1. Introduction

The sunflower family Asteraceae, the largest angiosperm family, encompasses approximately 26,000 named species across 1700 genera, 45 tribes, and 16 subfamilies (Mandel et al., 2019; Susanna et al., 2020; Zhang et al., 2024). This globally distributed family exhibits extraordinary ecological versatility, colonizing diverse habitats ranging from arid deserts to tropical rainforests, and displays remarkable morphological variation, including herbaceous, shrubby, and arboreal forms (Funk et al., 2009; Criado-Ruiz et al., 2024; Xu et al., 2025a, 2025b). Asteraceae species hold significant economic value as food crops (e.g., sunflowers and lettuce), medicinal plants (e.g., dandelion and yarrow), and ornamental species (e.g., daisies and chrysanthemums), while some have become notorious invasive weeds with widespread ecological impacts (Song et al., 2025). Their exceptional adaptive radiation and resilience in extreme environments render Asteraceae a compelling system for evolutionary, ecological, and horticultural research.

Genomic studies of Asteraceae plants have significantly advanced our understanding of their phylogeny and evolution, while facilitating the genetic improvement of crops, medicinal herbs, and ornamental varieties, as well as aiding in invasive species management. To date, plastid genomes (plastomes) have been sequenced for over 1300 species spanning at least 265 genera (NCBI Nucleotide, accessed December 5th, 2025), and no fewer than 57 nuclear genomes have been assembled (Schwacke et al., 2025; last accessed December 5th, 2025). These extensive genomic resources have enabled robust investigations into Asteraceae evolution, biosynthetic pathways, and genetic enhancement, with well-resolved phylogenetic relationships among subfamilies and tribes based on plastid and nuclear markers (Panero et al., 2014; Panero and Crozier, 2016; Zhang et al., 2021, 2024).

However, research on mitochondrial genomes (mitogenomes) in Asteraceae remains limited. The scarcity of published complete mitogenomes has hindered the exploration of mitogenomic variation and its implications for phylogeny, leaving gaps in our understanding of the family's evolutionary history. Notably, observed incongruences between plastid- and nuclear-derived phylogenies suggest potential discordant evolutionary signals, underscoring the need for mitogenomic data to provide a more comprehensive perspective on Asteraceae evolution. To date, only approximately 40 complete mitogenomes of Asteraceae species have been formally published—a number even lower than that of available nuclear genomes (e.g., Kozik et al., 2019; Fertet et al., 2021; Makarenko et al., 2021; Wu et al., 2022; Chen et al., 2023). Notably, most published mitogenomes are derived from Asteroideae, the largest subfamily, while representatives from other subfamilies, particularly early-diverging lineages, remain severely underrepresented. Current mitogenomic studies in Asteraceae have predominantly examined single or limited species (e.g., Jiang et al., 2023; Wu et al., 2023; Wang et al., 2024), leaving a critical gap in our understanding of mitogenome evolution across the entire family. Expanding mitogenomic investigations in Asteraceae will elucidate patterns of mitogenome evolution. Moreover, because plant mitogenomes encode numerous genes essential for oxidative phosphorylation (OXPHOS), and structural rearrangements in mitochondrial DNA can induce cytoplasmic male sterility (CMS) (Møller et al., 2021), a phenomenon widely exploited in crop breeding (e.g., see sunflower hybrid production) (Laver et al., 1991; Makarenko et al., 2018), expanded investigations will establish a foundation for future functional and applied research in this economically and ecologically significant plant family.

In this study, we newly assembled 38 mitogenomes from 12 subfamilies and 22 tribes in Asteraceae, including the first reported mitogenomes from 9 subfamilies and 3 additional tribes. Through comprehensive analyses of mitogenome size, structure, gene content, and RNA editing patterns, we characterized the diversity of mitogenomes across the family. Utilizing this expanded dataset, we reconstructed mitochondrial phylogenies using different sets of markers and evaluated their congruence with existing plastid and nuclear-based trees. Our results demonstrate substantial variation in mitogenome size across Asteraceae, with particularly compact mitogenomes observed in the Asteroideae subfamily. Comparative genomic analyses reveal both conserved features and dynamic evolutionary patterns in protein-coding sequences and non-coding regions. These findings provide new evolutionary insights and establish a valuable genomic framework for future studies of Asteraceae systematics and mitogenome evolution.

2. Material and methods 2.1. Sampling sequencing and data collection

We assembled mitogenomes for 38 Asteraceae species (Table S1), including 17 newly sequenced species collected across China. Their genomic DNA was extracted from silica-dried leaf tissue and sequenced on the Illumina HiSeq 4000 platform, generating 4–6 Gb of 150 bp paired-end reads per sample. The remaining 21 mitogenomes were reconstructed from publicly available sequencing data in the NCBI SRA database (https://www.ncbi.nlm.nih.gov/sra). We also included long-read sequencing data (Oxford Nanopore and/or PacBio SMRT) for four species: Arctium lappa, Carthamus tinctorius, Cynara cardunculus, and Tagetes erecta. In our analyses, we incorporated another nine published mitogenomes from GenBank, resulting in a dataset of 47 Asteraceae species. All the accession numbers for mitogenomes and plastomes are provided in Table S1.

2.2. Mitogenome assembly and annotation

We employed an organelle genome assembly pipeline similar to our previous study (Fig. S1; Feng et al., 2021). For Illumina short-read data, we first performed quality control using TRIMMOMATIC v.0.36 (Bolger et al., 2014) to remove low-quality bases, followed by de novo assembly with SPAdes v.3.13.1 (Bankevich et al., 2012). For the four species with long-read data (Arctium lappa, Carthamus tinctorius, Cynara cardunculus, and Tagetes erecta), we conducted hybrid assembly using Flye v.2.9.6 (Kolmogorov et al., 2019). Then mitochondrial and plastid contigs were extracted from the total contigs by BLAST v.2.12.0 search (Camacho et al., 2009) with default parameters, by aligning against reference mitogenomes and plastomes of Asteraceae, respectively. Candidate contigs were further manually screened based on read coverage and preliminary protein-coding gene annotation in Geneious R10 (Biomatters, Inc.). The plastomes were assembled by mapping sequencing reads to plastid contigs in Geneious. The initial contigs were iteratively extended through careful examination of overlapping flanking regions until achieving complete scaffold continuity. For plastome circularization, we first identified the characteristic inverted repeat (IR) regions by detecting their conserved boundary sequences, then manually verified and joined these regions to complete the circular genome structure.

Mitogenomes contain more repetitive sequences and mitochondrial plastid DNAs (MTPTs), resulting in more fragmented SPAdes contigs. We first pre-annotated our mitochondrial contigs and excluded the plastid contigs (they can also be retrieved due to the presence of MTPTs). Then using the "Annotate from Database" function of Geneious, we pre-annotated the selected mitochondrial contigs to identify present genes with a known Asteraceae mitogenome as the reference. We used two approaches to ensure all mitochondrial contigs were retrieved from BLAST searches. 1) We extracted all contigs longer than 500 bp with a similar coverage (including those 2–3 times higher than the single-copy). This approach excludes nuclear contigs having similar coverage through read mapping as their coverage is always unbalanced with high divergent reads. 2) We compiled a reference set containing all known angiosperm mitochondrial protein coding and rRNA genes. For genes not initially detected, we performed additional validation by mapping raw reads to the reference gene sequences and analyzing coverage patterns to distinguish between real gene loss and simply missed during our mitochondrial contigs selection. This approach identifies the missing contigs from the complete SPAdes assembly. This validation ensured complete representation of mitochondrial genes in our final assembly. Next, we assembled the mitogenome in an iterative mapping-extension approach similar to the plastome. Repeat regions and plastid insertions can be detected by their coverage. For example, a two-copy repeat exhibited double coverage of single-copy regions and MTPT may be 5 or 10 times higher than the normal mitochondrial sequences as the plastid reads can be mapped. During the manual extension, we resolved repeat regions through coverage and MTPT identified by mapping to the corresponding plastome (Fig. S1). We iteratively repeated the mapping-extension process until no further contig linkages could be established. Finally, mitogenomes from 30 species were assembled as 1–4 circular chromosomes (Table S1 and Fig. S2), while the other 8 species obtained linear chromosomes as we failed to find a way to connect the terminal repeats or MTPTs (Table S1 and Fig. S3). All assembled mitogenomes had a sequencing coverage over 50.

2.3. Gene and sequence annotations

The protein-coding and rRNA genes were annotated based on the annotations of known mitogenomes (e.g., Helianthus annuus, GenBank ID: NC_023337 and, GenBank ID: MK642355) with "Annotate from Database" with a similarity cutoff 70% in Geneious software. The tRNAs were predicted using tRNAscan-SE v2.0 (Chan and Lowe, 2019). RNA editing sites were predicted using Deepred-Mt (Edera et al., 2021), with sites exhibiting a probability greater than 90% retained. The long non-coding RNA (lncRNA) sequencing data of Helianthus annuus (SRA ID: SRR10984382) and Lactuca sativa (SRA ID: SRR28019647) were used to validate the coding region annotations and the RNA editing sites. MTPTs were identified by BLASTN against a set of Asteraceae plastomes (identity > 90%, length > 100 bp; Table S2). To characterize the distribution of MTPTs, we mapped all identified MTPTs to the plastome of Pertya phylicoides in Geneious (Fig. S4A; Geneious format read mappings are available on FigShare). Repeats were detected by BLASTN against the assembled mitogenome (identity > 95%, length > 30 bp), with only one representative of each repeat pair counted in the total repeat length (Table S2).

2.4. Comparative analysis of mitogenome size

We conducted comparative analyses of mitogenome sizes across angiosperms using complete mitogenomes retrieved from the NCBI Nucleotide database (https://www.ncbi.nlm.nih.gov/nuccore/), retaining only entries annotated as "complete" and including a single representative sequence per species to avoid redundancy. The final dataset, comprising mitogenomes from another 11 extensively sequenced families, i.e., Orchidaceae, Poaceae, Salicaceae, Fabaceae, Rosaceae, Malvaceae, Brassicaceae, Apiaceae, Lamiaceae, Solanaceae, and Convolvulaceae (Table S3), was analyzed and visualized using the Ggplot2 package (Wickham, 2016) in R software (https://www.r-project.org).

2.5. Sequence synteny analysis

To identify homologous regions, we first fragmented mitogenomes into 200 bp segments following the approach of Lin et al. (2022), then conducted pairwise comparisons using BLASTN with default parameter. The mitogenome synteny between and within the subfamilies were visualized by Python version MCscan of JCVI utility libraries v.1.1.17 (Tang et al., 2008).

2.6. Phylogenetic analysis

We reconstructed mitochondrial phylogenies using three distinct marker datasets. The first dataset comprised 36 coding sequences (CDS), including 24 core mitochondrial genes (atp, ccm, cob, cox, nad, mttB, and matR) along with large and small subunits (rpl/s) and succinate dehydrogenate genes (sdh). The second dataset contained 35 protein-coding genes (with intronic regions) and 3 rRNAs, excluding matR due to its inclusion within the nad1 intron. For the third dataset, we identified 53 homologous regions by using Ainsliaea angustata as a reference for BLASTN searches against all taxa (default parameter), followed by extraction and mapping of hits in Geneious to select conserved blocks present in most species (Fig. S4B). For each block, if there were more than one hits from the same species, we only retained the longest one. All sequences were aligned using MAFFT v.7 (Katoh and Standley, 2013) in "auto" mode, concatenated into one super matrix per dataset, and used to build a maximum-likelihood (ML) tree through IQ-TREE 2.0.7 (Nguyen et al., 2015) with the "GTR + I + G" model. Mitogenomes of Daucus carota (GenBank: NC_017855, Apiaceae), Apium leptophyllum (GenBank: MZ328723, Apiaceae), and Panax ginseng (GenBank: MW029460, Araliaceae) were used as outgroups.

We also constructed a plastid phylogeny using 79 CDS following identical construction procedures. Plastomes from Saposhnikovia divaricata (GenBank: MZ089852, Apiaceae), Apium leptophyllum (GenBank: MZ328721, Apiaceae), and Heptapleurum heptaphyllum (GenBank: KT748629, Araliaceae) were used as outgroups.

2.7. Protein structure prediction and comparison

Protein structures were predicted by AlphaFold 3 server (Abramson et al., 2024) and the visualization and comparison were done in UCSF ChimeraX v.1.7.1 (Meng et al., 2023).

2.8. Sequence analysis of mitochondrial DNA content

Due to the non-conserved nature of plant mitochondrial DNA, comparing mitogenome similarity and variation patterns across species presents significant challenges. To address this, we developed a comprehensive segment-based approach by constructing a library containing all Asteraceae mitochondrial DNA segments (including both conserved and non-conserved sequences) and analyzing their presence/absence patterns. Our analysis began by dividing all Asteraceae mitogenomes into consecutive segments (scanning with a sliding window of 200 bp and a step size of 200 bp), yielding 74,442 initial fragments. These segments were then processed using CD-HIT-EST v.4.8.1 (Li and Godzik, 2006) with an 80% nucleotide identity threshold to reduce redundancy, resulting in 19,606 unique representative segments that capture the full diversity of Asteraceae mitochondrial DNA (note: some overlapping segments were retained as they did not affect our results or conclusions). We then performed BLAST searches using this fragment library against both Asteraceae and 43 additional angiosperm mitogenomes (Table S4). For each segment, we determined its presence across species by calculating the cumulative coverage depth of all BLAST hits (disregarding sequence similarity). The resulting presence/absence patterns were visualized as a heatmap using the pheatmap package in R, with color gradients indicating relative coverage levels across species.

3. Results 3.1. Asteraceae mitogenomes vary in size and content

We assembled mitogenomes of a total of 38 species from the Asteraceae family, representing 12 subfamilies and 22 tribes (Fig. 1A and Table S1), which account for approximately 75% and 49% of the known subfamilies and tribes, respectively. Our sampling encompasses remarkable morphological diversity, including herbs, trees, and shrubs (Fig. 1B). These assemblies significantly expand the current mitogenomic resources for Asteraceae, especially for the basal lineages. Notably, our dataset includes the first reported mitogenomes for nine subfamilies (Barnadesioideae, Famatinanthoideae, Mutisioideae, Stifftioideae, Gochnatioideae, Wunderlichioideae, Pertyoideae, Gymnarrhenoideae, and Vernonioideae) and another three tribes from Asteroideae (Millerieae, Tageteae, and Gnaphalieae). By incorporating an additional nine mitogenomes from GenBank (Fig. 1A and Table S1), we analyzed a comprehensive dataset of 47 mitogenomes from this family. This represents the most extensive mitogenomic resource for Asteraceae to date, providing unprecedented insights into the evolution of their mitochondrial genomes.

The assembled mitogenomes consisted of 1–4 chromosomes per species, with a total of 40 circular chromosomes from 30 species (Fig. 1A and Table S1). Our circular assemblies are natural loops (i.e., contig termini were linked by overlapping reads) rather than any artificially circularize linear contigs or pseudochromosomal scaffolds (Fig. S2). Therefore, they can be seen as "complete" (disregarding potential mitochondrial plasmids; Warren et al., 2016). The remaining 9 chromosomes from 8 species were assembled as linear molecules, as terminal repeats or MTPTs prevented circularization (Fig. S3). While these linear assemblies may theoretically lack small repetitive or MTPT regions, their lengths are comparable to or even exceed those of closely related species (Fig. 1A), suggesting near-completeness. To validate assembly reliability, we provide read-mapping files demonstrating uniform coverage depth across circular and linear chromosomes (Figs. S2 and S3; Geneious format read mappings are available on FigShare).

The mitogenomes of Asteraceae species ranged from 194 to 427 kb in size with a GC content of approximately 45% (Fig. 1A and Table S1). Notably, the large mitogenomes (> 350 kb) were predominantly found in basal lineages, with the exception of Pentaphorus foliolosus. In contrast, derived taxa generally possessed smaller mitogenomes. Striking size reduction was observed in certain Asteroideae tribes (e.g., Anthemideae, Coreopsideae, and Gnaphalieae), where mitogenomes were as compact as 200 kb—approximately half the size of the largest assemblies in the family. This size variation suggests a potential mitogenome contraction event during Asteraceae evolution, particularly during the transition from basal lineages to Asteroideae. Remarkably, these reduced mitogenomes in Asteroideae are among the smallest reported in non-parasitic angiosperms, comparable in size to those of Brassicaceae (Fig. 2A). This pattern may contrast with the trend observed in most angiosperms, which typically exhibit larger mitogenomes (Fig. 2A).

Asteraceae mitogenomes contain relatively low repeat content (> 30 bp), accounting for no more than 16% of the total size (Table S2). Longer mitogenomes generally exhibit higher repeat content, except Pertya phylicoides (Fig. 1A). Notably, mitogenomes from the Asteroideae subfamily, particularly those in Anthemideae and Astereae tribes, show significantly reduced repeat content. Asteraceae mitogenomes also display minimal MTPTs, typically containing fewer than 10 MTPTs that collectively represent less than 3% of the mitogenome length. Striking examples include Achillea wilsoniana and Bidens pilosa, each possessing only a single MTPT fragment shorter than 300 bp (Table S2). Some MTPTs are shared among species, particularly those originating from the inverted repeat (IR) and the psaB regions (Fig. S4A).

3.2. Structural variations are rampant in Asteraceae mitogenomes

We investigated the structural diversity of mitogenomes across the Asteraceae family. Firstly, we selected representatives from seven subfamily and analyzed the collinearity. We found that mitogenomes from different subfamilies exhibited extensive rearrangements (Fig. 2B). In addition to intergenic regions, our analysis identified non-coding sequences that maintain homology across subfamilies. These elements, while limited in number, may represent valuable phylogenetic markers for resolving deep evolutionary relationships within Asteraceae.

We then examined the structural variations of mitogenomes within the six multi-species subfamilies (Fig. S5) and among species within each tribe of the Asteroideae subfamily (Fig. S6). More and longer homologous blocks were identified in these groups than at the subfamily-level, although rearrangements were prevalent and many regions that were not conserved could not be aligned. When it came to the genus level, the mitogenomes were nearly identical, with a few rearrangements (Kozik et al., 2019; Fertet et al., 2021b; Makarenko et al., 2021). These findings suggest that, similar to other angiosperms, the non-coding sequences of Asteraceae mitogenomes underwent rapid structural evolution. Additionally, our hierarchical analysis revealed that closely related species tended to exhibit similar mitogenomic structures. Over time, however, homologous sequences diminished, whereas recombination events increased.

3.3. Asteraceae mitochondrial genes maintain angiosperm traits with unique lineage variations

All examined mitogenomes contained 30–35 protein-coding genes, 3 rRNA genes, and 15–30 tRNA genes. The 24 core mitochondrial genes (including all atp, cob, cox, matR, mttB, and nad genes) were conserved across all species, consistent with patterns observed in most angiosperms. These core genes exhibited high sequence conservation with low substitution rates. Variations within core genes were also observed. For example, in Synotis nagensium, the 3′-end of atp1 was replaced by an unknown sequence (Fig. 3A), resulting in complete loss of the typical atp1 3′-terminus. AlphaFold 3 prediction indicated this alteration modifies the C-terminal protein structure (Fig. 3B), although the functional consequences remain unclear. Additional mutations were also observed in atp6 among some species.

The intron-exon structure largely followed typical angiosperm patterns, with introns present in 8 genes (ccmFc, cox2, nad1, nad2, nad4, nad5, nad7, and rps3). Of these, three genes (nad1, nad2, and nad5) contained trans-splicing introns, and the matR gene was located within nad1 exons 4 and 5. Notably, Asteraceae species possess a nad4 gene with only one intron, matching other Asterales but differing from Apiales – suggesting loss of the last two introns in the Asterales common ancestor (Fig. S7A). The transcriptome data confirmed that Asteraceae cox2 should contain two introns (Fig. S7B), correcting previous annotations in some published mitogenomes (e.g., Lactuca sativa [GenBank MK642355] and Paraprenanthes diversifolia [GenBank MN661146]) that reported only one intron (Fig. S7C). An exception was found in Artemisia, which has lost the first cox2 intron (Fig. S7C).

Although horizontal gene/DNA transfer (HGT) is commonly observed in parasitic plants from their hosts, it has also been detected in autotrophic plants, such as in the basal angiosperm Amborella (Rice et al., 2013). In Asteraceae, we identified a highly divergent region of cox1 in Arctium lappa (Fig. 3C). Comparison with other angiosperms revealed that this region is most similar to Amaranthaceae. Phylogenetic analysis of this region confirmed an HGT event, as Arctium was distantly related to its Asteraceae relatives but clusters together with Amaranthaceae species (Fig. 3D). Interestingly, a potential inter-family HGT was also observed. The ccmFn gene in Carthamus tinctorius shared a common deletion and several identical sites with species in the Anthemideae, suggesting a possible origin from this tribe (Fig. 3E). The deletion exhibited microhomology at both ends, indicating it may have resulted from microhomology-mediated end joining (MMEJ) following a double-strand break (Garcia-Medel et al., 2019) (Fig. 3F). This highlighted the occurrence of DNA exchange between mitochondria even among closely related plant species.

The ribosomal protein genes (rpl and rps) and sdh exhibited a higher frequency of gene losses and demonstrated greater sequence divergence across species (Fig. S8). Of these, only rpl5, rpl10, and rps13 were conserved in all investigated species. The loss of these genes may be attributed to their potential transfer to the nuclear genome (Adams and Palmer, 2003) or their functional replacement by dual-targeting nuclear-encoded genes (Kubo and Arimura, 2010).

3.4. Asteraceae mitogenomes exhibit lineage-specific losses of canonical RNA editing sites

RNA editing is a crucial post-transcriptional modification mechanism and mediates nucleotide alterations in RNA molecules, generating transcripts that diverge from their genomic DNA templates (Takenaka et al., 2008). In plant mitochondrial genomes, RNA editing primarily involves cytosine-to-uracil (C-to-U) conversions, with occasional reverse U-to-C editing events (Knoop et al., 2011). This molecular process plays an essential role in maintaining mitochondrial gene functionality by restoring conserved codons, rectifying missense mutations, and optimizing the structural and functional integrity of mitochondrial transcripts (Takenaka et al., 2008).

We systematically predicted RNA editing sites in protein-coding genes across Asteraceae species. Our analysis revealed that most Asteraceae species maintain approximately 500 RNA editing sites (Table S5). Validation using lncRNA data from Helianthus annuus and Lactuca sativa confirmed the accuracy of our predictions, although a few sites were not predicted (likely due to the stringent cutoff parameters employed in our analysis). Given the limited transcriptome data available for most species, we did further comparative analyses based on these prediction results to maintain consistency across taxa.

We observed a near-complete loss of RNA editing in the atp6 gene across the family, with only Cichorieae, Astereae, and Anthemideae members and Carpesium leptophyllum retaining a single conserved editing site (Table S5). This pattern of atp6 editing loss appears to be a derived characteristic in Asteraceae, as it contrasts markedly with the typical conservation observed in other Superasterids (Fig. S9A). Similarly, the nad4 gene exhibited significantly reduced editing (~18 sites) compared to Apiales (~45 sites) (Table S5). This reduction correlates with the Asterales-specific loss of two nad4 introns (Fig. S7A), supporting the hypothesis that retroprocessing mechanisms (Cuenca et al., 2016) may have contributed to both intron loss and editing site elimination in this lineage.

A particularly striking case was observed in Erigeron annuus, which exhibited a dramatic reduction in RNA editing, retaining only less than 300 sites (Fig. S9B and Table S5) – significantly fewer than other Asteraceae species. The pressures and molecular mechanism driving this exceptional case in Erigeron annuus warrant further investigation.

3.5. Mitochondrial phylogeny unravels complex evolutionary histories in Asteraceae

We constructed the phylogeny of Asteraceae using three distinct matrices: (1) 36 CDS (30,538 columns, including 1562 parsimony-informative sites, PIS), (2) 38 genes (35 CDS + 3 rRNA + introns; 56,778 columns, 2858 PIS), and (3) 53 conserved DNA blocks shared by most species (204,910 columns, 9029 PIS). The 53-DNA-block matrix yielded the highest resolution and was used to infer the Asteraceae phylogeny (Fig. 4). In contrast, the phylogenies derived from the 36-CDS and 38-gene matrices showed lower resolution and contained more anomalous clades unsupported by other evidence (Fig. S10).

Relationships among the 22 tribes were generally well-resolved, and the monophyly of all the tribes containing two or more sampled species received maximal support (Fig. 4). Barnadesieae (Barnadesioideae) was strongly supported as the basal-most tribe, followed by Famatinantheae as the next divergent lineage. Hyalideae and Stifftieae were moderately supported as sister groups and formed a clade sister to Mutisieae with moderate support too. These three tribes formed the third divergent clade of the family and the basal-most clade of the core Asteraceae, with the monophyly of the core Asteraceae being strongly supported (BS = 85). Wunderlichieae and Gochnatieae were strongly supported as sister groups (BS = 99) and formed the second branch of the core Asteraceae with maximal support. Pertyeae, Cardueae, and Gymnarrheneae diverged separately. Our results maximally supported the sister relationship of Cichorieae and Vernonieae, with them forming a clade as the sister of Asteroideae. Within Asteroideae, two major clades were recognized, with one clade containing four tribes belonging to the supertribe Senecionodae and the other containing six tribes belonging to the supertribe Helianthodae. Also, the relationships among these tribes were strongly supported with BS ≥ 90. Of the four Senecionodae tribes, Senecioneae and Anthemideae were the first and second clades to diverge, with Gnaphalieae and Astereae being resolved as sister groups. Of the Helianthodeae tribes, Inuleae was the first clade to diverge, and the Tageteae tribe and its sister Heliantheae formed the second divergent branch. Coreopsideae and Heliantheae were shown as sister groups, which were together sister to Millerieae.

The mitochondrial phylogenetic relationships within Asteraceae are generally consistent with those congruent in previous plastid and nuclear phylogenetic studies (Fig. 4) (Panero and Funk, 2008; Panero et al., 2014; Fu et al., 2016; Panero and Crozier, 2016; Zhang et al., 2024). Where the relationships of clades differ between the plastid and nuclear phylogenetic trees, our mitochondrial results are generally consistent with one of them. Both the mitochondrial and plastid trees strongly supported the sister relationship between Cichorieae and Vernonieae, which was also supported by previous plastid studies (Fig. 4) (Panero and Funk, 2008; Panero and Crozier, 2016). However, previous nuclear phylogenomic studies indicated that Vernonieae diverged earlier from the backbone of Asteraceae than did Cichorieae (Fig. 4) (Zhang et al., 2021, 2024). Furthermore, in our mitochondrial tree and previous nuclear trees, Stifftieae and Hyalideae are sister groups, whereas they were placed in distinct clades in our plastid tree and previous plastid phylogenetic studies (Fig. 4) (Panero and Funk, 2008; Panero and Crozier, 2016; Zhang et al., 2021, 2024). Also, our mitochondrial tree and previous nuclear phylogenetic studies consistently showed that Pertyeae diverged earlier than Cardueae, but our plastid tree and previous plastid studies supported the opposite (Fig. 4) (Panero and Funk, 2008; Panero and Crozier, 2016; Zhang et al., 2021, 2024). Within Asteroideae, Gnaphalieae were recognized as the sister of Astereae in both our mitochondrial results and previous nuclear phylogenetic studies, but our plastid tree and previous plastid studies did not support their close relationship (Fig. 4) (Panero and Funk, 2008; Panero and Crozier, 2016; Zhang et al., 2021, 2024). In addition, our mitochondrial and previous nuclear phylogenetic trees strongly support the sister relationships of Heliantheae and Coreopsideae, but they were not sister groups in our plastid tree and previous plastid analyses (Fig. 4) (Panero and Funk, 2008; Panero and Crozier, 2016; Zhang et al., 2021, 2024). Furthermore, some relationships found in our mitochondrial tree did not agree with either the plastid trees or the nuclear trees. For example, our mitochondrial tree strongly supports the sister relationship between Wunderlichieae and Gochnatieae, but they were placed as distinct lineages in plastid results and nuclear phylogenetic trees (Fig. 4) (Panero and Funk, 2008; Panero and Crozier, 2016; Zhang et al., 2021, 2024).

3.6. Small mitogenomes in Asteroideae contain less ancestral mitochondrial DNA

As previously demonstrated, the mitogenomes of Asteraceae exhibit significant size variation, with structural and sequence similarities decreasing as kinship becomes more distant, such as from the genus to the subfamily level (Figs. 2, S1, and S2). Although it is well-known in angiosperms that the non-coding DNA of mitogenomes generally varies, a comprehensive analysis of this dynamic process and potential origins remains underexplored. Here, we employed a new approach to detect the DNA content in Asteraceae (Fig. 5A). First, all the 42 Asteraceae mitogenomes were scanned with a sliding window of 200 bp and a step size of 200 bp, resulting in a total of 74,442 DNA fragments. Given the slow evolutionary rate of mitochondrial DNA sequences, many of these fragments are expected to share high identity. Thus, the redundancy was then reduced by applying a nucleotide identity cutoff of 0.8, yielding a dataset of 19,606 "non-redundant" mitochondrial DNA fragments. This dataset collectively represents the "total DNA" of Asteraceae mitogenomes. Finally, each fragment was compared against the mitogenomes of Asteraceae and other angiosperms (listed in Table S3) to assess its presence (measured by percentage) and to analyze the potential origin and HGT events.

We ultimately performed clustering analysis on the presence of these fragments across the plants and visualized the results in a heatmap (Fig. 5B). This map provides a clear visualization of the variation and dynamic evolutionary changes in Asteraceae mitogenomes. Based on the presence of mitochondrial DNA in Asteraceae, these sequences can be categorized into three main types.

Approximately 10% of these sequences are present in all Asteraceae species with only few exceptions. They are also be found in other angiosperms, although sometimes with a lower similarity. These represent the most conserved parts of the mitogenome in angiosperms, including the conserved genes (including introns) and their flanking regions. Excluding the influence of repetitive sequences, their contribution to mitogenome size variation is minimal in Asteraceae.

Another approximately 35% of the sequences are shared among most Asteraceae, although absence is also observed in some species. Unlike the first category, these sequences are rarely found in other angiosperms. Thus, they are presumed to be the ancestral sequences inherited from the common ancestor of the Asteraceae family. They have undergone multiple independent losses in different species. These sequences vary significantly in species and are positively related to mitogenome length. Many species in the subfamily Asteroideae have short mitogenomes, and they have evident reductions in these sequences (e.g., see species in Anthemideae and Coreopsideae). In contrast, other subfamilies with longer mitogenomes maintain a more complete set of these DNA segments. The two tribes with the longest mitogenomes in Asteroideae, Inuleae and Senecioneae, also contain more segments but with distinct features. Moreover, the map can also serve as a tool for assessing mitogenome identity. For example, species within the Millerieae and Mutisieae exhibit high similarity in DNA segment composition, indicating a high similarity in the mitogenomes, which aligns with the results of our mitogenomic collinearity analysis (Fig. S2).

The remaining approximately 55% of sequences are predominantly found in one or a few species, representing species- or lineage-specific sequences. Although these sequences possess a large proportion of the total Asteraceae mitochondrial DNA, their presence is rare among species. That explains why mitogenomes always have a large number of distinct sequences between closely related species. These segments are potential candidates for the novel sequences created from mutation events, or mobile elements obtained from other genomes. However, except for some short fragments, no long specific sequences were detected in the selected angiosperms. This may suggest that either HGT events are highly fragmented or they were from the species not included in this study.

4. Discussion

Mitochondria, known as the powerhouses of the cell, are pivotal organelles that furnish organisms with energy through the process of respiration (McBride et al., 2006). Despite their importance, research on plant mitogenomes has progressed more slowly than studies on plastomes and, at times, even lags behind nuclear genomic investigations. In this study, we aimed to address this gap within the Asteraceae family by assembled the mitogenomes of 38 species, including the first-ever reports for 9 subfamilies and 3 additional tribes. Our findings reveal distinctive features of Asteraceae mitogenomes. Similar to other angiosperms, they exhibit considerable structural variation, low nucleotide substitution rates, and a stable gene repertoire (Knoop et al., 2011; Wynn and Christensen, 2019). Asteraceae mitogenomes also display many unique characteristics, such as generally small genome sizes, few MTPTs, occurrences of inter- and intra-familial HGT, and specific RNA editing site and intron losses. These distinctive attributes provide a foundational resource for future omics and phylogenetic studies in Asteraceae.

In angiosperms, mitogenomes typically range from 400 to 600 kb in length (Mower, 2020). However, mitogenomes in Asteraceae are generally smaller than this, with those in the subfamily Asteroideae being particularly compact (some as small as approximately 200 kb), ranking among the smallest known in photosynthetic angiosperms (Fig. 2A). Through sequence segment analysis (Fig. 5), we observed that ancestral DNA inherited from the common ancestor of Asteraceae has undergone differential loss, significantly influencing mitogenome size variation. This finding offers important insights into the diversity within the family and serves as a valuable reference for understanding mitogenome evolution in other plant groups. Notably, the short mitogenomes in Asteroideae also show a relative scarcity of repeat sequences, contrasting with the pattern seen in most angiosperms (Wynn and Christensen, 2019). This evidence suggests that Asteroideae species may possess mechanisms that suppress the uncontrolled expansion of non-coding DNA, including repeats and MTPTs.

We constructed a family-scale mitochondrial phylogeny of Asteraceae. The tree based on mitochondrial sequences is largely congruent with topologies derived from plastid and nuclear data, indicating a broadly consistent evolutionary history among the three genomic compartments in Asteraceae (Panero and Funk, 2008; Huang et al., 2016; Panero and Crozier, 2016; Mandel et al., 2019; Zhang et al., 2021, 2024). Nevertheless, several discordances were observed. First, in some cases, the mitochondrial and plastid phylogenies supported the same topology, which was not recovered in nuclear trees. For example, Vernonieae and Cichorieae were resolved as sister groups in both mitochondrial and plastid trees, whereas nuclear data suggested that Vernonieae diverged earlier than Cichorieae. This pattern is consistent with the shared maternal inheritance of mitochondrial and plastid genomes and may reflect the influence of past hybridization events on the evolutionary history of these tribes (Dong et al., 2025). In other instances, mitochondrial and nuclear data agreed with each other but conflicted with plastid-based relationships—such as the placements of Stifftieae and Hyalideae, and the relative divergence order of Pertyeae—implying a distinct evolutionary trajectory for Asteraceae plastomes. Moreover, our mitochondrial phylogeny also revealed several clades not supported by either plastid or nuclear trees, pointing to a unique evolutionary history of mitochondrial genomes in this family.

Our study underscores the utility of mitochondrial markers for phylogenetic reconstruction in Asteraceae, while also revealing several strongly supported incongruences among trees derived from the three genomic compartments. Notably, certain conflicts (particularly those between mitochondrial and plastid genomes despite their shared maternal inheritance) highlight the complex evolutionary dynamics operating among the three genomes. Although some discordances, such as the relationship between Cichorieae and Vernonieae, may be explained by hybridization, others cannot be readily attributed to either hybridization or incomplete lineage sorting, the two mechanisms most frequently invoked to explain cytonuclear discordance (Bruun-Lund et al., 2017; Lee-Yaw et al., 2019; Morales-Briones et al., 2021; Fu et al., 2022). The systematic approaches for resolving deep phylogenetic conflicts, as demonstrated in recent angiosperm-wide studies (Nie et al., 2026), can provide valuable frameworks for interpreting such complex evolutionary histories in Asteraceae. Furthermore, while not explored in this study due to their high conservation within the family, variation in RNA editing sites may also influence phylogenetic inferences and represents a potential direction for future research (Dong et al., 2023).

Our assembly strategy incorporated both short- and long-read sequencing data. A comparison of the two approaches revealed that short-read assemblies may introduce artificial rearrangements when resolving repeat regions longer than the insert size. This did not compromise the overall completeness of our mitogenomes, as most chromosomes were assembled as circular units. It is also important to note that plant mitogenomes may not adopt a single "standard structure." Instead, they are highly dynamic, with recombination between repeats often leading to multiple coexisting isoforms within individuals (Palmer and Herbon, 1988; Woloszynska, 2010; Kozik et al., 2019). This structural plasticity can sometimes make long-read assembly challenging, as the same single-copy region may be resolved into multiple configurations, complicating the reconstruction of a consensus structure.

Acknowledgements

This study was supported by the National Natural Science Foundation of China (No. 32000158 and 32400188), the National Science & Technology Fundamental Resources Investigation Program of China (No. 2021XJKK0702), and the Hunan Provincial Natural Science Foundation of China (No. 2025JJ60203). We thank Jiahao Shen from Nanjing Zhongshan Botanical Garden and Caifei Zhang from Wuhan Botanical Garden, Chinese Academy of Sciences for providing some of the plant photographs.

Data availability

The mitogenomes assembled in this study have been deposited to GenBank under accessions OQ420717–OQ420738 and CNGBdb (https://db.cngb.org/) under project CNP0004742. The Geneious format read mapping files were saved in FigShare (https://doi.org/10.6084/m9.figshare.29860013).

CRediT authorship contribution statement

Zhixi Fu: Roles/Writing - original draft, Writing - review & editing, Methodology, Conceptualization, Data curation, Resources, Software. Penghao Yang: Software, Visualization. Jiazhen Wu: Data curation, Resources, Software. Guojin Zhang: Roles/Writing - original draft, Writing - review & editing, Methodology, Visualization. Yanlei Feng: Roles/Writing - original draft, Writing - review & editing, Methodology, Conceptualization, Data curation, Software, Visualization.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.pld.2025.12.013.