Complete mitochondrial genomes of the two high-valued bromeliads: comparative genomics and evolutionary insights

Abstract

Background

The Bromeliaceae family contains many horticulturally and economically important species, including Aechmea fasciata and pineapple (Ananas comosus). Compared to their well-studied chloroplast and nuclear genomes, the mitogenomes of bromeliads remain largely unexplored, hindering a full understanding of their organellar evolution and phylogenetic relationships. This study provides the first complete mitogenome assemblies and a comparative analysis for these two high-value bromeliads, revealing their phylogenetic affinities.

Results

Our findings suggested that the mitogenomes of Ae. fasciata and An. comosus both exhibit a simple master circle structure. They were 0.93–1.16 Mb in length, containing similar protein-coding genes (PCGs) (42–44), 39 tRNA genes, and three rRNA genes. Notably, the synonymous codons usage analysis revealed a preference for A/U endings, and a total of 542–548 RNA editing sites were detected. Most PCGs were under purifying selection, indicating a strong evolutionary constraint to preserve functional integrity. Additionally, the ATP synthase gene exhibited high nucleotide diversity, suggesting that such genes may need further investigation in the context of future CMS-related studies. Phylogenetically, analyses based on shared PCGs provided robust support for the clustering relationships within Bromeliaceae, confirming that An. comosus and Ae. fasciata form a distinct clade and are separate from Puya raimondii.

Conclusions

This study establishes a crucial comparative genomic framework for Bromeliaceae, providing valuable genetic resources for future evolutionary studies and potential marker-assisted breeding applications.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12864-026-12841-5.

Introduction

The Bromeliaceae family, comprising 58 genera and approximately 3,140 species, represents a morphologically distinct, ecologically diverse, and species-rich clade of flowering plants native to the New World tropics and subtropics, including a number of edible and ornamental crops [1–3]. Bromeliads hold significant economic and horticultural value. For example, the cultivated pineapple (Ananas comosus) is the world's third most important tropical fruit, with a centuries-long cultivation history and an annual global production of about 24.8 million tons [4, 5]. Pineapple also possesses nutritional value and diverse medicinal properties, largely attributed to the bromelain enzyme complex [5]. Previous studies have demonstrated its broad therapeutic potential, including antidiarrheal, anticoagulant, and immunomodulatory effects [6–8]. However, under natural conditions, bromeliads exhibit limited speciation through reproduction [9]. Consequently, the ornamental industry, which regards bromeliads as a key horticultural resource, depends not on natural diversification but on structured breeding programs to develop new hybrids with enhanced floral and foliar traits [10, 11].

As essential organelles, plant mitochondria are central to cellular energy production and are deeply involved in critical processes such as metabolic regulation, cell differentiation, and programmed cell death [12, 13]. Due to their typical maternal inheritance and structurally complex, highly variable genomes, mitochondrial DNA has become a powerful tool for population genetics, phylogenetic studies, and genetic breeding [12, 14–16]. Historically, accurate assembly and annotation of plant mitogenomes have been challenging due to their considerable size variation, dynamic structural rearrangements, and abundant repetitive sequences. Fortunately, recent advances in high-fidelity long-read sequencing have effectively overcome these obstacles [17, 18]. However, within the Bromeliaceae family, the mitogenome remains largely uncharacterized, and its potential for breeding applications still needs to be explored. This gap in mitochondrial genomics limits the development of breeding resources, restricting the identification of molecular markers and traits relevant to crop improvement [19]. While chloroplast and nuclear genomes have been investigated, the lack of well-assembled mitogenomes has also hindered studies on organellar co-evolution and phylogenetic diversification in this family [11, 20].

In this study, we presented the first assembly and annotation of the complete mitogenomes of two high-value bromeliads: Ae. fasciata and An. comosus. To characterize their genomic structures, we performed comprehensive analyses of architectural features, sequence transfers, and repetitive elements. To elucidate the evolutionary dynamics of protein-coding genes (PCGs), we examined RNA editing sites, codon usage bias, nucleotide diversity, and selective pressure, and conducted comparative genomics with closely related species. Furthermore, using available mitochondrial genomic data, we reconstructed the phylogenetic relationships within Poales to clarify the phylogenetic position of Bromeliaceae within the order. This work contributes to the understanding of the architecture and function of bromeliad mitogenomes, provides support for established phylogenetic relationships within the family, and offers valuable genomic resources for future evolutionary research and breeding in Bromeliaceae, including potential molecular markers and insights into cytoplasmic male sterility.

Materials and methods

Data acquisition, genome assembly, and annotation

Fresh leaf tissues of Aechmea fasciata and Ananas comosus were collected from Guangzhou, Guangdong Province, China (23.13° N, 113.26° E), and the plant materials were formally taxonomically identified by Professor Chao Shi based on morphological characteristics. Voucher specimens (QUST20250528-1 for Aechmea fasciata and QUST20250602-1 for Ananas comosus, respectively) were deposited in the Evolutionary Ecology Laboratory of Qingdao University of Science and Technology, snap-frozen in liquid nitrogen immediately after collection, and stored at − 80 °C for subsequent experiments. Genomic DNA was extracted from approximately 100 mg of tissue using the cetyltrimethylammonium bromide (CTAB) method [21]. DNA purity and concentration were assessed using a NanoDrop spectrophotometer (Thermo Fisher Scientific, USA). Samples with an A260/A280 ratio between 1.8 and 2.0 were selected for downstream processing. Integrity was confirmed by 1% agarose gel electrophoresis, which showed clear, high-molecular-weight bands. For long-read sequencing, SMRTbell libraries were constructed with the SMRTbell Express Template Prep Kit v2.0 (Pacific Biosciences, USA). Fragment size distribution was verified using a Femto Pulse system (Agilent Technologies, USA) to ensure fragments > 10 kb, followed by sequencing on a PacBio Sequel IIe platform to generate high-fidelity (HiFi) circular consensus sequencing (CCS) reads. For short-read sequencing, paired-end libraries (~ 500 bp insert) were prepared using the Illumina TruSeq DNA Library Prep Kit (Illumina, USA) and sequenced on an Illumina HiSeq 6000 platform (Novogene, China). Finally, we obtained approximately 14.2 GB and 27.4 GB of HiFi reads for Ae. fasciata and An. comosus, respectively. In addition, 10 GB of Illumina data was specifically used for chloroplast genome assembly of Ae. fasciata. Sequencing data quality was assessed using SeqKit v2.8.2 [22]. For An. comosus, the PacBio HiFi reads achieved a read N50 of 10.6 kb, with 97.6% of bases at Q30. For Ae. fasciata, the reads achieved a read N50 of 13.6 kb, with 96.5% of bases at Q30.

Using the HiFi reads, we performed Flye v2.9.2 [23] to obtain the complete mitogenome assemblies with the default parameters. Mitogenome contigs were identified via BLAST search in Bandage v0.8.1 [24]. We used the mitogenome of Nymphaea colorata (NC_037468.1) as a reference. This species belongs to the order Nymphaeales, a basal angiosperm lineage that diverged prior to the separation of monocots and eudicots [25]. Its intermediate phylogenetic position and relatively slow mitochondrial evolutionary rate [26, 27] make it a suitable outgroup reference for annotating and analyzing monocot mitogenomes, as it retains ancestral genomic features while being sufficiently distant to avoid confusion with lineage-specific innovations. Subsequently, we manually curated the initial assembly graph in Bandage v0.8.1 by merging overlapping contigs and selecting the primary circular conformation based on read depth consistency, to obtain a purified mitogenome [24]. Contigs showing high similarity (> 95% identity, > 80% query coverage) to the chloroplast genomes were identified as contaminants using BLASTn v2.13.0 and removed via the “Remove” function in Bandage [28]. Sequencing depth of mitochondrial contigs was calculated using Bandage. Regions exhibiting sequencing depth approximately equal to integer multiples of the average coverage depth were retained as duplicated sequences, and their copy numbers were noted in the annotation. The cleaned assembly graph was exported in GFA and FASTA formats for downstream analysis.

To investigate potential intracellular DNA transfer events, particularly the identification of mitochondrial plastid DNA (MTPT), we assembled the chloroplast genome of Ae. fasciata using Get-Organelle v1.7.7.1 with Illumina reads and the parameters "-R 15 -k 21,45,65,85,105 -F embplant_pt" [29]. The reference chloroplast sequences of An. comosus was obtained from the NCBI database (accession number: NC_026220.1).

The mitogenomes of two bromeliad species were annotated using the IPMGA online tools [30]. The tRNA and rRNA of the mitogenome were annotated by using the software tRNAscan-SE v2.0.11 [31] and BLASTn v2.13.0 [28], respectively. The above results were manually checked and corrected for more accurate annotation. Then, the chloroplast genomes were annotated using GeSeq v2.03 [32]. The visualization of the mitogenome was executed using the OGDRAW v1.3.1 (https://chlorobox.mpimp-golm.mpg.de/OGDraw.html) online tools [33].

Collinearity analysis

To assess genome structural evolution across phylogenetic groups, we performed collinearity analyses between the mitogenomes of two bromeliads. Alignments were generated using MUMmer v4.0.0, with a minimum collinear block size of 500 bp to ensure stringency [34]. Visualization was conducted using the NGenomeSyn v1.41 toolkit [35].

Identification of mitochondrial chloroplast DNAs (MTPTs)

In angiosperms, chloroplast-derived sequences are frequently transferred into the mitogenome occurs via MTPT events [36]. To identify MTPTs, we performed BLASTn v2.13.0 comparisons between the mitochondrial and chloroplast genomes of the two bromeliad species. Only homologous regions with alignment length ≥ 30 bp, sequence identity > 95%, and E-value < 1e-5 were retained as reliable MTPTs [28].

Repetitive sequence analysis, codon usage bias, and RNA-editing sites prediction

For repetitive sequence analysis, we firstly identified simple sequence repeats (SSRs) and tandem repeat sequences using MISA (v2.1) [37] and Tandem Repeats Finder (v4.09) [38], respectively. Dispersed repeat sequences were identified using the REPuter web server with the following parameters: Hamming distance = 0 (i.e., perfect matches only), minimum repeat length = 30 bp, and significance threshold (E-value) = 1 × 10⁻⁵. Tandem repeats were detected using Tandem Repeats Finder v4.09 [38] with default parameters (match = 2, mismatch = 7, indel = 7; minimum alignment score = 50; maximum period size = 2000). All results were visualized using the R package ggplot2 v3.5.1 [39, 40].

The usage of synonymous codons varies substantially across species, a pattern largely attributed to preferences shaped by long-term evolutionary pressures, including natural selection, mutation, and genetic drift. To analyze relative synonymous codon usage (RSCU), we first extracted all protein-coding sequences from the mitogenomes of two bromeliad species using Geneious Prime 2025.0.2 (https://www.geneious.com). Only genes with a length > 66 bp were retained. All annotated PCGs were included in the analysis to ensure comprehensive coverage of codon usage patterns across the mitogenome. RSCU values were subsequently calculated with the CAI v1.0.3 module [41].

In the mitochondria of most flowering plants (angiosperms), RNA editing events are prevalent in exon sequences and some noncoding regions. Mediated via C-to-U base substitution, these editing events induce changes in some amino acids as well as in the sites of promoters and terminators, ultimately generating more functional genes than the corresponding DNA coding sequences [42, 43]. RNA editing sites in the mitochondrial transcripts of the two bromeliad species were predicted using PREPACT3 v3.12.0, with plant mitochondrial-encoded proteins as references [44].

Selective pressure and nucleotide diversity analysis

Three parameters, including synonymous mutation frequency (Ks), nonsynonymous mutation frequency (Ka), and the nonsynonymous-to-synonymous mutation rate ratio (Ka/Ks), are commonly employed to characterize the occurrence rates of synonymous and nonsynonymous mutations in evolutionary analysis [45]. Pi is a quantitative index for measuring the degree of genetic variation at the nucleotide level within a population or among different sequences, which can be used to compare the degree of genetic differentiation among different gene regions [46].

To investigate the evolutionary selection pressure and nucleotide diversity, we selected these two bromeliad species and their closely related species to assess genetic variation. The shared PCGs were aligned individually using MAFFT v7.490 [47]. For each PCG, we calculated the Ka/Ks ratio and nucleotide diversity (Pi) on a gene-by-gene basis using DnaSP v6.12.03 [48], rather than concatenating all genes into a single alignment. This approach allows for the detection of gene-specific evolutionary patterns and avoids potential biases introduced by combining genes with different evolutionary rates or functional constraints.

Phylogenetic analysis

To explore the evolutionary relationships of bromeliad taxa in a broader context, we obtained all of the Poales plant available complete mitogenomes from the NCBI database (17 species), along with our two newly assembled mitogenomes for analysis (Table S1). Acacia confusa and Delonix regia were selected as the outgroups of the phylogenetic tree. Next, ten mitochondrial PCGs that were universally present and complete across all 21 species were extracted (cox2, rps7, cob, nad3, rps12, atp8, nad5, rps3, nad1, nad2). Although additional mitochondrial PCGs are present in some taxa, they were excluded due to absence, incomplete sequences, or evidence of transfer to the nuclear genome in other species, which could introduce bias in the phylogenetic reconstruction. These genes were aligned with MAFFT v7.490 [49]. We used Maximum likelihood (ML) and Bayesian phylogenetic inference (BI) to construct phylogenetic trees. The construction of phylogenetic tree based on ML method was conducted using IQ-TREE v2.2.2.7 [50], with the TVM + G4 model selected based on the Bayesian Information Criterion (BIC). Bootstrap support values for each branch in the phylogenetic tree were calculated based on 1,000 replicates. The BI phylogenetic tree was constructed using MrBayes v3.2.7 [51], with the TVM + G substitution model selected via jModelTest v2.1.10 under the BIC [52]. Posterior probabilities for branch support were estimated from two independent Markov chains run for 2 million generations, sampled every 1,000 iterations, and discarding the initial 25% of trees as burn-in. Consensus trees were visualized in iTOL v6 (https://itol.embl.de/). Acacia confusa and Delonix regia (Fabaceae) were selected as outgroups. While phylogenetically distant from Poales, Fabaceae exhibit relatively stable mitochondrial substitution rates [53, 54] and provide an intermediate evolutionary distance suitable for rooting monocot trees without introducing long-branch attraction artifacts. The use of two outgroup species also enables cross-validation of the inferred root position [55].

Results

Characteristics for the mitogenome of Aechmea fasciata and Ananas comosus

Based on HiFi sequencing data, we successfully assembled complete mitogenomes for two bromeliad species Ae. fasciata and An. comosus, respectively (Fig. 1). The initial assembly graphs of the bromeliad mitogenomes revealed multi-branched conformations, composing of ten contigs in Ae. fasciata and three in An. comosus (Fig. 1A, B). After decatenation, each mitogenome resolved into a single contiguous sequence (Fig. 1C, D), displaying a predominant circular master conformation (Fig. 1E, F). Notably, given the dynamic nature of plant mitogenomes, alternative isomeric forms may also coexist in vivo due to recombination between repetitive sequences [56]. In addition, beneath this conserved structural scaffold lies substantial size variation. The Ae. fasciata mitogenome spans 1.16 Mb, while An. comosus is considerably more compact at 0.93 Mb. The GC content of the two assemblies ranged from 44.3% to 44.4%, consistent with the typical AT-rich nature of plant mitogenomes [57].

Fig. 1.

Assembly and comparative analysis of mitogenomes in Aechmea fasciata and Ananas comosus. A, B Schematic of mitogenome contigs (colored segments). C, D Circular master conformation. E, F Assembly graphs with genes color-coded by function. G Collinearity analysis plot. Blue arcs indicate homologous regions

Despite this size disparity, gene content remains remarkably stable (Table 1). The An. comosus mitogenome contains 86 genes, comprising 44 PCGs, 39 transfer RNA (tRNA) genes, and three ribosomal RNA (rRNA) genes. The Ae. fasciata mitogenome harbors 84 genes, including 42 PCGs, 39 tRNA genes, and three rRNA genes. Both species share a conserved core of 31 unique PCGs, encompassing all essential functional categories: ATP synthase subunits (atp1, atp4, atp6, atp8, atp9), cytochrome c biogenesis factors (ccmB, ccmC, ccmFC, ccmFN), cytochrome c oxidase subunits (cox1, cox2, cox3), NADH dehydrogenase subunits, and ribosomal proteins. The observed size difference thus cannot be attributed to wholesale gene loss but rather to variation in intergenic regions, repetitive sequence content, and transferred DNA from other genomic compartments.

Table 1.

The coding genes of mitogenome for Ae. fasciata and An. comosus

Function	Aechmea fasciata	Ananas comosus
Group of genes	Gene name	Gene name
ATP synthase	atp1 atp4 atp6 atp8 atp9(2)	atp1 atp4 atp6 atp8 atp9(2)
Cytochrome c biogenesis	ccmB ccmC(2) ccmFC ccmFN	ccmB ccmC(2) ccmFC ccmFN
Ubiquinol cytochrome c reductase	cob	cob
Cytochrome c oxidase	cox1 cox2 cox3	cox1 cox2 cox3
Maturases	matR	matR
Transport membrane protein	mttB	mttB
NADH dehydrogenase	nad1 nad2 nad3(2) nad4 nad4L nad5 nad6(2) nad7 nad9	nad1 nad2 nad3(2) nad4 nad4L nad5 nad6 nad7(2) nad9
Ribosomal proteins (LSU)	rpl16 rpl2 rpl5	rpl16 rpl2 rpl5
Ribosomal proteins (SSU)	rps10 rps11 rps12 rps13 rps14 rps19 rps2 rps3 rps4 rps7	rps1 rps10 rps11 rps12 rps13 rps14 rps19(2) rps2 rps3 rps4 rps7
Succinate dehydrogenase	sdh4	sdh4
Ribosomal RNAs	rrn18 rrn26 rrn5	rrn18 rrn26 rrn5
Transfer RNAs	trnC-GCA(3) trnD-GUC(2) trnE-UUC(2) trnF-GAA(3) trnfM-CAU(2) trnH-GUG(2) trnI-AAU trnI-CAU(2) trnK-UUU(2) trnL-CAA(2) trnL-UAA trnM-CAU(3) trnN-GUU(2) trnP-GGG trnP-UGG trnQ-UUG(2) trnR-UCU trnS-CGA trnS-GCU trnS-GGA trnS-UGA trnV-GAC trnW-CCA trnY-GUA	trnC-GCA trnD-GUC(2) trnE-UUC(4) trnF-GAA(3) trnfM-CAU(3) trnG-GCC trnH-GUG(2) trnI-AAU trnI-CAU(2) trnK-UUU trnL-CAA(2) trnM-CAU(2) trnN-GUU(3) trnP-GGG trnP-UGG trnQ-UUG(2) trnR-UCU trnS-GCU trnS-GGA trnS-UGA trnV-GAC trnW-CCA trnY-GUA(2)

However, interspecific variations were observed in the copy numbers of several genes: for instance, the atp9 gene was duplicated in both species. While in the NADH dehydrogenase gene family, the nad6 gene possessed two copies in Ae. fasciata but only one in An. comosus. Conversely, the nad7 gene was present in two copies in An. comosus, but as a single copy in Ae. fasciata. Most notably, the ribosomal protein gene rps1 is present in An. comosus, but absent from Ae. fasciata. The loss is potentially compensated by functional transfer to the nuclear genome, a process commonly observed in angiosperm evolution [58, 59] (Table S2). However, the functional significance of nad6, nad7, and atp9 duplications remains unclear. Such duplications may result from repeat-mediated rearrangements [60] or represent transitional states in gene loss/transfer to the nucleus [59]. Further studies are needed to assess their transcriptional activity and potential roles in functional redundancy or subfunctionalization. In addition, we compared another bromeliad species (Puya raimondii) and found that the number of PCGs exhibits a significant loss of ribosomal genes in Pu. raimondii (Table S2).

Collinearity analysis reveals extensive sequence homology between the two Bromelioideae mitogenomes, yet the connecting ribbons display a characteristically disordered interlacing pattern rather than simple linear correspondence (Fig. 1G). This signature indicates that substantial genomic rearrangements have occurred since the divergence of these lineages, including inversions, translocations, and fragment shuffling. Such rearrangements are likely mediated by recombination between repetitive sequences, a hallmark of plant mitogenome evolution. Despite this structural reshuffling, the preservation of gene content and the circular master structure suggests strong functional constraints maintaining essential mitochondrial operations while permitting structural plasticity [15].

Analysis of homologous fragments between mitochondria and chloroplasts

During the evolution of organelles, gene transfer exhibits a distinct unidirectional pattern, with the primary flow occurring from the plastid genome to the mitogenome [58, 61]. Our analyses uncovered extensive but qualitatively different MTPTs landscapes in the two bromeliads, providing insights into the mechanisms and timing of organellar DNA exchange (Fig. 2). In Ae. fasciata, the total MTPT length reached 96,069 bp, accounting for 4.29% of the mitogenome and 60.3% of the plastome, with a total of 83 homologous fragments. Compared with Ae. fasciata, An. comosus had shorter MTPT sequences (75,215 bp), a smaller proportion of the plastome (47.1%), but a larger proportion of the mitogenome (6.06%), while the number of homologous fragments was identical.

Fig. 2.

Distribution of MTPTs fragments in Ae. fasciata (A) and An. comosus (B). a: The green arc represents the chloroplast genome (cpDNA), and the brown arc represents the mitogenome (mtDNA); b–d: gene density; e: GC content; f: connecting lines indicate homologous fragments, representing chloroplast-to-mitochondria DNA transfers

These MTPTs varied greatly in individual length, ranging from 31 bp to 6,656 bp in An. comosus and from 33 bp to 7,069 bp in Ae. fasciata, with the majority of fragments shorter than 500 bp (63.9% in An. comosus and 53.1% in Ae. fasciata), while a small proportion (2.4% in both species) represented large fragments exceeding 5,000 bp. Notably, Ae. fasciata exhibited a higher average fragment length (1,034.4 bp) compared to An. comosus (852.5 bp), suggesting more extensive chloroplast DNA migration events in its evolutionary history. In An. comosus, the majority of these MTPTs originated from the IR region of the chloroplast genome, accounting for 49.8% of the total MTPT length. In contrast, MTPTs originating from the chloroplast IR region accounted for 35.4% in Ae. fasciata. Finally, all MTPTs collectively contained 29 PCGs and two rRNA genes, while no tRNA genes were detected (Table S3).

Repetitive sequence analysis

To investigate the structural characteristics and potential sources of genomic plasticity, we analyzed the repetitive sequences, including simple sequence repeats (SSRs), dispersed repeats and tandem repeats in the mitogenomes of Ae. fasciata and An. comosus. SSRs are widely distributed in the mitogenome, but the proportions of the six types of nucleotide repeats show differences, which reflects structural variation between their mitogenomes (Fig. 3, Table S4).

Fig. 3.

Distribution of repetitive sequences in the mitogenomes of Ae. fasciata (A) and An. comosus (B). Circles: a represents mitochondrial gene backbone; b–d represents tandem repeat sequences, followed by SSRs, and dispersed repeat sequences; d: The innermost lines mean the connection levels. The detailed repetitive sequence statistics are shown in Table S4

Mononucleotide repeats were the most abundant category in Ae. fasciata and An. comosus, accounting for approximately 41.2% and 45.9% of its total SSRs. Notably, there are several hexanucleotide repeats in both species: only one locus was detected in Ae. fasciata, while An. comosus contained four loci of this type, revealing the structural complexity.

Secondly, both species possessed all four typical dispersed repeat types (forward, reverse, complementary, and palindromic) (Table S4). Ae. fasciata had more classified dispersed repeats (292 pairs in total) than An. comosus (150 pairs). Within these dispersed repeats, forward repeats predominated in An. comosus (33.33%) and palindromic repeats in Ae. fasciata (32.9%). In terms of length distribution, forward repeats showed the widest span in both species, ranging from 50 to 9,192 bp in Ae. fasciata, and 50 to 11,143 bp in An. comosus, with no obvious concentration in a specific short-length interval (Table S4).

On the other hand, in Ae. fasciata, a total of 25 tandem repeats, ranging in length from 25 to 76 bp, were found in the genome. In contrast, although fewer repeats (15) were identified in An. comosus, they exhibited a much wider length range (25 to 175 bp, Table S4). The match degrees of these two species were greater than 90%.

Codon usage bias analysis

Codon usage bias, as the non-random use of synonymous codons, reflects the cumulative effects of mutation, selection, and genetic drift over evolutionary time [62]. RSCU provides insights into the translational efficiency and mutational pressures shaping mitochondrial PCGs. Our analysis reveals striking conservation between the two bromeliad mitogenomes, suggesting shared evolutionary constraints despite their structural divergence.

We identified 25 codons with significant usage bias (RSCU > 1.1) in Ae. fasciata and 26 codons in An. comosus, which contained GTA of Valine (RSCU = 1.12) as the extra codon (Fig. 4). The distribution of preferred codons was highly consistent between the two species, with only minor differences in the RSCU values of codons corresponding to a few amino acids. Notably, both species shared the same set of highly preferred codons, namely GCU (Alanine), CAA (Glutamine) and CAU (Histidine), with RSCU values greater than 1.5. Importantly, all these preferred codons end in either A or U in both species (Fig. 4). This finding suggests a preference for A-ended or U-ended codons in their mitogenomes.

Fig. 4.

Analysis of codon preference in the mitogenome of Ae. fasciata (A) and An. comosus (B). Codons are color-coded by amino acid. RSCU > 1 preferential codon usage above its synonymous counterparts; < 1 implies a lesser frequency of usage

RNA-editing sites prediction analysis

RNA editing sites were predicted in the mitochondrial PCGs of these two bromeliad species. For Ae. fasciata, a total of 548 RNA editing sites were identified in these PCGs; in An. comosus, 542 RNA editing sites were detected (Fig. 5). The overall distribution of these sites was highly similar between the two species, indicating a deeply conserved RNA editing mechanism (Fig. 5A). However, the number of editing sites varied significantly across genes with different functions (Fig. 5B). NADH dehydrogenase and cytochrome c biogenesis genes were editing hotspots, for example, the nad4 gene was the most heavily edited in both species (55 sites). While ribosomal protein genes were minimally edited, like rps14 had only one editing site in each species (Fig. 5B). In addition, the number of RNA editing sites in the two species showed a right-skewed distribution pattern, meaning that genes with fewer editing sites account for a larger proportion (Fig. 5C).

Fig. 5.

RNA editing site analysis in bromeliad mitogenomes. A Editing sites across mitochondrial genes by functional category. B Distribution by category and species (Ae. fasciata: blue; An. comosus: purple). C Site distribution by species. D Between-species differences in editing sites across PCGs. E Proportion of editing positions

Despite the overall conservation, notable species-specific differences were observed in several genes. For example, nad5 had nine more editing sites in Ae. fasciata than in An. comosus, and rps1 was unedited in Ae. fasciata but contained seven sites in An. comosus. These marked differences suggest potential lineage-specific adaptations in energy metabolism and translational machinery (Fig. 5D) [63]. Moreover, a strong codon position bias was found, with ~ 75% of edits occurring at the second position, which was consistent with other research. (Fig. 5E) [19]. It is worth noting that these predictions are based solely on computational analyses and require experimental confirmation through transcriptome sequencing.

Analysis of selection pressure and nucleotide diversity

To assess the evolutionary forces shaping bromeliad mitochondrial genes, we calculated the ratio of nonsynonymous (Ka) to synonymous (Ks) substitution rates across 31 PCGs shared between species and their relatives (Fig. 6A). The Ka/Ks ratio provides a powerful indicator of selection pressure: values below 1.0 indicate purifying (negative) selection acting to eliminate deleterious mutations, values equal to 1.0 suggest neutral evolution, and values exceeding 1.0 indicate positive (Darwinian) selection favoring adaptive amino acid changes [64]. The results showed that most of the Ka/Ks ratios for shared PCGs were less than one, indicating these genes were predominantly under negative (purifying) selection. Notably, the cox1 gene exhibited the lowest average Ka/Ks ratio (≤ 0.0987 across all species pairs). In contrast, genes rps13 and sdh4 exhibited a Ka/Ks value greater than one, suggesting potential positive selection that may facilitate adaptive differentiation between species.

Fig. 6.

Selection pressure and nucleotide diversity of mitochondrial PCGs in bromeliads. A Ka/Ks ratios by gene family (red line = 1). B Nucleotide diversity (Pi) across gene families. C Distribution of Pi values. D Pi values by gene family (red: mean; blue: median)

Besides, to evaluate the genetic diversity and evolutionary dynamics of mitochondrial genes among the studied taxa, we analyzed the nucleotide diversity (Pi) of 31 mitochondrial genes, encompassing PCGs and ribosomal protein genes.

According to functional classification (Fig. 6B), genes belonging to different functional groups displayed distinct levels of variability. Among them, ATP synthase genes exhibited the highest overall variability. The atp8 gene showed particularly elevated diversity, placing it above the median for all genes analyzed. Ribosomal protein genes also displayed relatively high Pi values, consistent with previous observations of relaxed constraints on translation-associated genes in plant mitochondria [59]. In contrast, the NADH dehydrogenase gene family was relatively conserved, with Pi values ranging from 0.02678 to 0.07675 [19]. Density distribution analysis (Fig. 6C) revealed that Pi values were primarily concentrated in the range of 0.04 to 0.08, indicating that most genes exhibited moderate genetic variation. Notably, some genes (e.g., atp8 and cox2) exhibited higher Pi values (above the median), demonstrating high variability and thus serving as valuable resources for molecular marker development (Fig. 6D).

Phylogenetic analysis

Based on conserved mitochondrial PCGs, we investigated the phylogenetic relationships between bromeliads and their closely related species. We constructed both ML and BI trees based on ten shared mitochondrial PCGs of 19 Poales species, which included three Bromeliaceae species, and 16 closely related species (Poaceae, Cyperaceae, and Juncaceae) (Fig. 7, Table S1). Two species from Fabales and Rosales, respectively, served as outgroups. The results displayed robust nodal support, with bootstrap values ranging from 63 to 100 (Fig. 7). These two trees exhibited a highly consistent topological structure, which indicates that the phylogenetic relationships among the branches are relatively robust. Our phylogenetic analysis assigned all species into four monophyletic clades, corresponding to the families Cyperaceae, Poaceae, Bromeliaceae, and Juncaceae. This topology was fully consistent with the established relationships in the APG IV system [25]. Within the Bromeliaceae, Ae. fasciata and An. comosus clustered closely together, while Puya raimondii was phylogenetically distant from them.

Fig. 7.

Phylogenetic relationships constructed based on the mitogenomes of 19 Poales species. The topological structures of the ML tree and BI tree were basically consistent. The bootstrap values were listed at each node. Nodes with bootstrap values < 100 (ML) and posterior probabilities < 1 (BI) were marked in red. Colorful lines indicate the groups to which the specific species belong

Discussion

General features of the simple mitogenome structures

Mitochondria, recognized as the powerhouse of plants, show their dynamic evolution of genomic structure and now serve as a powerful tool for uncovering the diversity among different species [65, 66]. In this study, the assembled mitogenomes of the bromeliad species Aechmea fasciata and Ananas comosus display the intermediate-to-large size among reported plant mitogenomes (Fig. 1) [67, 68]. Consistent with previous studies on plant mitogenomes [69, 70], our current assembly and annotation results for Ae. fasciata and An. comosus appear to conform to the typical model of a single circular structure without isomeric forms (Fig. 1). This structural feature aligns with the reported mitogenome structure of Puya raimondii (GeneBank accession number: PQ221457.1) within the same family. Our results preliminarily support the conserved mitogenomic architecture within Bromeliaceae.

Generally, the differences in mitogenome sizes result from variations at the level of protein-coding gene loss/duplication, horizontal gene transfer, and expansion of repeat sequences [59, 60, 71, 72]. First, our comparative analysis of three Bromeliaceae mitogenomes reveals notable PCGs differences. The two Bromelioideae species retain a nearly complete and highly conserved set of PCGs, including almost all ribosomal protein genes, reflecting an ancestral and conserved mitogenome architecture. In contrast, Puya raimondii (Puyoideae) has undergone extensive PCGs loss (Table S2). This implies that the Puyoideae lineage may have experienced unique and substantial functional replacement or gene transfer in its mitogenome [14, 59].

Second, as a key component of horizontal DNA transfer, MTPTs influence the size of plant mitogenomes [58, 61, 72]. Our study reveals extensive gene transfer between mitochondrial and chloroplast genomes (Fig. 2). The MTPTs transfer hotspots of both bromeliad species exhibit variation (Table S3). In An. comosus, the enrichment of MTPTs hotspots in the IR region indicates that its mitogenome retains ancient and intact plastid DNA insertions [36]. In contrast, the MTPT pattern in Ae. fasciata, dominated by LSC-region insertions and lacking intact IR hotspots, implying more recent and dynamic gene transfer and genomic restructuring [73].

Repetitive sequences also contribute significantly to genomic size variation [60]. Differences in the proportion of dispersed repeats reflect ongoing genome evolution (Table S4), and abundant simple sequence repeats (SSRs) suggest active recombination (Table S4). In summary, our comparative analyses of PCGs, MTPTs, and repeat sequences revealed the structural characteristics and differences between the mitogenomes of these two economically important bromeliad species.

Patterns of molecular evolution and selection pressure

Consistent with findings in other monocots, such as Houttuynia cordata [16] and Cymbidium ensifolium [19], the total number of predicted RNA editing sites exceeded 500 among these two bromeliads (Fig. 5). Meanwhile, the distribution of RNA editing sites reveals minor differences between these related species. For instance, genes involved in cytochrome c biogenesis and NADH dehydrogenase exhibited the highest number of editing sites, whereas ribosomal protein genes showed minimal editing, highlighting significant variation in RNA editing frequency across different gene families (Fig. 5). The high editing load in cytochrome c biogenesis and NADH dehydrogenase genes primarily serves to restore evolutionarily conserved amino acid sequences, ensuring proper assembly and activity of respiratory complexes [63, 74, 75]. In contrast, the minimal editing in ribosomal protein genes may be linked to their ongoing transfer to the nuclear genome, which relaxes selective pressure on their mitochondrial copies [59]. Furthermore, the strong bias of first and second codon-position editing serves as direct proof that RNA editing functions to restore and maintain the conservation of protein sequences (Fig. 5E) [75]. Overall, these RNA editing patterns reveal the extensive post-transcriptional modification landscape in bromeliad mitochondria, providing a foundation for future functional studies of RNA editing in this family.

Analysis of codon usage bias is important for revealing phylogenetic relationships among organisms and patterns of genetic differentiation [76]. Our study also revealed a strong preference for A or U at the third codon position of mitochondrial mRNAs, reflecting the genomic AT-richness and characteristic mutational bias of the DNA genomes [77]. The distribution pattern of highly preferred codons was similar in the two species, except for a few codons (Fig. 4), further suggesting conservation of molecular evolution between these two species.

The Ka/Ks ratio can be used to determine whether specific PCGs have been subject to selective pressure during evolution [78]. For both Ae. fasciata and An. comosus, no matter compared within the Bromeliaceae family or with species from closely related families, the Ka/Ks values of most PCGs were less than one (Fig. 6A), reflecting the action of purifying selection to eliminate deleterious mutations and maintain normal mitochondrial function [19]. There were no statistically significant differences observed among the majority of functional gene families; in contrast, the Ka/Ks ratios of rps13 and sdh4 exceeded 1, likely reflecting positive selection during functional adaptation in mitochondrial translation and energy metabolism, as similarly reported in grasses [79]. Furthermore, the analysis of nucleotide diversity revealed that the NADH dehydrogenase genes were relatively conserved (Fig. 6D), indicating their functional stability in energy metabolism.

Cytoplasmic male sterility (CMS) is a mitochondrially encoded trait that disrupts pollen development and serves as a pivotal tool for the efficient production of hybrid seeds in agriculture. Notably, many CMS-associated loci in crops are chimeric genes originating from mitochondrial rearrangements, with fragments of ATP synthase subunits being frequently involved, like T-urf13 in maize CMS-T (containing atp6 sequences), orf79 in rice HL-CMS (co-transcribed with atp6), and orf224 in rapeseed pol CMS (associated with atp6) [14, 80–83]. In this study, the ATP synthase gene family (e.g., atp8) in bromeliads exhibited characteristics such as high nucleotide diversity (Pi) and a lower frequency of RNA editing, indicating relatively relaxed functional constraints and elevated sequence plasticity (Figs. 5B, 6B). This raises the possibility that such genes could, in principle, serve as a source for CMS-related variants, although direct evidence is currently lacking. Consequently, our assembly and characterization of these mitogenomes provide a foundational resource for future screening and utilization of CMS elements in the molecular breeding of economically important bromeliads, such as pineapple and ornamental varieties.

Phylogenetic inference of relationships at the genus level

With the advancement of mitogenome sequencing, it has become widely recognized that mitogenomes vary across species, which holds great potential as a tool for phylogenetic analysis [84, 85]. Our Poales phylogenetic tree (based on mitogenomes) shows that the clustering relationships among families align with previous research ((Bromeliaceae, (Poaceae, (Cyperaceae, Juncaceae)), supporting that Bromeliaceae is located at the base of the Poales phylogenetic tree and is distantly related to the core Poales (including Poaceae, Cyperaceae, etc., Fig. 7) [25]. This systematic position is consistent with the unique morphological and ecological traits exhibited by Bromeliaceae [1, 86].

Meanwhile, our results also show that within Bromeliaceae, the genera Ananas and Aechmea are more closely related to each other, which aligns with the findings of relevant studies based on chloroplast genomes [1, 86]. Based on the previous analyses, the loss of mitochondrial PCGs in Puyoideae concentrates in ribosomal genes, which may reflect lineage-specific evolutionary dynamics [59]. It also suggests that the mitogenome of Pu raimondii may have undergone significant evolutionary events such as functional reduction or gene transfer [59, 87]. On the other hand, despite the considerable morphological divergence between the two bromeliads, the short divergence time and the recovered phylogenetic relationships reflect the relatively conservative nature of their mitogenomes during speciation [86] (Table S2).

However, several limitations constrain the current study. First, taxonomic sampling within Bromeliaceae is limited to only three genera (Aechmea, Ananas, and Puya); the absence of mitogenome data from other subfamilies (e.g., Tillandsioideae) precludes a comprehensive assessment of mitogenome evolution across the family and limits resolution of deeper relationships within Bromeliaceae. Second, within the order Poales, complete mitogenomes are still lacking for key families (e.g., Restionaceae, Eriocaulaceae), hindering a full understanding of mitochondrial diversification at the order level. Third, the lack of transcriptomic data prevents experimental validation of predicted RNA editing sites and functional assessment of duplicated genes. Furthermore, the MTPTs, repetitive elements, and RNA editing sites reported here were identified based on computational predictions and require experimental validation via RNA-seq or bisulfite sequencing to confirm their biological activity and functional relevance.

In the future, more comprehensive plant mitogenome data will help to construct a more complete and higher-resolution phylogenetic tree, thereby elucidating the evolutionary relationships within Bromeliaceae from a mitogenomic perspective.

Conclusion

This study presents the first complete mitogenomes of two economically important Bromeliaceae species, Aechmea fasciata and Ananas comosus, revealing both conserved features and lineage-specific variations that shape mitogenome evolution in this family. The substantial size difference between the two mitogenomes primarily reflects dynamic changes in repetitive sequence content, chloroplast-derived DNA insertions, and gene copy number variations, highlighting these elements as key drivers of genome expansion and structural divergence. Conserved patterns of codon usage bias and RNA editing suggest strong selective constraints maintaining fundamental mitochondrial functions, whereas the high nucleotide diversity observed in ATP synthase genes highlights them as a focus for future investigations into potential CMS-related mechanisms in bromeliads. Phylogenetically, our analyses confirm the basal position of Bromeliaceae within Poales and clarify the close relationship between Ae. fasciata and An. comosus. Collectively, these findings establish a foundational genomic framework for understanding mitogenome evolution in bromeliads and offer valuable resources for evolutionary studies and marker-assisted breeding in this horticulturally and agriculturally important family.

Authors’ contributions

B.Z. and B.W. conceived and designed the project. B.W. assembled the mitogenomes and performed multiple analyses. P.Y. and B.Z. drafted the manuscript. B.Z., H.W., S.W., C.S., G.L., H.S., K.F., and J.Y. participated in the revision of the manuscript. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by the National Natural Science Foundation of China (NO. 32400371, 32570279, and 32500360), Shandong Provincial Natural Science Foundation (ZR2024QC396).

Data availability

The complete mitogenome sequences of *Aechmea fasciata* and *Ananas comosus* generated in this study have been deposited in the NCBI GenBank database under accession numbers [*Aechmea fasciata*: NC_087835.1] and [*Ananas comosus*: PX645088.1], respectively.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.