Assembly and comparative analysis of the complete mitochondrial genome of hexaploid oat (Avena sativa) Authors: Shiqiang Zhao, Ruohong Li, Zhicheng Jia, Tianqi Zhu, Chunjiao Mi, Juan Wang, Zhijia Lv, Peisheng Mao Published: 2026 ABSTRACT AbstractBackgroundCommon oat (Avena sativa) is a globally important hexaploid cereal crop with a large and complex genome. While its nuclear and chloroplast genomes have been extensively characterized, the mitochondrial genome of hexaploid oat has remained unresolved, limiting our understanding of organellar evolution and mitochondrial function in this polyploid species.ResultsWe report the first complete assembly and comprehensive analysis of the A. sativa mitochondrial genome, which comprises two circular chromosomes totaling 634,262 bp with a GC content of 44.11%. The mitogenome encodes 78 genes, including 37 protein-coding genes (PCGs), 27 tRNAs, 12 rRNAs, and 2 pseudogenes. Extensive repeat sequences were identified, providing a structural basis for recombination and genome plasticity. Codon usage analysis revealed a strong AT bias and distinct codon preferences, while Ka/Ks analysis indicated predominant purifying selection on core OXPHOS genes, with possible signals of positive selection in mttB and nad7. A total of 463 RNA editing sites were predicted, predominantly enhancing protein hydrophobicity and potentially optimizing mitochondrial function. Comparative analyses revealed high collinearity with the diploid oat A. longiglumis and frequent transfers of chloroplast-to-mitochondrial DNA. Phylogenetic reconstruction based on 32 conserved mitochondrial genes robustly resolved relationships within Poaceae and its relatives.ConclusionsThe first complete mitogenome of hexaploid oat provides new insights into mitochondrial genome structure, evolution, and organellar genome dynamics in polyploid species, and constitutes a valuable genomic resource for future studies of mitochondrial biology, comparative genomics, and marker development in oats.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12870-026-09191-5.Keywords: Avena sativa, Mitochondrial genome assembly, Mitogenome, RNA editing, Poaceae Background Common oat (Avena sativa, 2n = 6x = 42, AACCDD) ranks seventh in global cereal production and plays a crucial role in human nutrition and livestock feed due to its strong adaptability to diverse and marginal environments [1]. Belonging to the Avena genus within the Poaceae family [2], common oat exhibits a complex evolutionary history involving hybridization and polyploidization, giving rise to diploid, tetraploid, and hexaploid species with frequent intergenomic translocations [3, 4]. This results in a large genome (~ 11 Gb) characterized by high ploidy and substantial repetitive sequences (86.95%) [5], which have historically hindered molecular breeding and genomic research. Recent advances in high-throughput sequencing technology have enabled the assembly of multiple hexaploid oat genomes, including OT3098, Sanfensan [5] and Sang [6]. Gap-free genome assemblies, pangenome and pantranscriptome resources have also been developed [7–9], along with a multi-omics database [10], providing valuable resources for functional genomics and breeding research. Plastids and mitochondria are the only organelles with endosymbiotic genomes, and recent sequencing advances have greatly expanded organellar genome resources, revealing divergent evolutionary paths [11]. While chloroplast genomes of various Avena species have been extensively studied [12–14], only the diploid oat (A. longiglumis) mitochondrial genome has been assembled [15], leaving the hexaploid oat mitogenome unresolved. Understanding the mitochondrial genome is crucial because mitochondria play central roles in cellular energy production, signal transduction, and programmed cell death [16, 17]. The mitochondrial genome encodes core subunits of the oxidative phosphorylation (OXPHOS) machinery, including 18 subunits in Arabidopsis thaliana [18, 19], which are conserved across plant species and essential for cellular energy metabolism [20, 21]. Given the complexity of the hexaploid oat genome, assembling and characterizing its mitochondrial genome is essential for understanding the genetic basis of mitochondrial function in this important crop. Plant mitochondrial genomes exhibit extensive structural diversity, unlike the highly conserved plastid genomes and bilaterian mitogenomes [21, 22]. In most land plants, mitochondrial genomes display complex and dynamic molecular forms [23]. For instance, the mitochondrial genome of Angelica biserrata displays a multi-branched structure, composed of six circular contigs [24]. In contrast, the mitogenomes of two closely related species, Mikania micrantha and M. cordata, exhibit distinct differences in their structural configurations. The complete mitogenome of M. micrantha is a double-stranded DNA molecule, while M. cordata features a branching structure consisting of two small circular molecules and six linear molecules [25]. Plant mitochondrial genomes can generate various interchangeable isomers through repeat-mediated recombination, leading to stoichiometric variation among different structural forms [26]. Therefore, elucidating these dynamic structural variations, rather than adhering to the traditional single “master circle” model, is essential for unraveling the structural and functional complexity of plant mitochondrial genomes. RNA editing refers to the specific modification of RNA sequences at particular sites [27]. In plant organelles, RNA editing processes primarily convert C to U (or U to C), playing a pivotal role in mitochondrial gene expression and function [28, 29]. Many studies support the notion that each mitochondrion within a cell contains mtDNA [30]. For instance, human fibroblasts are estimated to have ≥ 1.4 copies of mtDNA per mitochondrion [31]. In contrast, plant cells maintain limited mtDNA with genomic information shared among multiple mitochondria. The restricted RNA-editing capacity in plants may represent an evolutionary bottleneck that limits mtDNA levels, which could also explain the differences in genome copy number between mitochondria and chloroplasts [32–34]. RNA editing of mitochondrial genes is believed to regulate plant cytoplasmic inheritance-related traits [35], making it an important mechanism for understanding oat mitogenome function. Plant mtDNA contains substantial noncoding and repeated sequences that drive frequent homologous recombination and rearrangement events [36]. Abnormal recombination can produce novel chimeric open reading frames (ORFs) causing cytoplasmic male sterility (CMS), with CMS chimeric genes often competing with or inhibiting the expression of mitochondrial functional genes [37]. Several CMS lines have been reported in rice (Oryza sativa) [38], wheat (Triticum aestivum) [39] and maize (Zea mays) [40], highlighting the potential of mitochondrial genomes in breeding and producing F1 hybrid seeds. Mitochondrial genomes encode a limited number of mitochondrial proteins, with most genes associated with subunits of the electron transport chain, making them indispensable for the proper functioning of the mitochondrial oxidative phosphorylation system [37]. Damage or depletion of mtDNA inevitably leads to disruptions in this system. Given the recent breakthroughs in oat molecular breeding, particularly in genetic transformation [41] and gene editing [42], assembling the complete mitochondrial genome of hexaploid oat has become crucial for advancing intelligent breeding programs. In this study, we assembled and annotated the first complete mitochondrial genome of hexaploid oat using hybrid sequencing approaches. Comprehensive analyses were conducted for the genome structure, gene content, repeat distribution, RNA editing patterns, codon usage, and selective pressure. Through comparative genomics, we investigated evolutionary relationships within Poaceae and characterized mitochondrial-chloroplast DNA transfers. Our findings provide fundamental genomic resources for understanding oat mitochondrial biology and its evolutionary context within the Poaceae family. Materials and methods Plant material and DNA sequencing Plants of oat (Avena sativa L., cultivar ‘Nelson’; Accession CN 54996, Plant Gene Resources of Canada) were cultivated at the China Agricultural University Greenhouse, Beijing, China (40°03′N, 116°28′E). Plant specimens were identified by Shiqiang Zhao and deposited in the China Agricultural University Herbarium (Index Herbariorum code: BAU, voucher specimen number: BAU-00002013). Young, healthy leaves were collected, immediately frozen in liquid nitrogen, and stored at -80 °C for DNA extraction and sequencing. Total cellular DNA was sequenced using a hybrid approach combining Illumina NovaSeq 6000 (paired-end libraries with 350 bp insert size) and Oxford Nanopore Technologies (ONT) long-read sequencing. Low-quality reads were filtered using fastp (version 0.23.4, https://github.com/OpenGene/fastp) for Illumina data and Filtlong (v0.2.1, https://github.com/rrwick/Filtlong) for ONT data. Assembly and annotation of mitochondrial genome Raw long-read data were aligned against a reference set of plant mitochondrial core genes (https://github.com/xul962464/plant_mt_ref_gene) using minimap2 (version 2.30) [43] to obtain the complete mitochondrial sequencing data. The raw long reads were error-corrected using Canu [44], and Illumina reads were aligned to the corrected sequences using Bowtie2 (version 2.3.5.1) [45]. The aligned Illumina data and corrected long reads were assembled using Unicycler (version 0.4.8) [46] with default parameters, and assembly results were visualized and manually refined using Bandage (version 0.8.1) [47]. To resolve complex branching structures and sub-ring configurations, corrected long reads were mapped to assembled contigs using minimap2, and branch orientations were manually determined to obtain the final mitochondrial genome assembly. Protein-coding genes (PCGs) and ribosomal RNAs (rRNAs) were annotated by BLAST (version 2.6, https://blast.ncbi.nlm.nih.gov/Blast.cgi) comparison with published plant mitochondrial sequences. Transfer RNAs (tRNAs) were predicted using tRNAscan-SE (http://lowelab.ucsc.edu/tRNAscan-SE) with default parameters (source: Organellar; search mode: Default). The secondary cloverleaf structures of all predicted tRNAs were generated by tRNAscan-SE and visualized using the tool’s built-in structure output [48]. ORFs were identified using ORF Finder (https://www.ncbi.nlm.nih.gov/orffinder/) with a minimum length of 102 bp, excluding redundant sequences and those overlapping with known genes. ORFs longer than 300 bp were aligned to the NCBI non-redundant (nr) protein database. All annotation results were manually checked and corrected to obtain the final annotation. Genome and gene structures were visualized using OGDRAW (https://chlorobox.mpimp-golm.mpg.de/OGDraw.html). Validation of genome assembly completeness and integrity. To assess the completeness, uniformity, and reliability of the final mitochondrial genome assembly, both the Illumina short reads and the Oxford Nanopore long reads were independently aligned back to the assembled mitogenome using minimap2 (v2.30) with the sr preset for Illumina paired-end reads and the map-ont preset for ONT reads. Alignments were sorted and indexed with SAMtools (v1.13), and per-base sequencing depth was calculated using samtools depth -a. Mapping statistics were summarized with samtools flagstat. Coverage plots were generated in Python using matplotlib. Pre-assembly contamination screening was performed by the sequencing provider (Nanjing Genepioneer Technology Co., Ltd.) by aligning a random subsample of 2,000 Illumina reads against the NCBI nucleotide (NT) database using BLAST (Table S4). Analysis of codon usage bias, repeat elements and prediction of RNA editing sites Relative synonymous codon usage (RSCU) values, which reflect the unequal usage of synonymous codons and are influenced by natural selection, mutation bias, and genetic drift, were calculated using a custom Perl script developed by Nanjing Genepioneer Technology Co., Ltd. (Nanjing, China) based on the standard genetic code (NCBI translation Table 1), which is applicable to plant mitochondrial genomes. The script also filtered unique CDS sequences for analysis. To evaluate codon usage bias from an evolutionary perspective, RSCU values were calculated for A. sativa and seven other representative Poaceae species (A. longiglumis, T. aestivum, H. vulgare, L. perenne, O. sativa, Z. mays, and E. indica) using a custom Python script. The comparative RSCU profiles were visualized using the Codon Usage Visualization online tool (pcg-lab.shinyapps.io/RSCU-Plot/) [49–52]. An additional RSCU heatmap was generated using the R package pheatmap (v1.0.12) to display cross-species codon usage patterns in matrix format. Table 1Statistics of the mitochondrial genome of A. sativaNCBI accession numberContigsTypeLengthGC content (%)PV640483Chromosome 1circular559,944 bp44.00PV640484Chromosome 2circular74,318 bp44.92PV640483-PV640484Chromosome 1–2-634,262 bp44.11 Statistics of the mitochondrial genome of A. sativa Simple sequence repeats (SSRs) were identified using MISA (version 1.0, https://webblast.ipk-gatersleben.de/misa/) [53] with minimum repeat thresholds of 10, 5, 4, 3, 3, and 3 for mono-, di-, tri-, tetra-, penta-, and hexanucleotide motifs, respectively. Tandem repeats were detected using Tandem Repeats Finder (version 4.09, https://tandem.bu.edu/trf/trf.html) [54] with alignment parameters of 2, 7, and 7 for matches, mismatches, and indels, respectively, and a minimum alignment score of 80 and maximum period size of 2000. Dispersed repeats were identified using BLASTn (version 2.10.1, https://blast.ncbi.nlm.nih.gov/Blast.cgi) with word_size = 7 and E-value=1e-5, excluding redundant sequences and tandem repeats. Repeat distributions were visualized using Circos (version 0.69-5, http://circos.ca/). To further characterize SSR distribution across functional genomic compartments, each SSR was classified as genic (located within annotated CDS, rRNA, or tRNA regions) or intergenic (IGS) based on coordinate overlap with the GenBank annotation using a custom Python script with Biopython (v1.79). To compare SSR characteristics across Poaceae, the same MISA analysis with identical parameters was applied to the mitochondrial genomes of seven additional species (A. longiglumis, T. aestivum, H. vulgare, L. perenne, O. sativa, Z. mays, and E. indica). RNA editing sites were predicted using the Plant Predictive RNA Editor for mitochondrial genes (PREP-Mt, http://prep.unl.edu/) [55]. To experimentally validate the predicted editing sites, ccmB and nad4L were amplified by RT-PCR from young leaves of A. sativa ‘Nelson’ using gene-specific primers (Table S1), with cDNA generated from DNase I-treated total RNA by random N9 primers (TransScript AT311, TransGen Biotech) and a parallel gDNA control processed identically. Bidirectional Sanger sequencing of cDNA and gDNA amplicons was performed, and C-to-U editing events were identified by aligning cDNA and gDNA chromatograms to the assembled CDS in SnapGene, scoring positions showing a T peak in cDNA with a corresponding C peak in gDNA. Ka/Ks and nucleotide diversity analysis We performed pairwise comparisons between A. sativa, A. longiglumis, T. aestivum, O. sativa subsp. japonica, and O. sativa subsp. indica, extracting homologous gene pairs for analysis. These gene pairs were aligned using MAFFT (version 7.310, https://mafft.cbrc.jp/alignment/software/) [56]. After alignment, the Ka (nonsynonymous substitution rate) and Ks (synonymous substitution rate) values for each gene pair were calculated using KaKs_Calculator (version 2.0, https://sourceforge.net/projects/kakscalculator2/) [57], with the MLWL method selected for the calculation. Finally, the Ka/Ks ratios for each gene pair were visualized using boxplots. We performed global alignments of homologous gene sequences from the above different species using MAFFT (version 7.427, https://mafft.cbrc.jp/alignment/software/) with auto mode. The pi values for each gene were calculated using DnaSP (version 5.0, http://www.ub.edu/dnasp). Comparative genomics and phylogenetic analysis Mitochondrial genome collinearity was assessed using NUCmer (version 4.0.0beta2, https://github.com/mummer4/mummer/releases) with the maxmatch parameter, and results were visualized as dot plots. Whole-mitogenome pairwise alignments between A. sativa and A. longiglumis (GenBank: OQ450323.1) were performed using MUMmer4 (v4.0) with the nucmer and dnadiff modules to compute alignment coverage, average sequence identity, SNP counts, and structural variant statistics. Gene content variation was analyzed for 43 mitochondrial genes across five species. Gene annotations were obtained from NCBI GenBank for reference species, while A. sativa genes were annotated as described in Sect. 2.2. Genes were classified as present (1), absent (0), or pseudogenes (-1) based on the integrity of ORFs and functional annotation. Pseudogenes were initially identified during the automated annotation pipeline when open reading frames were found to be disrupted by internal stop codons or substantial truncation relative to functional orthologs, and were subsequently verified by manual BLAST alignment against homologous sequences in closely related species. We obtained 33 Poaceae complete mitochondrial genome sequences from five different subfamilies (Pooideae, Bambusoideae, Oryzoideae, Chloridoideae, and Panicoideae) from the National Center for Biotechnology Information (NCBI) database, along with four additional species, Cyperus esculentus (Cyperaceae), Carex breviculmis (Cyperaceae), Cyperus stoloniferus (Cyperaceae) and Suaeda glauca (Amaranthaceae) as outgroups (Table S2). These sequences, along with the newly assembled A. sativa genome, were used for phylogenetic analysis. A total of 32 shared PCGs across the analyzed species were identified and extracted using PhyloSuite software (version 1.2.1). These 32 conserved PCGs were identified: atp1, atp4, atp6, atp8, atp9, ccmB, ccmC, ccmFc, ccmFn, cob, cox1, cox2, cox3, matR, mttB, nad1, nad2, nad3, nad4, nad4L, nad5, nad6, nad7, nad9, rpl16, rps1, rps12, rps13, rps2, rps3, rps4 and rps7. Coding sequences were aligned using MAFFT (version 7.427), concatenated and trimmed using trimAl (version 1.4.rev15) with gt = 0.7, and the best-fit substitution model (GTR) was selected using jModelTest (version 2.1.10, https://github.com/ddarriba/jmodeltest2). A maximum-likelihood phylogenetic tree was constructed using RAxML (version 8.2.10, https://cme.h-its.org/exelixis/software.html) with the GTRGAMMA model and 1,000 bootstrap replicates. Chloroplast-to-mitochondrion and mitochondrion-to-nucleus DNA transfer Homologous sequences between the A. sativa chloroplast and mitochondrial genomes were identified using BLAST (version 2.6, https://blast.ncbi.nlm.nih.gov/Blast.cgi) with an E-value threshold of 1e-5 and minimum sequence similarity of 70%. Homologous regions and gene distributions were visualized using Circos (version 0.69-5, http://circos.ca/). Nuclear mitochondrial DNA segments (NUMTs) were identified by BLAST searches between the mitochondrial genome and the nuclear genome reference [6] with an E-value of 1e-5. The chromosomal and length distributions were analyzed and visualized. Results Mitochondrial gene organization and features The complete mitogenome of A. sativa exhibited a complex, multi-branched conformation consisting of two circular chromosomes (Fig. 1): chromosome 1 (559,944 bp) and chromosome 2 (74,318 bp), with a total length of 634,262 bp and a GC content of 44.11% (Table 1). During assembly, the mitochondrial genome was initially fragmented into multiple discontinuous segments. These segments were subsequently merged via overlapping regions (Fig. 1A and B), yielding two complete circular chromosomes (Fig. 1C). The final assembly accurately reconstructed the full mitochondrial genome structure, facilitating downstream functional analyses. The complete mitochondrial genome sequences have been deposited in GenBank under the accession numbers listed in Table 1. Assembly graph and structural resolution of the A. sativa mitochondrial genome. (A) Initial assembly graph visualized in Bandage, exhibiting a complex, multi-branched conformation composed of multiple segments. (B) Progressive resolution of the graph structure through overlapping regions. (C) Final assembly consists of two independent circular chromosomes Annotation of the mitochondrial genome revealed 78 genes, comprising 37 PCGs, 27 tRNA genes, 12 rRNA genes, and 2 pseudogenes (Fig. 2; Table 2). Among the 37 PCGs, atp6 and mttB each had two copies. Of these, 26 were core genes and 11 were non-core genes. The core genes were functionally classified into seven groups: ATP synthase, cytochrome c biogenesis, ubiquinol cytochrome c reductase, cytochrome c oxidase, maturases, transport membrane proteins, and NADH dehydrogenase. The non-core genes included two large ribosomal subunit genes (rpl5, rpl16), eight small ribosomal subunit genes (rps1, rps2, rps3, rps4, rps7, rps12, rps13, rps14), and one succinate dehydrogenase gene (sdh4). Transfer RNA prediction identified 27 tRNA genes recognizing all standard amino acids, with multiple copies observed for trnC, trnD, trnE, trnM, trnP, and trnS. All tRNAs displayed the characteristic cloverleaf secondary structure (Fig. S1), indicating proper folding and functionality. Circular gene maps of the A. sativa mitochondrial genome. (A) Chromosome 1 (559,944 bp). (B) Chromosome 2 (74,318 bp). Genes shown on the outside of the circle are transcribed clockwise, whereas those on the inside are transcribed counterclockwise. The inner gray ring illustrates the GC content variation. Functional categories of genes are color-coded as shown in the legend Table 2Genes identified in the A. sativa mitochondrial genomeGroup of genesGene namesATP synthase (Complex Ⅴ)atp1, atp4, atp6(2), atp8, atp9Cytochrome c biogenesisccmB, ccmC, ccmFc*, ccmFnUbiquinol cytochrome c reductase (Complex Ⅲ) cob Cytochrome c oxidase (Complex Ⅳ)#cox1, cox1, cox2*, cox3Maturases matR Transport membrane proteinmttB(2)NADH dehydrogenase (Complex Ⅰ)nad1****, nad2****, nad3, nad4***, nad4L, nad5****, nad6, nad7****, nad9Ribosomal proteins (LSU)rpl5, rpl16Ribosomal proteins (SSU)#rps19, rps1, rps2, rps3*, rps4, rps7, rps12, rps13, rps14Succinate dehydrogenase sdh4 Ribosomal RNAsrrn5(4), rrn18(4), rrn26(4)Transfer RNAstrnC-GCA(3), trnD-GTC(2), trnE-TTC(2), trnF-GAA, trnH-GTG, trnK-TTT, trnM-CAT(5), trnN-GTT, trnP-TGG(3), trnQ-TTG, trnR-TCT, trnS-GCT, trnS-GGA, trnS-TGA(2), trnW-CCA, trnY-GTA*: intron number; The “#” prefix denotes pseudogene status (e.g., #cox1, #rps19); Gene(2): Number of copies of multi-copy genes Genes identified in the A. sativa mitochondrial genome *: intron number; The “#” prefix denotes pseudogene status (e.g., #cox1, #rps19); Gene(2): Number of copies of multi-copy genes Validation of assembly completeness by read mapping To evaluate the completeness and uniformity of the final assembly, both short-read and long-read sequencing data were aligned back to the assembled mitochondrial genome. Mean per-base sequencing depths were 181× for Illumina short reads and 223× for ONT long reads across the 634,262 bp assembly (Fig. S2), both well above the standard thresholds used for high-confidence plant mitogenome assemblies. Only 8 and 2 positions, respectively, had zero coverage in the Illumina and ONT datasets (< 0.002% of the assembly), indicating that essentially all regions of the assembly — including non-coding intergenic spacers — were supported by independent sequencing evidence. Mean coverage was consistent between the two assembled chromosomes (Chr1: 180×/226×; Chr2: 193×/199× for Illumina/ONT), providing additional support for the authenticity of both chromosomes. Localized coverage peaks observed in both sequencing datasets predominantly co-localized with annotated rRNA gene clusters (rrn5, rrn18, rrn26; orange shading in Fig. S2; Table S3), with additional minor peaks corresponding to repetitive sequences shared between the two chromosomes, consistent with the known multi-copy structural features of plant mitochondrial genomes. The uniform read depth observed across intergenic regions, together with the successful mapping of long reads spanning chromosomal junctions, provides direct evidence that the reference-gene-guided assembly strategy did not introduce bias against divergent, non-coding, or structurally complex regions of the mitogenome. Furthermore, pre-assembly contamination screening confirmed that all aligned reads matched Poaceae species, with no detectable non-plant contamination (Table S4). Repeat sequence analysis We identified SSRs, tandem repeats, and dispersed repeats within the A. sativa mitochondrial genome (Fig. 3). This analysis revealed 138 SSRs, 19 tandem repeats, and 383 dispersed repeats (Table S5). The distribution of these repetitive sequences was uneven across the two chromosomes. For SSRs, chromosome 1 harbored the majority (127), whereas chromosome 2 contained only 11. Similarly, of the 19 tandem repeats, 15 were located on chromosome 1 and 4 on chromosome 2. On chromosome 1, SSR composition was dominated by tetrameric motifs (39.37%), followed by monomeric (22.83%) and dimeric (16.54%) repeats. In contrast, chromosome 2 exhibited a higher proportion of monomeric and dimeric repeats (27.27% each). Overall, monomeric and dimeric motifs accounted for 23.19% and 17.39% of the total SSRs, respectively (Table S6). Classification of the 138 SSRs by genomic location revealed that the vast majority (132, 95.7%) were located in intergenic spacer (IGS) regions, whereas only 6 (4.3%) fell within genic regions (CDS, rRNA, or tRNA) (Fig. S3). This IGS enrichment was consistent across all motif types, with tetranucleotide repeats showing the most pronounced bias (52 of 53 in IGS). Distribution of repetitive sequences in the A. sativa mitochondrial genome. The outermost circle represents the two mitochondrial chromosomes drawn to scale, with chromosome 1 in cyan and chromosome 2 in purple. Moving inward, blue and red tick marks indicate the positions of simple sequence repeats (SSRs) and tandem repeats, respectively. The inner ribbons (chords) connect pairs of dispersed repeats identified across the genome To place these SSR characteristics in a broader evolutionary context, we performed the same MISA analysis on seven additional Poaceae mitochondrial genomes using identical parameters (Fig. S4; Table S7). The total number of SSRs varied from 103 (A. longiglumis) to 248 (Z. mays), with tetranucleotide repeats being the dominant motif type across all eight species (34–42%). Notably, Z. mays exhibited a markedly higher SSR content (435.4 per Mb), driven primarily by expansions of pentanucleotide and hexanucleotide repeats. SSR density in the remaining species ranged from 187.8 to 268.9 per Mb. For dispersed repeats, chromosome 1 contained 338 sequences (165 forward and 173 palindromic), whereas chromosome 2 had only 7 (4 forward and 3 palindromic). Additionally, 38 inter-chromosomal dispersed repeats (21 forward and 17 palindromic) were identified between the two chromosomes. In total, 383 dispersed repeats were identified across the mitochondrial genome (Table S8). Notably, the length distribution was highly skewed: most repeats were short (< 100 bp, 82%), whereas ten large repeat pairs (≥ 10 kb, up to 44,994 bp) with near-perfect sequence identity (≥ 99.98%) were identified exclusively within chromosome 1 (Table S9). In contrast, all 38 inter-chromosomal repeats were shorter than 500 bp. Analysis of codon usage Codon usage analysis of the 37 PCGs identified a total of 10,475 sense codons in the mitochondrial genome of A. sativa (excluding 35 stop codons). These codons encoded all 20 standard amino acids. Among them, leucine (Leu) was the most abundant amino acid (1,131 codons, 10.80%), followed by serine (Ser) (934 codons, 8.92%), whereas cysteine (Cys) was the least abundant (152 codons, 1.45%) (Table S10). Overall, RSCU analysis indicated a biased codon usage pattern, with A/T-ending codons generally preferred, which is consistent with the AT-rich composition of the mitochondrial genome. A comparative RSCU heatmap across the eight species is provided in Fig. S5. To further examine whether this codon usage pattern is conserved across grasses, we compared the RSCU profiles of A. sativa with those of seven additional representative Poaceae species (Fig. 4). The overall RSCU patterns were broadly similar among the eight mitochondrial genomes, revealing a highly conserved codon usage bias within the Poaceae family. While limited variation was observed for a subset of codons, the predominant synonymous codon preferences remained remarkably consistent across species. These results indicate that the mitochondrial genome of A. sativa shares stable evolutionary constraints on protein-coding sequences with other grass species. Comparative analysis of relative synonymous codon usage (RSCU) among eight Poaceae species. Each panel shows the stacked RSCU values for 20 amino acids, with individual codons distinguished by color Analysis of Ka/Ks ratios To assess evolutionary pressure on mitochondrial PCGs, we calculated Ka/Ks ratios across pairwise comparisons of five species. The majority of genes exhibited Ka/Ks ratios below 1 (Fig. 5A), with ribosomal protein genes and Complex IV subunits showing particularly low values (typically < 0.5), indicating strong evolutionary constraints. Notable exceptions were mttB and nad7, which showed Ka/Ks ratios exceeding 1 in multiple comparisons, suggesting possible positive selection or relaxed selective constraints. Ka/Ks ratios for 37 shared PCGs across five species. (A) Boxplots showing the distribution of Ka/Ks ratios for each gene across all pairwise species comparisons. The red dashed line indicates Ka/Ks = 1. (B) Bar plots showing Ka/Ks ratios for each gene in individual species pair comparisons. Different colors represent different species pairs, as indicated in the legend Pairwise comparisons revealed substantial lineage-specific variation. Most notably, atp4 exhibited an exceptionally high Ka/Ks ratio (> 4.0) in the comparison between A. sativa and A. longiglumis. Similarly, nad5 showed elevated Ka/Ks values (> 2.0) in the comparison between A. sativa and O. sativa subsp. indica, while remaining relatively conserved in other comparisons. Additional variation was observed in ribosomal and cytochrome c maturation genes across different species pairs (Fig. 5B). The prediction of RNA editing A total of 463 RNA editing sites were identified across 31 mitochondrial PCGs using PREP-Mt (Fig. 6). The distribution of editing sites was uneven among genes. The cytochrome c maturation (ccm) genes exhibited the highest editing frequency, with ccmFn containing 36 sites, followed by ccmC (35 sites) and ccmB (33 sites). In contrast, genes such as rps7 contained only a single editing site. Distribution of RNA editing sites across 31 PCGs in the A. sativa mitochondrial genome. Each bar represents the total number of predicted editing sites for a given gene. Colors indicate the type of amino acid hydropathy change resulting from C-to-U editing: hydrophilic to hydrophobic (red), hydrophobic to hydrophobic (orange), hydrophilic to hydrophilic (light blue), hydrophobic to hydrophilic (dark blue), and hydrophilic to stop codon (grey). Genes are ordered by decreasing total number of editing sites Analysis of amino acid substitutions revealed that RNA editing substantially impacted protein properties (Table S11). The majority of editing events (48.6%, 225 sites) caused a shift from hydrophilic to hydrophobic amino acids (e.g., Ser to Leu). Other edits maintained hydrophobicity (28.9%) or hydrophilicity (11.9%), while a smaller proportion (9.9%) converted hydrophobic residues to hydrophilic ones. Notably, 3 sites (0.65%) introduced premature stop codons (e.g., Gln to Stop). The most frequent specific substitutions were Serine to Leucine (62 occurrences), followed by Serine to Phenylalanine (41) and Proline to Leucine (40). Overall, these editing events resulted in a net increase in the hydrophobicity of mitochondrial proteins, with 225 hydrophilic-to-hydrophobic conversions compared to 46 hydrophobic-to-hydrophilic changes. To experimentally assess the reliability of these computational predictions, eight predicted positions in ccmB and nad4L (five and three, respectively) were experimentally confirmed as C-to-U editing events (Fig. S6; Table S12), all producing non-synonymous codon changes consistent with the tendency of plant mitochondrial editing to restore evolutionarily conserved amino acid residues. Nucleotide diversity Nucleotide diversity (Pi) values were calculated for 38 mitochondrial genes across five Poaceae species to assess sequence polymorphism (Fig. 7; Table S13). The Pi values varied considerably, ranging from 0.00 to 0.21481. Among the analyzed genes, rps2 exhibited the highest nucleotide diversity (Pi = 0.21481), containing 261 polymorphic sites within a 1497 bp region. This was followed by atp9 (Pi = 0.05156) and atp6 (Pi = 0.04340), which contained 20 and 52 polymorphic sites, respectively. The high variability in these genes identifies them as evolutionary hotspots and potential candidates for molecular marker development. In contrast, genes such as rrn5, rps14, and sdh4 showed no nucleotide diversity (Pi = 0.00), indicating high conservation across the sampled species. Nucleotide diversity (Pi) values of 38 mitochondrial genes across five Poaceae species Comparative genomic analysis Genome-wide synteny analysis revealed substantial variation in collinearity levels across species (Fig. 8A). A. sativa and A. longiglumis exhibited remarkably high collinearity (92.46% and 93.04%, respectively), reflecting their close evolutionary relationship. Quantitative comparison using MUMmer4 (nucmer/dnadiff) revealed an exceptionally high average sequence identity of 99.91% between the aligned regions, with only 167 SNPs and limited small-scale structural differences, demonstrating exceptional sequence conservation between the two mitogenomes (Table S14). In contrast, comparisons with rice subspecies showed much lower synteny (O. sativa subsp. indica: 37.74%; O. sativa subsp. japonica: 37.76%), while T. aestivum displayed intermediate collinearity (52.18% and 42.37%). Syntenic blocks were not arranged in the same order across species, suggesting frequent structural rearrangements during evolution. Comparative genomic analysis of A. sativa and related Poaceae species. (A) Synteny dot-plots showing pairwise collinearity. Red and blue lines represent forward and reverse alignments, respectively, with numbers in parentheses indicating the percentage of aligned bases (alignment coverage). (B) Presence/absence heatmap of 43 mitochondrial genes. Red, white, and gray squares indicate gene presence, absence, and pseudogenes, respectively. A cladogram reflecting the phylogenetic relationships among the species is shown on the left Gene content analysis revealed lineage-specific patterns (Fig. 8B). Core OXPHOS genes (atp, cox, nad, cob) were universally conserved, while ribosomal protein genes showed considerable variation. Five genes (sdh3, rps8, rps10, rpl6, rpl10) were absent across all species. The oat lineage showed additional losses of rps11 and rpl2, with rps19 being absent in A. longiglumis and pseudogenized in A. sativa. Rice subspecies displayed pseudogenization of rps11 and rps14, with rpl16 additionally pseudogenized in japonica. T. aestivum exhibited the most extensive gene loss pattern. Phylogenetic inference Phylogenetic relationships were reconstructed using mitochondrial genomes from 34 Poaceae species and 4 outgroup species (Fig. 9). The maximum likelihood tree showed clear clustering by taxonomic groups, with bootstrap support values ranging from 45% to 100%. A. sativa clustered with A. longiglumis (bootstrap support 100%), forming a distinct clade within the Pooideae subfamily. The Avena lineage was positioned as a sister group to the Triticeae tribe (including Triticum, Hordeum, and Aegilops). Other major clades corresponded to their respective subfamilies, including Panicoideae (Zea, Sorghum, Setaria) and Oryzoideae (Oryza species). Outgroup species (Cyperus, Carex, Suaeda) formed basal clades, with Suaeda glauca showing the greatest divergence (bootstrap 100%). Maximum likelihood phylogenetic tree based on mitochondrial genomes from 34 Poaceae species and 4 outgroup species. Numbers at nodes indicate bootstrap support values (%). The scale bar represents substitutions per site Chloroplast-derived mitogenomic sequences Comparative analysis revealed extensive homologous sequences between the chloroplast and mitochondrial genomes of A. sativa. Based on BLASTn results, 60 homologous alignments were identified, with alignment lengths ranging from 39 bp to 7418 bp, and sequence similarities spanning 73.57% to 100.0% (Table S15). For instance, a homologous fragment exhibited a high identity of 99.11% over a 7,418 bp alignment (41 mismatches, 9 gaps), containing chloroplast genes such as atpA (partial: 83.66%), trnR-UCU, trnfM-CAU, rps14, psaB, and psaA, which matched mitochondrial genes trnM-CAT and trnR-TCT. Similarly, trnW-CCA showed 88.71% identity across 124 bp (11 mismatches, 3 gaps), fully contained within its homologous fragment, aligning with an mt region. In contrast, rps12 was partially aligned, with only 4.00% of its length (130 bp) matching an mt region at 98.46% identity. The Circos plot visually confirmed these connections, highlighting frequent transfers of photosynthetic genes (e.g., atpA, psaB) and tRNAs (e.g., trnR-UCU, trnW-CCA) (Fig. 10). Homologous sequences between the chloroplast and mitochondrial genomes of A. sativa. Genes from the same functional complex are marked in identical colors, with connecting lines indicating homologous regions. Outer and inner circle blocks represent genes on positive and negative strands, respectively Nuclear mitochondrial DNA segments (NUMTs) Beyond the chloroplast-mitochondrion DNA transfer described above, extensive integration of mitochondrial DNA into the nuclear genome was also observed. BLAST analysis identified 16,232 NUMTs distributed across all 21 chromosomes of the hexaploid oat genome (Fig. S7). However, this chromosomal distribution was highly non-uniform. Notably, chromosome 3 C exhibited the most extensive accumulation of mitochondrial sequences, with a total length approaching 1 Mb, representing the highest density of insertion among all chromosomes. In contrast, chromosomes such as 6 A showed relatively limited integration events. Analysis of length distribution revealed that while small fragments predominated, large insertions were also substantial (Fig. S8). The majority of NUMTs were short, with 5,140 segments (31.7%) ranging from 200 to 300 bp and 2,415 segments (14.9%) ranging from 300 to 400 bp. However, contrary to the typical fragmentation pattern, we detected a significant number of large insertions; specifically, 1,273 NUMTs exceeded 3 kb in length, including 921 segments that were longer than 4 kb. The coexistence of abundant short fragments and substantial large segments suggests a complex history of mitochondrial DNA integration involving both continuous insertion events and subsequent fragmentation during oat evolution. Discussion Structural complexity and repeat-mediated recombination of the A. sativa mitogenome Mitochondria are crucial for energy production in plant cells, profoundly influencing their growth and development [58]. The functionality of a genome is frequently modulated by its structural attributes, and most mitochondrial genomes in higher plants exhibit complex circular or linear structures [23]. Notably, there is substantial structural variation in mitochondrial genomes across different plant species. In this study, we assembled and reported the first complete mitochondrial genome of A. sativa, an important cereal and forage crop in the Poaceae family. The mitochondrial genome of A. sativa exhibits a multi-branched conformation that resolves into two independent circular chromosomes, with a total length of 634,262 bp. Chromosome 1 is 559,944 bp, and Chromosome 2 is 74,318 bp, with an overall GC content of approximately 44.11%. The genome length and GC content are comparable to those of other Poaceae mitochondrial genomes, such as T. aestivum (452,638 bp, 44.36%) [39] and Lolium perenne (678,580 bp, 44.10%) [59]. Numerous repetitive sequences were identified in the A. sativa mitogenome, including 138 SSRs, 19 tandem repeats, and 383 dispersed repeats. A closer examination of this repeat landscape reveals two features that directly shape the multipartite organization of the A. sativa mitogenome. First, the length distribution is strongly skewed: most dispersed repeats are short (< 100 bp), while ten large repeat pairs (≥ 10 kb, up to 44,994 bp) with near-perfect sequence identity (≥ 99.98%) are located exclusively within chromosome 1. Repeats of this size and similarity are well-established substrates for intramolecular homologous recombination in plant mitogenomes [60, 61] and are known to generate alternative sub-stoichiometric conformations [62] underlying the multipartite organization typical of this organelle [63]. Second, the 38 inter-chromosomal repeats are all short (< 500 bp), indicating that homology-dependent exchange between the two chromosomes is limited. Together, these observations suggest that chromosome 1 likely exists as a dynamic population of recombination-derived isoforms, whereas the two chromosomes behave as largely autonomous molecules rather than as interconverting components of a single master circle [63]. Consistent with this view, our comparative analysis showed that although A. sativa shares high sequence similarity with A. longiglumis, the syntenic blocks are not arranged in the same order, reflecting the structural variation commonly observed in plant mitochondrial genomes [25, 64]. Furthermore, comparative SSR analysis across eight Poaceae species revealed a conserved predominance of tetranucleotide motifs, while Z. mays showed a pronounced expansion of pentanucleotide and hexanucleotide repeats, consistent with the well-documented repeat-rich nature of the maize mitochondrial genome. These cross-species SSR profiles provide a useful reference for future development of mitochondrial SSR markers in Poaceae. Evolutionary selection and post-transcriptional modification of mitochondrial genes Over time, the mitochondrial genomes of most eukaryotes have been further streamlined, with many genes either lost or relocated to the nucleus through endosymbiont gene transfer (EGT) [65]. Consequently, plant mitochondrial genomes often experience the loss of PCGs. Comparative analysis with the diploid A. longiglumis [15] revealed lineage-specific variations in gene content that reflect this ongoing process and the complex polyploid origin of A. sativa. Notably, rps19 is pseudogenized in A. sativa but is completely absent in A. longiglumis. This pattern suggests that A. sativa acquired rps19, possibly from a different ancestral diploid genome during polyploid formation, though the gene has subsequently lost function. The retention of this pseudogene indicates relatively recent acquisition, as non-functional sequences are typically eliminated over evolutionary time. Beyond gene content variation, codon usage analysis of the A. sativa mitochondrial genome revealed a clear bias toward A/T-ending codons. Notably, AT-rich codons such as UUU (Phe) and AUU (Ile) were more frequently used than their GC-rich counterparts, consistent with the overall nucleotide composition of the mitochondrial genome. Comparative analysis across eight Poaceae species further showed that the overall RSCU patterns were broadly similar, suggesting that codon usage bias is relatively conserved during grass mitochondrial genome evolution. To understand the evolutionary pressures acting on these retained genes, Ka/Ks analysis was performed. The results showed predominant purifying selection (Ka/Ks < 1), particularly in ribosomal and respiratory chain complex subunits, maintaining functional integrity. In contrast, elevated Ka/Ks ratios in mttB and nad7 may indicate adaptive evolution, indicating diverse evolutionary pressures within the conserved OXPHOS system [66]. Post-transcriptional modification provides another layer of regulation. In this study, we identified 463 RNA editing sites in A. sativa, which were similar to other angiosperms, such as O. sativa (491) [67], Setaria italica (417) [68] and Agrostis stolonifera (460) [69]. The most frequent amino acid changes involved serine to leucine, serine to phenylalanine, and proline to leucine. The prevalence of hydrophobicity-enhancing edits, particularly in genes associated with cytochrome c biogenesis (ccm family) and ATP synthesis (atp genes), suggests a potential role in optimizing protein stability or mitochondrial function. Future research should focus on functional analyses to determine the impact of these RNA editing events on protein structure and function [70]. This computational portrait was experimentally supported by Sanger sequencing of ccmB and nad4L, in which eight predicted sites were confirmed as C-to-U editing events, all producing non-synonymous amino acid changes consistent with the editing-mediated restoration of conserved residues described above. Finally, the nucleotide diversity of the 38 mitochondrial genes varied considerably. rps2 exhibited the highest diversity (Pi = 0.21481), indicating substantial sequence variation and potential as a genetic marker. In contrast, nad7 showed low diversity (Pi = 0.00169), suggesting stronger selective constraints. These hypervariable regions hold significant potential for the development of specific molecular markers. Future population-scale studies are warranted to confirm species-specific differences, providing a basis for studying Poaceae evolution. Phylogenetic inference and intracellular gene transfer (IGT) dynamics Phylogenetic analysis of 34 Poaceae mitochondrial genomes revealed distinct clustering of subfamilies, with strong support for the Pooideae clade (91 ~ 100% bootstrap) and basal positioning of outgroups like Suaeda glauca, consistent with previous studies [15]. Short intra-clade branch lengths contrasted with longer inter-order distances, underscoring tight evolutionary relationships within orders and broader divergence across Poaceae. These findings demonstrated the utility of mitochondrial genomes in resolving plant phylogenies, as organelle-derived sequences reliably reflect taxonomic relationships, consistent with previous studies on plant organellar genomes [71]. Beyond broader Poaceae-level relationships, the specific maternal origin of the hexaploid (AACCDD) A. sativa mitogenome traces unambiguously to the A-subgenome lineage. This is evidenced by the exceptional sequence conservation (99.91% identity over > 92% of aligned bases; Fig. 8A, Table S14) between A. sativa and the A-genome diploid A. longiglumis. This finding is strongly corroborated by previous organelle phylogenomic analyses, which demonstrated that cultivated oats acquired their maternal genomes from an A-genome diploid via an AC-tetraploid intermediate [72]. While further comparisons are constrained by the absence of publicly available C-genome mitogenomes and extant D-genome diploids [73, 74], the highly conserved A-genome reference firmly establishes the evolutionary trajectory of the hexaploid oat mitogenome. Horizontal gene transfer between chloroplast (cp.) and mitochondrial (mt) genomes is well-documented in plants and is critical for understanding plant mitogenome evolution [75–77]. Comparative analysis of A. sativa revealed 60 homologous cp-mt sequences, with alignments (39 ~ 7418 bp) exhibiting 73.57 ~ 100% similarity. Notable transfers included photosynthetic genes (atpA, psaB) and tRNAs (trnR-UCU), indicating frequent cp-to-mt gene flow. These results exhibited dynamic organellar genome interactions and support distinct mitochondrial evolutionary patterns in Poaceae [64]. In parallel with chloroplast-to-mitochondrion transfers, our study also uncovered extensive migration of mitochondrial DNA into the nuclear genome (NUMTs). The chromosomal distribution of these NUMTs is highly non-uniform, with chromosome 3 C harboring a striking accumulation of mitochondrial sequences (approaching 1 Mb). Such biased distribution patterns align with observations in other plant genomes [78] and may stem from variations in chromatin accessibility or local recombination rates [79]. Whether this preferential accumulation reflects specific structural features of chromosome 3 C remains a subject for future investigation. Furthermore, the length distribution of NUMTs provides additional evolutionary insights. Contrary to the expectation that nuclear-integrated fragments degrade rapidly, we detected a substantial number of large segments, including 921 NUMTs exceeding 4 kb. The retention of such lengthy, intact sequences is comparable to the 641-kb insertion reported in Arabidopsis thaliana [80], suggesting that gene transfer from mitochondria to the nucleus is a relatively recent or ongoing evolutionary process in hexaploid oat. These large segments potentially serve as physical intermediates in endosymbiotic gene transfer [81], contributing substantially to the complexity of the nuclear genome. Conclusions This study presents the first complete mitochondrial genome of hexaploid oat (A. sativa), comprising 634,262 bp across two circular chromosomes with 78 genes. Comparative analysis with the diploid oat (A. longiglumis) revealed lineage-specific gene content variations, notably rps19 pseudogenization in A. sativa versus its absence in A. longiglumis, reflecting complex polyploid genome evolution. The genome harbors abundant repetitive sequences (138 SSRs, 383 dispersed repeats) facilitating structural plasticity, and 463 RNA editing sites enhancing protein function. Most notably, we uncovered extensive migration of mitochondrial DNA to the nucleus (NUMTs), with a striking enrichment hotspot on chromosome 3 C. Phylogenetic analysis confirmed close evolutionary relationships within Pooideae. This genomic resource is helpful for the understanding of mitochondrial evolution in polyploid crops and provides a genomic foundation for future studies of mitochondrial biology and comparative genomics in oats. Supplementary Information Supplementary Material 1. Acknowledgements Not applicable. Abbreviations Protein-coding Gene Oxidative Phosphorylation Simple Sequence Repeat Relative Synonymous Codon Usage Non-synonymous to synonymous substitution rate ratio Nuclear mitochondrial DNA segment Intracellular Gene Transfer Cytoplasmic Male Sterility Coding Sequence Open Reading Frame Mitochondrial DNA Chloroplast DNA Authors’ contributions Shiqiang Zhao: Writing–review & editing, Writing–original draft, Visualization, Validation, Software, Resources, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Ruohong Li: Data curation. Zhicheng Jia: Software. Tianqi Zhu: Data curation. Chunjiao Mi: Visualization. Juan Wang: Visualization. Zhijia Lv: Formal analysis. Peisheng Mao: Supervision, Project administration, Investigation, Writing–review & editing, Funding acquisition. Funding This research was supported by the earmarked fund for CARS (CARS-34). Data availability The assembled mitochondrial genome sequences generated in this study have been deposited in the NCBI GenBank database under the accession numbers PV640483 (https://www.ncbi.nlm.nih.gov/nuccore/PV640483) and PV640484 (https://www.ncbi.nlm.nih.gov/nuccore/PV640484). Declarations Ethics approval and consent to participate Not applicable. Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1. Data Availability Statement The assembled mitochondrial genome sequences generated in this study have been deposited in the NCBI GenBank database under the accession numbers PV640483 (https://www.ncbi.nlm.nih.gov/nuccore/PV640483) and PV640484 (https://www.ncbi.nlm.nih.gov/nuccore/PV640484).