2001) were employed. The results were visualized using the RCircos (Zhang et al., 2013) package.

Deepred-mt (Edera et al., 2021), a tool based on the convolutional neural network (CNN) model, was utilized for predicting C to U RNA editing sites. Mitochondrial protein-coding genes were extracted for prediction analysis, and only results with probability values exceeding 0.9 were selected for further consideration.

The chloroplast genome was assembled and annotated using GetOrganelle (Jin et al., 2020), and CPGAVAS2 (Shi et al., 2019), respectively. The BLASTn (Chen et al., 2015) program was utilized to compare two organelle genomes of E. stephanianum. In this process, the mitogenome was established as the database with makeblastdb, and the chloroplast genome was employed as the query sequence. All results were visualized using the RCircos (Zhang et al., 2013) package.

Related species of E. stephanianum were selected based on their genetic relationship, and their complete mitogenome sequences were downloaded from NCBI (https://www.ncbi.nlm.nih.gov) (Supplementary Table S3). PhyloSuite was utilized to extract shared mitochondrial genes across these species. Multiple sequences alignment was carried out using MAFFT (Katoh et al., 2019) with a bootstrap value of 1,000. IQ-TREE (Minh et al., 2020) was used for phylogenetic analysis. The resulting phylogenetic analysis was visualized using iTOL (Letunic and Bork, 2021).

Using the BLASTn, we identified conserved homologous sequences, which are referred to as co-linear blocks, with commonly parameters “-value 1e-5, -word_size 9, -gapopen 5, -gapextend 2, -reward 2, -penalty -3”. As a result, only co-linear blocks longer than 500 bp were considered. Based on sequence similarity. The mitochondrial genome of E. stephanianum was compared with multiple synteny regions from closely related species using MCscanX (Wang et al., 2024).

The mitochondrial genome of E. stephanianum exhibits a circular structure. We utilized Bandage software to visualize the draft mitochondrial genome assembled from long-read data. The final results are illustrated in Figure 1, which includes three nodes (Alternative conformations are illustrated in Supplementary Figures S1, S2). The assembly graph of the mitochondrial genome of E. stephanianum includes three distinct contigs/nodes (ctg). The ctg1 has a length of 251,257 base pairs; the ctg2 measures 107,879 base pairs in length; the ctg3 is significantly shorter, at 3,139 base pairs. Detailed information can be found in Supplementary Table S1.

The mitochondrial genome of E. stephanianum is primarily structured as a single circular molecule. After eliminating duplicate regions using Nanopore sequencing data, we identified a predominant circular contig with a total length of 365,414 bp and a GC content of 42.72% (Supplementary Table S2). Annotation of the mitochondrial genome yielded a total of 28 unique protein-coding genes, which include 24 core mitochondrial genes and four non-core genes, as well as 18 tRNA genes (of which 7 are multicopy) and 3 rRNA genes. The core genes comprise 5 ATP synthase genes (atp1, atp4, atp6, atp8, and atp9), 9 NADH dehydrogenase genes (nad1, nad2, nad3, nad4, nad4L, nad5, nad6, nad7, and nad9), 4 cytochrome c biogenesis genes (ccmB, ccmC, ccmFC, and ccmFN), 3 cytochrome c oxidase genes (cox1, cox2, and cox3), one membrane transport protein gene (mttB), one maturation enzyme gene (matR), and one coburol-ferredoxin reductase gene (cob). The non-core genes include one large ribosomal subunit gene (rpl10), two small ribosomal subunit genes (rps1 and rps3), and one succinate dehydrogenase gene (sdh4) (Figure 2; Table 1).

In our previous study, we present the first annotated chloroplast genome of E. stephanianum, describing its structure. The genome has a total length of 158,809  bp and contains 76 annotated protein-coding genes. Phylogenetic analysis confirms that E. stephanianum belongs to the Erodium genus in the family Geraniaceae (Gao et al., 2024) (Figure 3). Interestingly, several plastid genes were also annotated in the mtDNA, albeit mostly as fragments. These include rps3, atp1, trnC-GCA, et al. This finding suggests a notable sequence migration was observed between the chloroplast DNA (cpDNA) and mtDNA of E. stephanianum, which was accompanied by gene transfer, which will be discussed in detail below.

An analysis of codon preference among the 28 unique protein-coding genes (PCGs) of E. stephanianum mitochondria was conducted, with the usage of various codons for each amino acid summarized in Supplementary Table S3. Codons with a relative synonymous codon usage (RSCU) greater than one are considered to be favored by their respective amino acids. As illustrated in Figure 4, with the exception of the start codon AUG and the tryptophan codon (UGG), which both have an RSCU value of 1, there is a general preference for codon usage among the mitochondrial PCGs. For instance, alanine (Ala) exhibits a strong preference for the codon GCU, achieving the highest RSCU value of 1.6 among the mitochondrial PCGs, followed by histidine (His), which favors the codon CAU with an RSCU value of 1.54. Notably, the highest RSCU values for phenylalanine (Phe) and stop codons are both below 1.2, indicating a lack of strong codon usage preference.

Microsatellites, also known as simple repeat sequences (SSRs), were analyzed to determine the presence of repeat sequences in the mitogenome. An analysis of repetitive sequences in the E. stephanianum mitochondrial genome revealed a total of 103 simple sequence repeats (SSRs), with monomeric and dimeric forms accounting for 49.51% of the total SSRs (Supplementary Table S4). Thymidine (T) monomeric repeat sequences comprised 50.00% (16 out of 32) of the monomeric SSRs. Tandem repeats, also known as satellite DNA, are characterized by core repeating units of approximately 7–200 bases that are repeated in succession. These sequences are widely distributed in the genomes of eukaryotes and prokaryotes. Within the E. stephanianum mitochondrial genome, 52 tandem repeat sequences were identified, with a similarity greater than 76% and lengths ranging from 9 to 43 bp. Additionally, scattered repeats within the Es mitochondrial genome were detected, yielding a total of 604 repeat sequences with lengths of 30 bp or greater. This included 291 palindromic repeats, 313 forward repeats, and no instances of reverse repeats or complementary repeats. The longest palindromic repeat measured 704 bp, while the longest forward repeat extended to 3,139 bp. A bar char in Figure 5 illustrates the different types of repeat sequences.

We observed significant sequence transfers from the chloroplast genome to the mitogenome in E. stephanianum. Based on sequence similarity analysis, a total of 55 segments were identified as homologous fragments between the mitochondrial and chloroplast genomes, with an aggregate length of 58,305 bp, which accounts for 15.96% of the total mitochondrial genome length (Figure 6). The longest segment, MTPT22, measures 5,586 bp. Annotation of these homologous sequences revealed the presence of 33 complete genes within the 55 homologous fragments, including 28 protein-coding genes (atpB, atpE, atpF, atpH, atpI, ccsA, ndhA, ndhI, petA, petB, petD, petG, petL, petN, psaJ, psbB, psbF, psbJ, psbL, psbN, psbT, rpl33, rpoA, rpoC2, rps18, rps2, rps7, ycf2) and five tRNA genes (trnC-GCA, trnD-GUC, trnM-CAU, trnN-GUU, trnW-CCA) (Supplementary Table S5).

To investigate the synteny relationship between E. stephanianum and closely related species, we utilized MCscanX to generate multiple synteny plots based on the sequence similarity. Figure 7 illustrates that the co-linear blocks exhibit varying arrangements across individual mitochondrial genomes, the red arched regions represent inverted sequences, while the gray areas indicate regions with high sequence homology. Although a substantial number of blocks were detected between E. stephanianum and Geranium maderense and Citrus unshiu, these co-linear blocks appeared to be shorter in length. Additionally, several blank regions were identified, corresponding to sequences unique to E. stephanianum and lacking homology with other species. These results suggest inconsistent collinear block arrangements among the five mitochondrial genomes, indicating that the Es mitochondrial genome has undergone rearrangement compared to its relatives (Supplementary Table S6), more extensive results have been compiled in Supplementary Table S8.

Phylogenetic trees were constructed for 32 species across four orders of angiosperms based on the DNA sequences of 24 conserved protein-coding genes (PCGs). The specific mitochondrial genome sequences of the studied plant species can be found in Appendix 1. The protein-coding genes examined included atp1, atp4, atp6, atp8, atp9, ccmB, ccmC, ccmFC, ccmFN, cob, cox1, cox2, cox3, matR, mttB, nad1, nad2, nad4, nad4L, nad5, nad6, nad9, and rps3. The mitochondrial genomes of two species from the family Zingiberaceae were designated as the outgroup (Supplementary Table S7). The phylogenetic topology based on mitochondrial DNA aligns with the most recent classification of the Angiosperm Phylogeny Group (APG). The species E. stephanianum belongs to the order Geraniales and the family Geraniaceae, clustering with the Geraniaceae species Geranium maderense, illustrating its close evolutionary relationship with Geranium maderense (NC_027000.1) (Figure 8).

RNA editing events were identified in 28 unique protein-coding genes (PCGs) from the mitochondrial genome of E. stephanianum using Deepred-mt. The cutoff value was set at 0.9. Under this threshold, a total of 109 potential RNA editing sites were detected across the 28 mitochondrial PCGs, all characterized by the conversion of cytosine (C) to uracil (U). The ccmFN gene exhibited the highest number of RNA editing sites, with 27 identified, followed by the mttB gene, which demonstrated 15 RNA editing events (Figure 9).