Complete mitogenome assembly of Lilium amoenum E. H. Wilson ex Sealy revealed its multiple chromosomal configurations, genome evolution, phylogenetic implications, and RNA-editing

Abstract

Background

Lilium amoenum E. H. Wilson ex Sealy is a wild lily species endemic to Yunnan, China, and is of considerable ornamental and scientific interest owing to its solitary, pendulous, fragrant, rose-purple flowers. Although the chloroplast genome of this species has been reported, its mitogenome has not yet been characterized. To address this gap, we employed a hybrid sequencing strategy combining Illumina short-read and Nanopore long-read data to assemble and annotate, for the first time, the complete mitogenome of L. amoenum.

Result

The mitogenome of L. amoenum exhibits a distinctive multichromosomal architecture comprising 13 circular chromosomes and one linear chromosome, with a total length of 789,335 bp and an overall GC content of 45.11%. In total, 55 genes were annotated, including 37 protein-coding genes (PCGs), 3 rRNA genes, and 15 tRNA genes. We also identified 171 simple sequence repeats (SSRs), eight tandem repeats, and 509 dispersed repeats. Homology analysis detected 14 plastid-derived fragments totaling 4,723 bp, representing 0.60% of the mitogenome. Collinearity analysis revealed extensive genomic rearrangements between L. amoenum and its close relatives, indicating substantial structural divergence during mitogenome evolution. Phylogenetic analysis based on the mitogenomes of 26 species showed that L. amoenum clusters with other members of Lilium, consistent with the APG IV classification system. In addition, 493 C-to-U RNA editing sites in the coding regions of mitochondrial PCGs were predicted. Lastly, we successfully validated five randomly selected RNA editing sites.

Conclusion

This study provides the first complete mitogenome of L. amoenum, revealing its unique multi-chromosomal configuration. Moreover, it offers a valuable genomic resource for breeding, species identification, germplasm conservation, and phylogenetic studies within Lilium species.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12870-026-09675-4.

Introduction

Mitochondria are widely accepted to have originated through an ancient endosymbiotic event, in which an alphaproteobacterial ancestor was engulfed by a primordial eukaryotic host cell. This integration was subsequently accompanied by extensive gene loss from the endosymbiont genome and large-scale intracellular gene transfer to the host nuclear genome [1, 2]. In plants, mitochondria serve as central hubs for cellular energy production via oxidative phosphorylation and play critical roles in maintaining metabolic homeostasis. Although plant mitogenomes have undergone substantial genomic reduction and retain only a limited repertoire of PCGs, these genes remain functionally indispensable, underpinning oxidative phosphorylation, organelle biogenesis, developmental regulation, abiotic and biotic stress responses, and cytonuclear coordination [3, 4]. Beyond their fundamental biological significance, plant mitogenomes are also directly relevant to agronomic applications. Notably, cytoplasmic male sterility (CMS), a maternally inherited trait of significant value in hybrid crop breeding, often arises from chimeric open reading frames produced by structural rearrangements in the mitogenome. The phenotypic expression of CMS is modulated by nuclear-encoded restorer-of-fertility genes [5–7]. Consequently, elucidating the structure, function, and evolutionary dynamics of plant mitogenomes is of broad significance to both fundamental organelle biology and applied crop improvement.

Compared with animal mitogenomes, plant mitogenomes exhibit much greater complexity and plasticity. Their major features include striking variation in genome size, diverse physical conformations, and the coexistence of relatively conserved coding regions with highly dynamic genome structures [8–13]. Plant mitogenomes vary in size from tens of kilobases (e.g., 15.8 kb in Chlamydomonas reinhardtii) to several megabases (e.g., 18.99 Mb in Cathaya argyrophylla). This expansion is driven primarily by the accumulation of extensive non-coding regions, repetitive sequences, plastid-derived fragments, and other foreign DNA, rather than by a proportional increase in functional gene content [8–10, 14, 15]. Meanwhile, plant mitochondrial DNA is rarely maintained as a single stable circular molecule; instead, it often exists as a heterogeneous mixture of circular, linear, branched, subgenomic, and multichromosomal forms that can interconvert through repeat-mediated homologous recombination [8–10]. Notably, plant mitogenomes display the characteristic combination of low nucleotide substitution rates and high structural variability, in which core coding regions are generally conserved over evolutionary time, whereas repeat-mediated recombination, insertions/deletions, and chromosomal rearrangements remain highly active [10–13]. In addition, plant mitochondrial gene expression relies on complex post-transcriptional processes, including group II intron splicing, trans-splicing, and extensive RNA editing, all of which contribute to transcript maturation and functional maintenance [3, 6, 16]. Collectively, these features make plant mitogenomes an important molecular system for investigating organelle genome evolution, cytonuclear interactions, and the genetic basis of complex traits.

Recent advances in high-throughput sequencing technologies and bioinformatic approaches have substantially promoted research on plant mitogenomes. Although short-read-based assembly strategies are efficient for organelle sequence enrichment and preliminary genome assembly, they remain limited in resolving long repeats, complex recombination junctions, and low-abundance structural isoforms [17]. In contrast, third-generation long-read sequencing platforms, including PacBio HiFi and Oxford Nanopore, together with graph-based or HiFi-oriented assembly methods such as Master graph, PMAT, TIPPo, and HiMT, have markedly improved the resolution of repetitive regions, recombination-derived connections, and coexisting structural conformations [18–22]. Benefiting from these advances, publicly available plant mitogenome resources have increased rapidly, with the number of plant mitochondrial genomes deposited in the NCBI database rising from approximately 291 in 2020 to more than 636 by 2025, including about 345 additional genomes added during 2021–2025 [23, 24]. At the same time, an increasing number of plant mitogenomes with complex structural features have been characterized [8, 25, 26]. For example, the mitogenome of Silene noctiflora comprises more than 50 circular-mapping chromosomes [27], whereas that of Picea sitchensis exhibits a multipartite structure with branched physical configurations [28]. Together, these studies indicate that plant mitogenome research is moving beyond basic sequencing of a limited number of species toward broader phylogenetic sampling and deeper investigation of structural heterogeneity and dynamic evolutionary mechanisms, thereby prompting a re-evaluation of the traditional “master circle” model.

The genus Lilium comprises economically and horticulturally important species, and recent genomic studies have substantially expanded the resources available for this group. Plastome-based analyses have clarified phylogenetic relationships within the genus and provided insights into the maternal origins of cultivated lilies [29, 30]. In addition, nuclear genome assemblies have been reported for L. davidii var. unicolor, L. sargentiae, and L. regale, further enriching genomic resources for Lilium [31–33]. By contrast, L. amoenum, a wild lily species endemic to Yunnan, China, remains poorly studied despite its importance as a wild germplasm resource. It mainly grows on forested slopes and in shrubby grasslands at elevations of 1,800-3,000 m and is distinguished by its solitary, nodding, fragrant, rose-purple flowers [29]. To date, studies on this species have focused mainly on taxonomic treatment, phylogenetic analysis, and chloroplast genome characterization [29, 30]. At the mitogenome level, available resources for Lilium remain limited. The first complete mitogenome reported for the genus, that of L. tsingtauense, spans 1,125,108 bp and consists of 27 independent circular chromosomes, representing a typical multipartite and multichromosomal architecture accompanied by frequent repeat-mediated recombination [34]. In the closely related genus Cardiocrinum, the mitogenome of C. giganteum comprises 19 circular molecules, further suggesting that extensive fragmentation and dynamic reorganization may be common features of mitogenomes in Liliaceae [35]. Although the chloroplast genome of L. amoenum has been reported [29, 36], the mitogenome of L. amoenum has not yet been systematically characterized. Given the central roles of plant mitogenomes in energy metabolism, cytoplasmic inheritance, organ development, and cytoplasmic male sterility formation, characterization of the L. amoenum mitogenome will not only enrich organelle genomic resources for Lilium, but also provide valuable insights into mitochondrial structural evolution, phylogenetic relationships, and the utilization of wild germplasm.

In this study, we combined second-generation Illumina short-read sequencing with third-generation Oxford Nanopore Technologies long-read sequencing to generate a high-quality assembly and annotation of the mitogenome of L. amoenum. Based on this assembly, we systematically analyzed genome structure, codon usage bias, repetitive sequences, phylogenetic relationships, and synteny. In addition, the chloroplast genome was assembled using Illumina data, and intracellular gene transfer between the chloroplast and mitogenome was investigated. RNA editing sites were also predicted and validated by PCR amplification and Sanger sequencing. These results provide valuable genomic resources for elucidating the structural features and evolutionary history of the L. amoenum mitogenome, and lay a foundation for further research on this ornamental species and for the conservation and utilization of Lilium germplasm resources.

Materials and methods

Plant material sampling, DNA extraction, and sequencing

Young, healthy bulbs of L. amoenum were collected from the tissue culture room at the College of Landscape and Horticulture, Yunnan Agricultural University, located in Panlong District, Kunming, Yunnan Province, China (102°75’E, 25°13′N). The samples were immediately flash-frozen in liquid nitrogen and stored at -80 °C. Genomic DNA was extracted with a Plant Genomic DNA Kit (DP305, Tiangen Biotech, Beijing, China) following the manufacturer’s protocol. DNA quality and concentration were evaluated by agarose gel electrophoresis and NanoDrop spectrophotometry, respectively.

Sequencing was performed using a combined Illumina and Nanopore strategy. For Illumina sequencing, paired-end libraries with an insert size of 350 bp were constructed and sequenced on the Illumina HiSeq 2500 platform. For Nanopore sequencing, DNA fragments longer than 20 kb were selected using the BluePippin automated size-selection system (Sage Science, MA, USA) and subsequently purified with magnetic beads. The libraries were then quantified using a Qubit Fluorometer and sequenced on the PromethION platform using an R9.4.1 (FLO-PRO002) flow cell (Oxford Nanopore Technologies, Oxford, UK).

Mitogenome assembly and annotation

The L. amoenum mitogenome was initially assembled using ONT long-read data. Long reads were directly assembled using Flye (v2.9.2) [37] with default parameters, generating a graph-based assembly in GFA format. All assembled contigs in FASTA format were then formatted into a local BLAST database, and candidate mitochondrial contigs were identified by BLASTn (v2.16.0) [38] searches using the mitogenome of the related species L. tsingtauense (OP973783.1-OP973810.1) as the query sequence. The BLASTn parameters were set to ‘-evalue 1e-5 -outfmt 6 -max_hsps 10 -word_size 7 -task blastn-short’. The GFA file was visualized using Bandage (v0.8.1) [39], and candidate mitochondrial contigs were screened based on the BLASTn results to obtain a preliminary mitochondrial assembly graph.

Subsequently, long reads and short reads were independently mapped to the filtered mitochondrial contigs using minimap2 (v2.26-r1175) [40], and the mapped mitochondrial reads were extracted for subsequent hybrid assembly. Using both long-read and short-read data, hybrid assembly was performed with Unicycler (v0.8.1) [41] under default parameters to generate candidate mitogenome structures, which were again visualized using Bandage (v0.8.1) [39]. To resolve repeat regions in the assembly graph, long reads were further mapped to repeat regions and their flanking sequences to determine whether reads spanned these regions. The most likely genomic paths were then inferred based on the long-read evidence. Ultimately, the mitogenome was resolved into 14 chromosomes, among which Chr2 was linear and the remaining 13 were circular.

Genome annotation was performed using the plant mitogenome annotation tool PMGA (http://www.1kmpg.cn/pmga/) [42], with a reference database containing 319 mitogenomes. tRNA genes were annotated using tRNAscan-SE (v2.0.11) [43], whereas rRNA genes were identified using BLASTN (v2.16.0) [38]. Potential annotation errors were manually corrected using Apollo (v1.11.8) [44], and the final mitogenome map was generated using OGDRAW [45].

Analysis of relative synonymous codon usage (RSCU) and repeat sequences

Protein-coding sequences of all annotated mitochondrial PCGs were extracted using PhyloSuite (v1.1.16) [46] and checked for complete reading frames. Terminal stop codons were removed before analysis. Codon usage bias was then assessed in MEGA v7.0 [47]. The sequences were imported into the Sequence Data Explorer, and codon frequencies and RSCU values were calculated using the built-in Statistics > Codon Usage function. For each codon, RSCU was defined as its observed frequency divided by the expected frequency under equal usage of all synonymous codons for the corresponding amino acid. An RSCU value greater than 1 indicates more frequent usage than expected, a value below 1 indicates less frequent usage, and a value of 1 indicates no synonymous codon usage bias.

SSRs, tandem repeats, and dispersed repeats were identified using MISA (v2.1) [48], TRF (v4.09) [49], and RousFinder [50], respectively. The resulting data were summarized and visualized using Excel (2021) and the Circos package (v0.69.9) [51].

Identification of mitochondrial plastid DNAs (MTPTs) and synteny analysis

The chloroplast genome was assembled from short-read data using GetOrganelle (v1.7.7.0) [17], annotated with CPGAVAS2 [52], and further corrected using CPGView [53]. Homologous fragments shared between the chloroplast and mitogenomes were identified using BLASTN (v2.16.0) [38], and the distribution of MTPTs was visualized using the Circos package (v0.69.9) [51].

For synteny analysis, conserved homologous sequences were identified as syntenic blocks using BLASTn (v2.16.0) [38] with the following parameters: ‘-evalue 1e-5 -word_size 9 -gapopen 5 -gapextend 2 -reward 2 -penalty − 3’. Only syntenic blocks longer than 500 bp were retained for further analysis. Based on the BLASTn results, pairwise comparisons and multiple synteny plots were generated using MCScanX [54] to visualize conserved syntenic blocks among species.

Phylogenetic analysis

To infer the phylogenetic position of L. amoenum, 26 complete mitogenomes were selected for phylogenetic reconstruction, including 10 species from Asparagales, five from Liliales, three from Pandanales, and six from Alismatales as the ingroup, with two Ranunculales species used as the outgroup (Table S1). Shared genes were extracted using PhyloSuite (v1.1.16) [46], aligned using MAFFT (v7.525) [55], and used to reconstruct a maximum-likelihood (ML) phylogenetic tree with IQ-TREE (v2.3.6) [56] under the parameters ‘--alrt 1000 -B 1000’. The best-fit substitution model was ‘GTR + F+I+R3’. The resulting phylogenetic tree was visualized using iTOL (v7) [57].

Identification and validation of RNA editing sites

All mitochondrial PCG sequences of L. amoenum were used as input for Deepred-mt [58] to predict C-to-U RNA editing sites in mitochondrial PCGs. Only predictions with probability values greater than 0.9 were retained.

To validate the predicted RNA editing sites, we randomly selected five genes for experimental verification. Primers flanking the predicted editing sites were designed (Table S2), and both genomic DNA (gDNA) and cDNA served as amplification templates. Genomic DNA was extracted using the protocol described earlier. Total RNA was isolated from the bulb with an RNA extraction kit (TIANGEN, DP432, Beijing, China), and first-strand cDNA was synthesized using a FastKing RT Kit (TIANGEN, KR116, Beijing, China). PCR amplification was conducted under the following conditions: 94 ℃ for 3 min; 35 cycles of 95 ℃ for 15 s, 55 ℃ for 15 s, and 72 ℃ for 15 s; and a final extension at 72 ℃ for 5 min. The PCR products were then sequenced by the Sanger method (Sangon Biotech (Shanghai) Co., Ltd.). RNA editing events were identified by comparing the sequences of the amplified gDNA and cDNA products.

Results

Genomic features of the mitogenome of L. amoenum

Based on the preliminary assembly generated from ONT long-read data, the assembly graph of the L. amoenum mitogenome was visualized using Bandage (v0.8.1) [39] (Fig. 1A). The graph comprised 18 contigs, among which ctg1, ctg2, ctg3, ctg16, ctg17, and ctg18 were interconnected, forming a complex branched structure, whereas the remaining 12 contigs appeared as independent circular molecules. Long reads were subsequently mapped back to these contigs to evaluate read support for alternative connections among contigs, and the paths supported by larger numbers of reads were preferentially resolved. Based on the long-read evidence spanning repeat and junction regions, the contigs were integrated in the order ctg1-ctg18_copy-ctg17_copy-ctg2-ctg17-ctg18-ctg16 to generate a new circular molecule, whereas ctg3 was resolved as a linear molecule and the remaining contigs were retained as independent circular molecules (Fig. 1B). Ultimately, 14 sequences were obtained, representing the complete mitogenome configuration of L. amoenum. The correspondence between each final sequence and the nodes in Fig. 1B is summarized in Table S3.

Fig. 1.

The assembly result of the mitogenome of L. amoenum. A The preliminary results of the mitogenome assembly. B The structure of the assembled mitogenome.

The mitogenome of L. amoenum exhibited a typical multichromosomal architecture, consisting of 13 circular chromosomes and one linear chromosome, with a total length of 789,335 bp and an overall GC content of 45.11% (Fig. 2). The lengths of the 14 chromosomes were 275,108 bp, 65,668 bp, 54,008 bp, 51,648 bp, 48,838 bp, 42,399 bp, 40,053 bp, 37,792 bp, 35,174 bp, 30,853 bp, 30,115 bp, 26,718 bp, 26,594 bp, and 24,367 bp, respectively; the corresponding GC contents were 45.08%, 44.91%, 44.76%, 45.46%, 44.11%, 46.07%, 45.34%, 45.83%, 44.60%, 45.17%, 45.40%, 44.90%, 45.27%, and 45.38%, respectively (Table 1). A total of 55 genes were annotated, including 37 unique PCGs, 15 tRNA genes, and three rRNA genes, among which four tRNA genes were present in multiple copies. Of the 37 PCGs, 24 were core mitochondrial genes and 13 were variable genes (Fig. 2, Table 2). The core genes included five ATP synthase genes (atp1, atp4, atp6, atp8, and atp9), nine NADH dehydrogenase genes (nad1, nad2, nad3, nad4, nad4L, nad5, nad6, nad7, and nad9), four cytochrome c biogenesis genes (ccmB, ccmC, ccmFC, and ccmFN), three cytochrome c oxidase genes (cox1, cox2, and cox3), one membrane transporter gene (mttB), one maturase gene (matR), and one cytochrome b gene (cob). The variable genes comprised two large-subunit ribosomal protein genes (rpl5 and rpl16), 10 small-subunit ribosomal protein genes (rps1, rps2, rps3, rps4, rps7, rps10, rps12, rps13, rps14, and rps19), and one succinate dehydrogenase gene (sdh4).

Fig. 2.

Map of the L. amoenum mitogenome. Genes transcribed clockwise are shown inside the circle, while those transcribed counterclockwise are placed outside. Different functional groups of genes are distinguished by color

Table 1.

Basic mitogenome information

Contigs	Length (bp)	GC content (%)
Chromosome 1–14	789,335	45.11
Chromosome1	275,108	45.08
Chromosome2	65,668	44.91
Chromosome3	54,008	44.76
Chromosome4	51,648	45.46
Chromosome5	48,838	44.11
Chromosome6	42,399	46.07
Chromosome7	40,053	45.34
Chromosome8	37,792	45.83
Chromosome9	35,174	44.60
Chromosome10	30,853	45.17
Chromosome11	30,115	45.40
Chromosome12	26,718	44.90
Chromosome13	26,594	45.27
Chromosome14	24,367	45.38

Table 2.

Gene composition of the L. amoenum mitogenome

Group of genes	Name of genes
ATP synthase	atp1, atp4, atp6, atp8, atp9
NADH dehydrogenase	nad1, nad2, nad3, nad4, nad4L, nad5, nad6, nad7, nad9
Cytochrome b	cob
Cytochrome c biogenesis	ccmB, ccmC, ccmFC, ccmFN
Cytochrome c oxidase	cox1, cox2, cox3
Maturases	matR
Protein transport subunit	mttB
Ribosomal protein large subunit	rpl5, rpl16
Ribosomal protein small subunit	rps1, rps2, rps3, rps4, rps7, rps10, rps12, rps13, rps14, rps19
Succinate dehydrogenase	sdh4
Ribosome RNA	rrn5, rrn18, rrn26
Transfer RNA	trnC-GCA, trnD-GUC, trnE-UUC, trnfM-CAU (x6), trnH-GUG, trnI-CAU, trnK-UUU, trnL-GAG, trnM-CAU (x3), trnN-GUU (x2), trnQ-UUG (x2), trnT-UGU, trnV-GAC, trnW-CCA, trnY-GUA

The numbers in parentheses represent gene copy numbers

PCG codon usage analysis

Codon usage bias can reflect long-term patterns of nucleotide composition and translational selection during genome evolution. Therefore, RSCU analysis was performed for the mitochondrial PCGs of L. amoenum. The codon usage frequencies for each amino acid are listed in Table S4. In general, codons with RSCU values greater than 1 are considered preferentially used. Except for the start codon AUG and the tryptophan codon UGG, both of which had RSCU values of 1, most amino acids exhibited varying degrees of codon usage bias (Fig. 3). Among them, the stop codon UAA showed the strongest bias, with the highest RSCU value of 1.73, followed by GCU encoding alanine (Ala), with an RSCU value of 1.62. These results indicate that the mitochondrial PCGs of L. amoenum exhibit an obvious codon usage bias.

Fig. 3.

L. amoenum mitogenome codon preferences

Repeat sequence analysis

In addition to variation in intergenic regions, the accumulation of repetitive sequences and exogenous fragments is considered an important factor shaping structural diversity in plant mitogenomes. Repeat sequence analysis of the L. amoenum mitogenome identified 171 simple sequence repeats (SSRs), 8 tandem repeats, and 509 dispersed repeats (Fig. 4A and B).

Fig. 4.

Repeat sequences in the L. amoenum mitogenome. A Type and number of SSRs. The x-axis represents the mitochondrial genome, and the y-axis represents the number of repeat fragments. The gray, orange, blue, yellow, green, and purple legend indicates monomeric, dimeric, trimeric, tetrameric, pentameric, and hexameric SSRs, respectively. B Analysis of Repeat Sequences Using Circos Plot. The colored lines on the innermost circle connect the two repeat sequences of dispersed repeats, with purple lines representing palindromic repeats and blue lines representing forward repeats. The black segments on the second circle denote tandem repeat sequences, while the black segments on the outermost ring represent microsatellite repeat sequences

Six categories of SSRs were detected, including mono-, di-, tri-, tetra-, penta-, and hexanucleotide repeats, and their abundances differed substantially. Chr1 contained the largest number of SSRs (61), whereas Chr9, Chr12, and Chr14 each harbored only three. By repeat type, tetranucleotide repeats (35.09%, 60 loci) and dinucleotide repeats (28.07%, 48 loci) were predominant, followed by mononucleotide repeats (22.22%), trinucleotide repeats (10.53%), pentanucleotide repeats (5.26%), and hexanucleotide repeats (0.58%) (Table S5). Among the mononucleotide repeats, A/T motifs were overwhelmingly dominant, accounting for 97.14% of all mononucleotide SSRs, suggesting that these loci may serve as potential molecular markers for resource identification and genetic analysis of L. amoenum.

A total of eight tandem repeats with sequence identity greater than 88% and lengths ranging from 9 to 39 bp were identified in the mitogenome (Table S6). No tandem repeats were detected on Chr3, Chr5, Chr6, Chr7, Chr10, Chr11, Chr13, and Chr14. Overall, the limited number and short length of tandem repeats suggest that they may contribute only marginally to overall mitogenome expansion in L. amoenum.

In addition, 509 dispersed repeats with lengths of at least 30 bp were identified, including 225 forward repeats and 284 palindromic repeats, whereas no reverse or complementary repeats were detected (Table S7). The longest dispersed repeat unit, designated R1, was 5,912 bp in length. These dispersed repeats ranged from 30 to 5,912 bp, with a substantial proportion represented by long fragments: 217 repeats exceeded 500 bp and 171 exceeded 1 kb. Collectively, the 509 dispersed repeats spanned 334,966 bp, accounting for 42.44% of the entire mitogenome. These findings indicate that abundant repetitive sequences, particularly dispersed repeats, constitute an important basis for the structural complexity of the L. amoenum mitogenome and may have played a major role in genomic rearrangement and evolutionary diversification.

Analysis of homologous fragments between chloroplast and mitogenomes

In order to identify the homologous fragments between mitochondrial and chloroplast genomes of L. amoenum. Here, we assembled and annotated the L. amoenum chloroplast genome, which contained 152,290 bp and 132 annotated genes, as shown in Fig. 5A. Sequence similarity analysis identified 14 homologous fragments shared between the chloroplast and mitogenome, with a total length of 4,723 bp, accounting for 0.60% of the complete mitogenome (Fig. 5B). These homologous fragments were mainly distributed on Chr1, Chr2, Chr3, Chr4, Chr5, Chr7, Chr9, Chr11, and Chr13. The longest fragment, MTPT8, was 1,930 bp in length and was located on Chr4 (Fig. 5B, Table S8). Further annotation revealed that these 14 homologous fragments contained three intact genes, including one PCG (psbC) and two tRNA genes (trnH-GUG and trnW-CCA). These results indicate that intracellular transfer of chloroplast-derived sequences into the L. amoenum mitogenome was relatively limited, although such transfers may still have contributed to its structural evolution.

Fig. 5.

Genome map of the L. amoenum chloroplast and sequence migration. A L. amoenum chloroplast genome map. Genes residing inside and outside of the outer circle are in the forward and reverse directions, respectively. The dark and light gray bars in the inner circle denote the G + C and A + T contents, respectively. B Schematic of 14 MTPTs of L. amoenum. The orange arcs represent the mitogenome, and the green arcs represent the chloroplast genome. The pink connecting lines between the arcs correspond to homologous genomic fragments

Synteny analysis

To evaluate the structural conservation of the L. amoenum mitogenome relative to those of closely related taxa, comparative synteny analysis was conducted among five species in Liliaceae, including two Fritillaria species (F. taipaiensis and F. usuriensis), two Lilium species (L. amoenum and L. tsingtauense), and one Cardiocrinum species (C. giganteum) (Fig. 6). The results showed that L. tsingtauense and C. giganteum shared the largest number of homologous syntenic blocks, with 191 blocks identified between them (Table S9). Moreover, the cumulative length of these syntenic blocks was the greatest among all pairwise comparisons, reaching 67.10 kb, indicating a certain degree of local conservation between their mitogenomes.

Fig. 6.

Mitogenome synteny. The red areas indicate inversion, grey areas indicate good homology

Notably, 137 syntenic blocks were identified between the two Lilium species. Among them, Chr1 of L. amoenum contained the largest number of syntenic blocks (54), and 116 blocks were longer than 1,000 bp. The longest syntenic block, approximately 30.94 kb in length, was located on Chr9 of L. amoenum and was also the longest syntenic block detected among all comparisons (Table S9). These results indicate that Lilium species still retain several relatively long conserved homologous regions. However, the arrangement of syntenic blocks between L. amoenum and its close relatives was highly inconsistent, suggesting that these mitogenomes have undergone extensive genomic rearrangements during evolution.

Phylogenetic analysis

To clarify the phylogenetic position of L. amoenum, sequence alignments of 27 conserved mitochondrial PCGs from 26 species representing five angiosperm orders were used to reconstruct a phylogenetic tree. The analyzed genes included atp1, atp4, atp6, atp8, atp9, ccmB, ccmC, ccmFC, ccmFN, cob, cox1, cox2, cox3, matR, mttB, nad1, nad2, nad3, nad4, nad4L, nad5, nad6, nad7, nad9, rps3, rps12, and rps19. Two species from Ranunculales were designated as the outgroup.

The resulting phylogenetic tree comprised 23 nodes, of which 18 had bootstrap support values greater than 90% and 17 had support values of 100%, indicating a high level of phylogenetic reliability (Fig. 7). The tree topology was fully consistent with the latest Angiosperm Phylogeny Group IV (APG IV) classification system. L. amoenum was placed within Liliaceae in the order Liliales and clustered with other Lilium species, indicating a close phylogenetic relationship with this genus.

Fig. 7.

Phylogenetic analysis of the L. amoenum mitogenome based on 27 conserved mitogenome PCGs. Numbers related to the branches are bootstrap-supported values

RNA editing site prediction and validation of predicted RNA editing events

A total of 493 putative RNA editing sites were predicted in the mitochondrial PCGs of L. amoenum, and all were of the C-to-U type (Fig. 8, Table S10). Among these genes, nad4 harbored the largest number of predicted RNA editing sites (47), followed by ccmB and ccmC, each of which contained 36 predicted editing sites.

Fig. 8.

Number of RNA editing sites detected on each PCG in L. amoenum mitochondria

To evaluate the reliability of these predictions, five PCGs containing predicted RNA editing sites were randomly selected, and five sites associated with start- or stop-codon formation (nad1-2, rps10-2, atp9-223, ccmFC-1306, and rps10-334) were validated by PCR amplification and Sanger sequencing (Fig. 9A and B, Fig. S1). The results confirmed RNA editing events at all five sites. Among them, the rps10-334 site exhibited a high editing efficiency, with the cytosine residue almost completely edited to uridine, whereas only partial editing was observed at the nad1-2 and ccmFC-1306 sites, indicating marked variation in editing efficiency among different RNA editing sites.

Fig. 9.

Validation of mitochondrial RNA editing sites in L. amoenum. A Cloning of mitochondrial RNA editing sites in L. amoenum. B Sequencing chromatograms of mitochondrial RNA editing sites in L. amoenum

Discussion

Plant mitogenomes usually retain a conserved set of essential genes but vary greatly in size, chromosome number, and physical conformation because repeats promote recombination and generate alternative subgenomic forms [3, 8]. The mitogenome of L. amoenum fits this general pattern. It comprised 13 circular molecules and one linear molecule, totaling 789,335 bp with an overall GC content of 45.11%. This multipartite organization is less expanded and less fragmented than that of the congeneric L. tsingtauense (1,125,108 bp; 27 circular molecules) [34], and it is also smaller than the previously reported mitogenomes of L. pumilum and L. davidii [29]. By contrast, it is closer in scale to F. ussuriensis (737,569 bp) [59] and F. taipaiensis (847,160 bp) [60], but far smaller than the extremely expanded mitogenome of C. giganteum (2,066,679 bp) [35]. These comparisons indicate that mitogenome size and chromosome number do not necessarily correspond to taxonomic relatedness within Liliaceae. Instead, they are more likely to reflect lineage-specific differences in repeat accumulation, recombination frequency, DNA acquisition and loss, and intergenic sequence expansion. The coexistence of circular and linear assembly units is also consistent with the view that plant mitochondrial DNA occurs as a dynamic population of branched, linear, and circular forms rather than as a single stable master circle [8, 61]. Although the complete mitogenome of L. amoenum was assembled using a hybrid sequencing strategy, the dynamic nature of plant mitogenome architecture indicates that some alternative conformations or low-abundance structural variants may not have been fully resolved. Therefore, the 14 assembled molecules should be regarded as well-supported genomic configurations rather than definitive evidence that all molecules exist as stable and independent chromosomes in vivo. Future studies should investigate the structural diversity of the Lilium mitogenome at the population level. In addition, cryo-electron microscopy could be employed to visualize the actual in vivo structural organization of the Lilium mitogenome.

The repeat landscape provides a plausible mechanistic basis for this structural dynamism. The L. amoenum mitogenome contained 171 simple sequence repeats, eight tandem repeats, and 509 dispersed repeats, with dispersed repeats accounting for 42.44% of the total genome. Intermediate and long repeats can serve as substrates for homologous or ectopic recombination, generating alternative genomic configurations, inversions, translocations, and subgenomic molecules [62–64]. However, SSR abundance alone does not explain structural complexity: F. ussuriensis and F. taipaiensis contain 192 and 178 SSRs, respectively [59, 60], despite differences in their genome organization. The large number and cumulative length of dispersed repeats in L. amoenum, including numerous repeats longer than 500 bp and 1 kb, may therefore be more relevant to its rearrangement potential than the total number of SSRs. This interpretation is consistent with the highly fragmented mitogenome of L. tsingtauense, in which repeat-mediated recombination has been proposed as an important driver of structural variation [34], and with C. giganteum, in which the accumulation of several repeat classes is associated with extensive genome expansion [35]. Nevertheless, repeat abundance demonstrates recombination potential rather than recombination activity. Direct confirmation will require long reads spanning alternative repeat junctions, quantitative assessment of structural isoforms, and comparisons among individuals or populations.

In contrast, plastid-derived DNA made only a limited contribution to the L. amoenum mitogenome. We detected 14 homologous chloroplast-derived fragments totaling 4,723 bp, representing only 0.60% of the genome. Given that plastid-to-mitochondrion DNA transfer is widespread but highly variable among angiosperms [65], this low proportion suggests that internal repeat proliferation and recombination, rather than extensive intracellular DNA transfer, was the primary force shaping mitogenome structure in L. amoenum.

In contrast to its structural plasticity, the coding features of the L. amoenum mitogenome were relatively conserved. Synonymous codons generally showed a preference for A- or U-ending codons, as also reported in L. tsingtauense, C. giganteum, F. ussuriensis, and F. taipaiensis [34, 35, 59, 60]. The persistence of similar codon preferences across mitogenomes with markedly different sizes and configurations suggests that codon usage and genome architecture are influenced by partly independent evolutionary processes. Genome structure is highly responsive to repeat-mediated recombination and intergenic sequence turnover, whereas codon usage in conserved protein-coding genes is more strongly constrained by nucleotide composition, mutational bias, and the requirements of mitochondrial translation. Nevertheless, RSCU patterns alone cannot distinguish the relative contributions of mutational pressure and translational selection, and further comparisons incorporating nucleotide composition and gene-expression data would be required to evaluate these mechanisms. RNA editing provides an additional link between conserved mitochondrial function and lineage-specific evolution. A total of 493 C-to-U editing sites were predicted in the PCGs of L. amoenum. This number is comparable to those reported for F. taipaiensis (488) and F. ussuriensis (505), but lower than those of L. tsingtauense (591) and C. giganteum (595) [34, 35, 59, 60]. Thus, RNA-editing abundance varies among Liliaceae species, although the overall magnitude remains relatively constrained. In L. amoenum, nad4 contained the largest number of predicted editing sites, followed by ccmB and ccmC. The enrichment of editing sites in genes involved in respiratory-chain activity and cytochrome c biogenesis across several Liliaceae species suggests that editing is maintained primarily by functional constraints on mitochondrial protein complexes. However, the identity of the most extensively edited gene differs among lineages: nad4 is predominant in L. amoenum and F. ussuriensis, whereas ccmB is a major editing hotspot in L. tsingtauense [34, 60]. This variation may reflect evolutionary changes in mitochondrial cis-recognition sequences and their interactions with nuclear-encoded editing factors, including pentatricopeptide repeat proteins [66]. Because most sites in the present study were predicted computationally and only five were validated by PCR and Sanger sequencing, conclusions concerning lineage-specific editing patterns remain provisional. Transcriptome-based validation across tissues and developmental stages, together with comparative analysis of nuclear-encoded editing factors, would provide a stronger mechanistic explanation.

Comparative synteny analysis revealed a marked decoupling between sequence conservation and structural evolution in Liliaceae mitogenomes. L. amoenum and L. tsingtauense shared 137 syntenic blocks, including 116 longer than 1 kb, with the longest block extending approximately 30.94 kb on L. amoenum Chr9. These conserved blocks indicate that substantial ancestral mitochondrial segments remain recognizable within Lilium, whereas their markedly different order and orientation point to extensive rearrangement following lineage divergence. Across Liliaceae, L. tsingtauense and C. giganteum exhibited the greatest cumulative syntenic coverage, while L. amoenum and L. tsingtauense shared the longest individual block. These measures capture complementary aspects of genome conservation: cumulative coverage reflects the retention of multiple fragmented homologous modules, whereas maximum block length indicates the preservation of a longer contiguous ancestral segment. Collectively, the coexistence of local sequence conservation, altered block order, abundant dispersed repeats, and multipartite organization is consistent with repeated recombination and rearrangement having substantially reshaped mitogenome architecture while preserving portions of the ancestral sequence [67–69]. However, these genomic patterns represent indirect evidence of recombination; direct confirmation will require long reads spanning alternative repeat junctions and quantitative assessment of structural isoforms [70].

In contrast to this structural dynamism, conserved mitochondrial genes retained a robust phylogenetic signal. Analysis of 27 mitochondrial protein-coding genes placed L. amoenum within Lilium with strong support, consistent with its accepted generic assignment. Published plastome studies provide complementary evidence for this placement and further clarify its relationships within the genus. The 152,280 bp plastome of L. amoenum exhibits the conserved quadripartite structure typical of Lilium [36]. Complete-plastome analysis of 22 Lilium species grouped L. amoenum with L. bakerianum and L. taliense [36], while a broader phylogeny of 70 wild taxa placed it in Clade IV together with L. bakerianum, L. primulinum var. ochraceum, L. nepalense, and L. taliense (BS = 85%) [30]. Because these species occur predominantly in the Himalayas and Hengduan Mountains, their plastid affinity may reflect a shared regional evolutionary history. An independent analysis of 48 complete Lilium plastomes further recovered L. amoenum as sister to L. bakerianum, strengthening evidence for their close maternal relationship [71]. These results also illustrate that, despite the high structural conservation of Lilium plastomes, SNPs, InDels, and variable non-coding regions provide sufficient information to resolve relationships among closely related species [29]. The consistent association of L. amoenum with L. bakerianum and other Himalayan-Hengduan species is particularly informative because its sectional assignment has varied among traditional classifications, including placement in Sinomartagon subsection 5c, section Oxypetalum, or section Lophophorum [30, 36]. This inconsistency indicates that floral morphology alone may not adequately reflect the evolutionary history of this lineage and highlights the need to reassess sectional boundaries using genomic evidence [72–74]. Overall, the mitochondrial and plastid results are congruent at the genus level and jointly support the placement of L. amoenum within Lilium. Nevertheless, its precise sister relationship cannot be evaluated without comparable taxon sampling across datasets. Moreover, mitochondrial and plastid genomes are generally inherited uniparentally and behave as largely nonrecombining linkage groups; consequently, they may reflect correlated organellar histories rather than independent estimates of the species tree [75]. Integrating broader taxon and population sampling with low-copy nuclear markers will therefore be essential for distinguishing shared species history from incomplete lineage sorting, hybridization, introgression, or organellar capture [72–74]. Such a cytonuclear framework will also be necessary to determine whether the pronounced mitogenomic structural variation in Liliaceae corresponds to recognized taxonomic lineages or instead reflects lineage-specific genomic dynamics.

Overall, the L. amoenum mitogenome reveals a coordinated evolutionary pattern in which repeat accumulation and recombination promote rapid structural reorganization, while core gene sequences, codon usage, and major RNA-editing targets remain comparatively conserved. This combination of structural flexibility and functional constraint provides a mechanistic framework for understanding mitochondrial genome diversification in Liliaceae. However, population-level structural validation, transcriptomic confirmation of RNA-editing sites, and phylogenetic integration of mitochondrial, plastid, and nuclear evidence will be required to test this framework more comprehensively.

Conclusion

In this study, we assembled the mitogenome of L. amoenum using a hybrid sequencing strategy that integrated Nanopore long reads and Illumina short reads. The mitogenome exhibited a distinctive multichromosomal architecture, spanning 789,335 bp across 14 separate chromosomes, including 13 circular contigs and one linear contig. A total of 55 genes were annotated, comprising 37 PCGs, 15 tRNA genes, and three rRNA genes. We further characterized codon usage bias, repetitive sequences, MTPTs, and syntenic relationships. Phylogenetic analysis based on the mitogenomes of L. amoenum and 25 other taxa confirmed its evolutionary placement within Lilium. In addition, 493 C-to-U RNA editing sites were identified, five of which were experimentally validated. Collectively, these findings provide valuable genomic resources for future studies on the evolutionary history, germplasm identification, and molecular breeding of Lilium.

Abbreviations

PCG

Protein-coding genes

tRNA

Transfer RNA

rRNA

Ribosomal RNA

SSRs

Simple sequence repeats

MTPTs

Mitochondrial plastid DNA transfer events

RSCU

Relative synonymous codon usage

ML

Maximum likelihood

Authors’ contributions

Y.G. offered the conceptualization; Y.G. and Z.L. wrote the main manuscript text; S.Y., S.L., and W.D. interpreted the data; Y.G., Z.L., W.D., and H.W. reviewed and revised the manuscript. All authors reviewed and approved the final manuscript.

Funding

Our work was funded by the Joint Special Project for Basic Agricultural Research of Yunnan Province (202301BD070001-144), the Academician Expert Workstation from the Yunnan Provincial Science and Technology Department (202605AF350078), and the Caiyun Postdoctoral Program Innovation Project (2025) of Yunnan Provincial Department of Human Resources and Social Security.

Data availability

The complete mitogenome and chloroplast genome of L. amoenum are available at GenBank (https://www.ncbi.nlm.nih.gov/) with the following accession numbers: PZ270155-PZ270168 and PZ270244. The raw sequencing data for the Illumina and Nanopore platforms have been deposited in NCBI (https:// www.ncbi. nlm. nih. gov/) with accession numbers PRJNA1452599, SAMN57239136, SRR38072817, and SRR38072818, respectively.

Declarations

Ethics approval and consent to participate

The authors declare that all methods were carried out in accordance with relevant guidelines and regulations. No specific permits are required for sample collection in this study. We comply with relevant institutional, national, and international guidelines and legislation for plant study.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.