Multipartite mitochondrial genome evolution in Halophila seagrasses: repeat-driven structural plasticity and selection signatures of marine adaptation

Abstract

Background

The degradation of seagrass beds, which serve as pivotal indicators of coastal ecosystem health, has emerged as a critical reflection of global nearshore environmental stress. In this study, we employed two closely related seagrass species, Halophila beccarii and H. ovalis, as case studies to investigate the structural and evolutionary dynamics of their mitochondrial genomes through comparative genomics and evolutionary analyses.

Results

Our findings reveal striking structural disparities in the mitochondrial genomes of the two species: H. beccarii comprises 28 free circular elements totaling 1.98 Mb, whereas H. ovalis harbors 12 circular units spanning 0.58 Mb. Codon usage analysis demonstrated a significant A/U preference at the third position of high-frequency codons. Analyses of relative substitution rates (Ka/Ks) showed that the nad3 gene consistently exhibited ratios greater than 1 in both marine-freshwater pairwise comparisons and across the phylogeny, distinguishing it from other mitochondrial genes. Halophila species showed a reduced complement of RNA editing sites. Specifically, genes nad3 and nad6 in H. beccarii, and cox1, nad3, and nad6 in H. ovalis completely lacked RNA editing under the experimental conditions employed.

Conclusions

Collectively, this work characterizes the diverse mitogenomic architecture in seagrasses, offering a genomic foundation for future studies on seagrass evolution and environmental response.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12870-026-08236-z.

Background

Seagrass beds serve as critical indicator systems for coastal ecosystem health, with their degradation rate closely reflecting the cumulative effects of nearshore environmental stresses [1, 2]. Although covering only 0.1% of the global ocean area, seagrass beds contribute approximately 20% of marine carbon sinks [3, 4] and provide nursery grounds for thousands of species [5–7]. However, approximately 30% of seagrass beds have vanished over the past century [8], highlighting the urgency to clarify their adaptive mechanisms in response to marine environmental pressures.

Adaptive evolution of plant organelle genomes has been confirmed to couple with environmental stress responses [9–11]. For instance, the mechanism by which the chloroplast FtsH5/VAR1 protease maintains singlet oxygen and salicylic acid homeostasis involves regulating chloroplast biogenesis at low temperatures, thereby influencing plant cold tolerance [12]; meanwhile, mitochondrial gene variations correlate with stress tolerance in energy metabolism [13]. These findings imply that genomic features in organelles may serve as potential molecular indicators for evaluating ecosystem resilience. Nevertheless, in marine angiosperms, understanding of mitochondrial genomes remains limited—only a few of the world’s 74 seagrass species have undergone mitochondrial genome annotation [14], and systematic research on mitochondrial genome–environment adaptation associations is lacking.

The ecological differentiation of Halophila beccarii and H. ovalis, spanning diverse aquatic habitats and adaptive traits, provides an ideal model for investigating the mechanisms of species divergence and adaptive evolution in seagrasses [15–17]. However, whether their organelle genomes drive adaptive evolution through structural variations, selection pressure differentiation, or regulatory innovations remains unclear. Based on this, in this study, weaimed to: (1) resolve the dynamic characteristics of mitochondrial genomes of the two Halophila species through comparative genomics and evolutionary analysis; (2) explore the evolutionary patterns of gene loss, repetitive sequence accumulation, and collinear structure; (3) evaluate the regulatory effects of natural selection and mutation pressure on codon usage bias and key functional genes; and (4) clarify the potential role of RNA editing events in marine environment adaptation. This study thus seeks to dissect the adaptive evolutionary mechanisms of mitochondrial genomes in H. beccarii and H. ovalis, illuminating how these seagrasses have navigated unique marine challenges through genomic adaptations.

Materials and methods

DNA extraction, purification, sequencing, and bioinformatics analyses

Halophila beccarii and H. ovalis were collected in October 11, 2024 from Jinpai Port, Lingao, Hainan, China, during low tide (accessed by entering the exposed mudflats post-tide recession). The plant material was formally identified by Prof. Shiquan Chen. All determinations were further verified by Dr. Zefu Cai. Voucher specimens (Collection numbers: 20240315 and 20240316) are deposited in the specimen repository of Qukou Scientific Research Base, Hainan Academy of Ocean and Fisheries Sciences, Haikou, China (Photo of the specimen: Fig. S1-S2). A total of 10 individuals were sampled for each species. One individual per species was subjected to sequencing. Genomic DNA was extracted from leaves using a modified CTAB method [18]. The quality and purity were evaluated using a Nanodrop 2000 spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA) and 1.0% a garose gel electrophoresis. Subsequently, we constructed 15 kb sequencing libraries using a SMRTbell Express Template Prep Kit 2.0 (PacBio Biosciences, CA, USA). Eventually, the HiFi sequencing data were generated on the PacBio Revio platform. Meanwhile, the libraries with an average fragment length of 350 bp were constructed using a NexteraXT DNA Library Preparation Kit (Illumina, San Diego, CA, USA). Sequencing was then performed on the DNBSEQ-T7 platform. Raw sequence reads underwent quality control processing using Fastp v 0.19.7 [19]. We then used minimap2 v 2.15-r905 [20] to align the HiFi sequencing data with the reference sequence (NCBI Reference Sequence: H. beccarii PP936079.1–PP936106.1; Zantedeschia aethiopica NC_073008.1; and Ruppia sinensis NC_088727.1). Next, we used miniasm v 0.3-r179 [21] to assemble the aligned sequences to obtain the initial assembly results. We then used nextPolish (v 1.3.1, https://github.com/Nextomics/NextPolish) to correct the initial assembly results and then used bowtie2 v 2.4.1 [22] to align the DNBSEQ-T7 sequences with the corrected results. We used Unicycler v 0.4.8 [23] to assemble the DNBSEQ-T7 data after alignment to obtain contigs, and then used Bandage v 0.8.1 [24] for visualization. Finally, we used minimap2 v 2.15-r905 [20] to align the HiFi data with the contig sequences, combine the connection situation from Bandage v 0.8.1 [24], confirm the contig connection relationship, and manually organize and present the final results.

We obtained the complete mitochondrial genomes of H. ovalis and H. beccarii, which were annotated using MITOFY [25]and MFANNOT [26] and mapped/visualized using OGDRAW [27]. The quality and completeness of the de novo assembled mitochondrial genomes were rigorously assessed. The assembly for H. beccarii achieved a contig N50 of 73,764, while the assembly for H. ovalis achieved a contig N50 of 49,953. To evaluate the sequencing depth, clean reads were mapped back to the final assemblies using BWA v 0.7.17-r1188 [28], calculating depth with SAMtools v 1.9 [29], and visualizing the distribution using Python 3.12.11. The resulting genome-wide coverage profiles are presented in Supplementary Fig. S3-S6. The raw sequencing data generated in this study have been deposited in the NCBI Sequence Read Archive (SRA) under the accession numbers PRJNA1303958 and PRJNA1303959. The assembled complete mitochondrial genome sequences were then submitted to the GenBank and are openly available under the accession number: PV999242- PV999253, PV999277- PV999304.

Chloroplast genome assembly and annotation were primarily performed using a method based on the ptGAUL pipeline [30]. First, HiFi PacBio reads were mapped to the reference chloroplast genome (H. beccarii NC_051970) using minimap2 v 2.15-r905 [20] to generate a PAF file. Mapped reads were then extracted from the PAF file. Next, the coverage depth distribution of these reads across the reference genome was analyzed, and regions with coverage depths below 50× were filtered out. The filtered reads were then assembled de novo using Flye v 2.9.6 [31] to generate an assembly graph. Using Bandage v 0.8.1 [24], the assembly graph was visualized to identify and remove redundant contigs or branches, followed by manual inspection and editing to form a single circular contig. Finally, the complete chloroplast genome was annotated using PGA [32], and the automated annotation results were manually curated to ensure accuracy. The chloroplast genome sequence have been deposited in the GenBank under the accession number: PV993969, PV993971.

Genomic characteristics and repetitive sequence dynamics analysis

Comparative analyses were performed using a dataset of 16 monocot mitochondrial genomes. This included the novel mitochondrial genomes of H. beccarii (accessions PV999277-PV999304) and H. ovalis (accessions PV999242-PV999253) assembled in this study, integrated with 14 publicly available mitochondrial genomes retrieved from GenBank. The mitochondrial genomes from GenBank included the previously published H. beccarii sequence (accessions: PP936079-PP936106), which was jointly used with the newly assembled H. beccarii genome from this study for interspecific comparative analysis. The specific composition of datasets used for each analysis is detailed in Table S1. The mitochondrial genomic information was statistically analyzed using Perl v 5.32.1. Based on the gene names in various species, we drew a core gene distribution map using R v3.6.0.

For codon bias analysis, the protein-coding genes (PCGs) of H. ovalis and H. beccarii were extracted using Perl v 5.32.1 with the following screening criteria: sequence length ≥ 300 bp, start codon ATG, and stop codons TAA/TAG/TGA. CodonW v 1.4.4 [33] and the online tool CUSP (https://www.bioinformatics.nl/cgi-bin/emboss/cusp) were employed to calculate the relative synonymous codon usage (RSCU), while various parameters, such as the effective number of codons (ENC) and GC content at each codon position (GC1/GC2/GC3/GCall/GC3s), were quantified. The results of codon preference were visualized using R v 3.6.0.

For repetitive sequence identification, microsatellites were detected using MISA v1.0 [34] with the following parameters: mononucleotide repeats ≥ 10, dinucleotide repeats ≥ 5, trinucleotide repeats ≥ 4, and tetranucleotide to hexanucleotide repeats ≥ 3. Tandem repeats were identified using TRF v4.09 [35]. Dispersed repeats were analyzed via the REPuter online platform (minimum repeat length 30 bp, similarity ≥ 90% [36]).

Sequence evolution and selection pressure quantification

Inter-organellar gene transfer events were identified using BLASTn v2.9.0+ [37] with parameters E-value ≤ 1e⁻⁵ and word size = 7, retaining only homologous sequences ≥ 1,000 bp in length. Circos v 0.69-6 [38] was used to visualize the distribution of transferred fragments.

Nucleotide diversity, collinearity, and Ka/Ks were analyzed for 11 aquatic monocot plants (Butomus umbellatus KC208619, H. beccarii PP936079–PP936106, H. beccarii, H. ovalis, Stratiotes aloides KX808393, Phyllospadix iwatensis OP441721, Zostera japonica OP441720, Z. marina OR336317-0R336318, Z. marina KX808392, R. sinensis PP438605, and Scheuchzeria palustris PQ031343). For nucleotide diversity (Pi) analysis, MAFFT v7.429 [39] was used to perform a global alignment of homologous genes, and DnaSP v6 [40] was used to calculate the Pi values, with a sliding window of 200 bp and step size of 100 bp to resolve gene variability. Conserved collinear blocks were identified via BLASTn v 2.9.0+ [37] homologous sequences, and multi-collinearity maps were generated using TBtools v2.119 [41]. For selection pressure analysis, ParaAT v2.0 [42] was used to align homologous gene pairs, and KaKs_Calculator v2.0 [43] was used to compute the Ka and Ks values for each gene pair using the YN method. Finally, the Ka/Ks ratios of each gene pair were statistically analyzed.

To detect gene-wide and site-specific positive selection, we performed a series of phylogenetically-informed analyses. The codon sequences of 23 homologous mitochondrial genes (atp1, atp4, atp6, atp8, ccmB, ccmC, ccmFc, ccmFn, cob, cox1, cox2, cox3, matR, mttB, nad1, nad2, nad3, nad4L, nad4, nad5, nad6, nad7, nad9) from 11 species were aligned using MAFFT v7.429 [39] with the codon alignment mode (Codon Table 1) and the auto strategy enabled. The alignments were filtered with Gblocks 0.91b [44] to retain conserved blocks (parameters: -t = c -b1 = 8 -b2 = 10 -b3 = 5 -b4 = 3 -b5 = h). Filtered sequences were converted to PAML format using PhyloSuite v1.2.1 [45]. A maximum likelihood phylogenetic tree was constructed from the filtered concatenated alignment using IQ-TREE v1.6.12 [46], which served as the species tree for subsequent selection analyses implemented in the CodeML program of PAML v4.10.7 [47]. First, the overall ratio of non-synonymous to synonymous substitution rates (ω) for each gene across all lineages was estimated under two neutral models: one incorporating the species tree (runmode = 0; model = 0; NSsites = 0) and another without tree information (runmode=-2; model = 0; NSsites = 0). Second, site-specific models were employed to test for positively selected sites across the entire evolutionary history by contrasting two pairs of selection and neutral models: (i) M2a (positive selection model; model = 0; NSsites = 2) versus M1a (neutral model; model = 0; NSsites = 1), and (ii) M8 (positive selection under beta distribution; model = 0; NSsites = 8) versus M7 (neutral model under beta distribution; model = 0; NSsites = 7). These models allow ω to vary among sites, enabling the detection of episodic positive selection that may not be captured by gene-wide averages. Furthermore, to investigate whether positive selection occurred specifically along the branch of interest (with H. beccarii set as the foreground branch and the remaining species as background), a branch-site model was applied. This model allows ω to vary both among sites and across branches, with the alternative model (model = 2; NSsites = 2; fix_omega = 0) permitting ω > 1 in the foreground branch and the null model (model = 2; NSsites = 2; fix_omega = 1) constraining ω = 1 for all lineages. Positively selected sites were identified using the Bayes Empirical Bayes (BEB) method implemented in PAML, with a posterior probability ≥ 0.95 considered significant. For all likelihood ratio tests (LRT), the test statistic was calculated as twice the absolute difference in log-likelihood values between the alternative model and the null model. P-values derived from LRTs were adjusted for multiple comparisons using the False Discovery Rate (FDR) correction procedure, with an FDR threshold of < 0.05 considered statistically significant. When model comparisons yielded significant results, the Akaike Information Criterion corrected for small sample sizes (AICc) was employed to select the most appropriate model for inferring positively selected sites.

Table 1.

Distribution of tandem and dispersed repeats in mitochondrial genomes across 11 selected aquatic monocot species

	Butomus umbellatus KC208619	Halophila beccarii PP936079-PP936106	Halophila beccarii	Halophila ovalis	Stratiotes aloides KX808393	Phyllospadix iwatensis OP441721	Zostera japonica OP441720	Zostera marina OR336317-OR336318	Zostera marina KX808392	Ruppia sinensis PP438605	Scheuchzeria palustris PQ031343
Tandem repeats distribution
5–19 bp	1	41	42	20	7	6	28	92	53	17	5
20–29 bp	5	15	14	16	1	5	13	10	7	5	8
30–39 bp	0	3	3	3	0	8	36	53	31	4	0
40–49 bp	1	2	2	2	0	0	2	0	0	0	1
50–59 bp	0	0	0	0	0	0	4	5	2	1	0
60–69 bp	0	0	0	9	0	0	7	12	9	1	0
70–79 bp	0	0	0	1	0	0	1	0	0	0	0
80–89 bp	0	0	0	0	0	0	0	8	3	0	0
90–99 bp	0	0	0	0	0	0	0	1	0	0	0
≥ 100 bp	0	0	0	1	2	0	0	5	1	1	0
SSR Motifs
Mono	21	286	292	1	10	5	12	6	3	1	26
Dimer	18	237	241	10	29	6	9	17	8	46	23
Trimer	10	70	70	14	7	7	4	4	2	1	8
Tetramer	29	573	580	51	34	13	12	16	10	16	31
Pentamer	4	37	36	2	8	2	2	0	0	13	4
Hexamer	0	2	3	0	0	0	0	2	1	0	0
Sequence orientation and complementation types of dispersed repeats
Palindromic	110	27	29	813	684	1494	2983	3885	2300	2869	249
Forward	118	59	61	851	711	1514	3292	4994	2674	2875	231
Reverse	0	4	4	0	0	0	0	0	0	0	0
Complement	0	1	1	0	0	0	0	0	0	0	0
Dispersed repeats distribution
30–39 bp	119	76	79	1096	1134	2159	4694	6670	3802	3753	398
40–49 bp	32	10	11	109	162	328	712	829	491	990	21
50–59 bp	17	3	3	62	48	206	364	584	273	802	18
60–69 bp	13	2	2	93	18	110	203	147	107	89	15
70–79 bp	3	0	0	15	10	65	85	150	89	64	8
80–89 bp	4	0	0	58	8	29	35	169	66	14	6
90–99 bp	7	0	0	12	2	18	54	47	40	7	1
100–199 bp	15	0	0	204	11	65	111	226	84	18	11
≥ 200 bp	18	0	0	15	2	28	17	57	22	7	2

Phylogenetic reconstruction and RNA editing detection

Sixteen monocot plants were subjected to phylogenetic analysis. Six single-copy homologous genes (atp8, ccmB, cob, mttB, nad1, and nad2) were extracted using Perl scripts. Multiple sequence alignment was performed using MAFFT v7.429 [39], with ambiguous regions trimmed using Gblocks 0.91b [44]. The concatenated sequences were used to construct a maximum-likelihood phylogenetic tree via IQ-TREE v1.6.12 [46]. Nucleic acid substitution models were optimized using ModelFinder, selecting GTR + F + R2 under the BIC criterion, and bootstrap resampling (1,000 replicates) was performed to validate the nodal support.

RNA editing events were investigated using a combination of computational prediction and transcriptome-based validation. Initially, RNA editing events were predicted using PmtREP [48] with a cutoff value of at 0.2. To empirically validate these events, we analyzed RNA-Seq data from GenBank (accession number of H. beccarii: SRR26321227; accession number of H. ovalis: ERR2109154). RNA-seq data were aligned to CDS sequences using bowtie2 v 2.4.1 [22], and the resulting alignments were sorted with Samtools v 1.9 [29]. Next, BCFtools v 1.10.2 [29] was employed to detect SNP sites between the sequencing data and the mitochondrial genome, which were considered potential RNA editing sites.

Results

Genomic structural characteristics and repetitive sequence dynamics

Whole-genome sequencing of H. ovalis (n = 1) and H. beccarii (n = 1) generated DNBSEQ-T7 data (67.54 Gb and 51.45 Gb, respectively) and HiFi data (15.7 Gb and 19.9 Gp, respectively), enabling successful assembly of their mitochondrial and chloroplast genomes. The mitochondrial genomes of both seagrass species exhibited multi-circular structures but differed significantly in size. H. ovalis comprised 12 free circles (total length = 584,100 bp) which are autonomous and no recombination detected under the conditions examined, harboring 26 PCGs, 13 tRNAs, and 3 rRNAs, with a GC content of 48.35% (Fig. 1, 8; Table S2); that of H. beccarii consisted of 28 free circles (total length = 1,976,414 bp) that are also autonomous and lack inter-circle recombination under the conditions examined, containing 26 PCGs, 10 tRNAs, and 3 rRNAs, with a GC content of 46.71% (Fig. 2, S8; Table S3). Their chloroplast genomes showed typical circular structures: H. ovalis (166,241 bp) contained 90 PCGs, whereas H. beccarii (168,964 bp) had 88 PCGs; both harbored 40 tRNAs and 8 rRNAs, with GC contents of 38.85% and 38.48%, respectively (Fig. S9, S10; Tables S4, S5).

Fig. 1.

Annotated map of the mitochondrial genome structure of H. ovalis. Genes inside the circle indicate a clockwise transcription direction, while genes outside the circle indicate a counterclockwise transcription direction. Genes with different functions are marked with distinct colors. The inner gray histogram shows the genomic GC content, with the middle gray line representing the 50% threshold

Fig. 8.

The phylogenetic relationships of Halophila with other monocot plants (a). Number of predicted RNA editing sites in individual mitochondrial protein-coding genes (PCGs) of H. ovalis (b) and H. beccarii (c). The data charts of the two H. beccarii samples are identical, so one is used as a substitute

Fig. 2.

Annotated map of the mitochondrial genome structure of H. beccarii. Genes inside the circle indicate a clockwise transcription direction, while genes outside the circle indicate a counterclockwise transcription direction. Genes with different functions are marked with distinct colors. The inner gray histogram shows the genomic GC content, with the middle gray line representing the 50% threshold

Codon usage bias analysis of mitochondrial PCGs from three Halophila specimens (two H. beccarii and one H. ovalis) revealed significant preferences for most codons, except the start codon AUG and the tryptophan codon UGG (with RSCU = 1 for the alanine codon GCC in H. ovalis). High-frequency codons (RSCU > 1) in two H. beccarii samples and H. ovalis predominantly featured A/U at the third base, whereas low-frequency codons (RSCU < 1) had G/C at this position. Notably, the arginine codon AGA showed the highest RSCU values across all samples: 1.7 in H. beccarii and 1.59 in H. ovalis (Fig. 3; Table S6). GC content analysis of Halophila mitochondrial genomes exhibited significant site-specificity: GC1 (47.75%–48.34%) > GC2 (40.10%) > GC3 (40.50%–41.30%). ENC-GC3 correlation analysis indicated that most genes fell below the standard curve, suggesting natural selection as the dominant evolutionary driver. Neutrality plots showed regression slopes of 0.0639–0.0703 between GC12 and GC3, further confirming the limited impact of mutational pressure on codon bias (Fig. 4; Table S7, S8).

Fig. 3.

The relative abundance of codon in mitochondrial genes of H. beccarii and H. ovalis. From left to right, the bar chart shows H. beccarii (PV999277-PV999304), H. beccarii (PP936079–PP936106) and H. ovalis

Fig. 4.

Codon usage bias analyses in mitochondrial protein-coding genes of H. beccarii and H. ovalis. ENC-GC3 plot analysis in H. beccarii (a) and H. ovalis (d). Pearson’s correlation analysis heatmap of codon usage indicators of H. beccarii (b) and H. ovalis (e). Neutrality plot analysis of GC12 and the third codon position (GC3) in H. beccarii (c) and H. ovalis (f). The data charts of the two H. beccarii samples are identical, so one is used as a substitute

Cross-species comparisons revealed a universal loss of ribosomal protein genes (RPGs) in aquatic plants: all seagrass and most freshwater aquatic plants mitochondrial genomes lacked at least 12 RPGs (rpl16, rps2, rps4, rps10, rps11, rps13, rps14, rps19, rpl2, rpl5, sdh3, and sdh4), with the Halophila genus retaining only rps12 and rpl10; terrestrial plants had more RPGs preserved (Fig. 5).

Fig. 5.

Comparison of the mitochondrial genes of the 16 monocot plants. Yellow denotes genes present in the mitochondrial genome, white denotes genes lost from the mitochondrial genome, and gray denotes pseudogenes. For duplicated gene copies, red indicates 3 copies, green indicates 2 copies, and yellow indicates 1 copy

Characteristics and distribution of sequence repeats

Mononucleotide, dinucleotide, trinucleotide, and tetranucleotide repeat sequences were all found in the mitochondrial genomes of the 11 aquatic plants compared. Except for Z. marina, pentanucleotide repeat sequences were found in the mitochondria of the nine other species compared. However, hexanucleotide repeat sequences were only found in small amounts in H. beccarii and Z. marina. In addition, tetranucleotide repeat sequences were the most abundant in nine species, except for Z. marina (OR336317–OR336318) and R. sinensis (Table 1, S9). Interestingly, sequence repeat analysis uncovered significant expansion in H. beccarii, where the SSR counts far exceeded those in other species (Table 1, S9).

In the mitochondria of the 11 aquatic plants investigated, Z. marina (OR336317-OR336318) had the highest number of tandem repeat sequences (186), followed by Z. marina (KX808392) with 106. In contrast, B. umbellatus had the fewest tandem repeat sequences, with only seven. Repeat sequences 5–19 bp long were the most abundant across all samples, totaling 312, followed by repeat sequences 30–39 bp long, with a total of 141 such repeats across the mitochondria of the 11 samples (Table 1, S10).

Analysis of interspersed repeats in the mitochondrial genomes of the 11 compared samples revealed that, except for the mitochondrial genomes of H. beccarii, which detected forward repeats, palindromic repeats, reverse repeats, and complement repeats, the mitochondrial genomes of the other nine species only contained forward repeats and palindromic repeats. The abundance of each repeat type varied by sample. Except for S. palustris, where palindromic repeats were the most abundant, forward repeats were the most abundant in the mitochondria of the other 10 species. The 30–39 bp repeat category was the most numerous across all samples, with a total of 23,980, followed by 40–49 bp repeat sequences, which totaled 3,695 (Table 1, S11).

Sequence evolution and selection pressure

Four, ten, and ten chloroplast-derived sequences were identified in the mitochondrial genomes of H. ovalis, H. beccarii (PV999277- PV999304), and H. beccarii (PP936079–PP936106), respectively, with total lengths of 4,598, 16,552, and 16,526 bp, respectively (Fig. 6a and b; Table S12). Collinearity analysis revealed the strongest structural conservation within the Halophila genus, with 32 homologous blocks (covering 49,610 bp) identified between H. beccarii and H. ovalis. In cross-genus comparisons (e.g., B. umbellatus vs. H. beccarii), only 11 blocks (15,038 bp) were detected (Fig. 6c; Table S13).

Fig. 6.

The gene transfers that occurred between the chloroplast and mitochondrial genomes of H. beccarii (a) and H. ovalis (b); lines connecting the arcs correspond to genomic segments that are homologous between chloroplasts and mitochondria, with lengths exceeding 1000 bp; The data charts of the two H. beccarii samples are identical, so one is used as a substitute. Synteny analysis of mitochondrial genomes between Halophila and other aquatic monocotyledonous plants (c); regions connected by arcs represent regions with high homology, wherein red arcs denote regions with reverse sequence orientation and gray regions denote regions with forward sequence orientation

Nucleotide diversity (Pi) was highest for the atp8 gene (Pi = 0.10791) and lowest for nad4L (Pi = 0.0327). Similarly, genes cob, cox3, nad1, nad2, nad3, nad4L, nad5, nad7, and nad9 exhibited Pi values below 0.05 (Fig. 7a; Table S14). In the pairwise Ka/Ks analysis across lineages, the nad3 gene exhibited a mean Ka/Ks > 1, while atp1 and cox1 showed Ka/Ks much less than 1. A subset of 12 genes demonstrated Ka/Ks > 1 in specific pairwise comparisons. In the direct comparison between H. beccarii and H. ovalis, cox1, nad1, nad3, and nad7 had the lowest Ka/Ks ratios (< 0.1) (Fig. 7b; Table S15).

Fig. 7.

Nucleotide diversity analysis of mitochondrial genomes across Halophila and other aquatic monocotyledonous plants (a). Ka/Ks analysis of mitochondrial protein-coding genes across Halophila and other aquatic monocotyledonous plants (b)

Phylogeny-aware selection analysis using CodeML (PAML) further delineated distinct evolutionary regimes. The results showed that most genes (22 out of 23) had a gene-wide dN/dS ratio (ω) less than 1. The ω values ranged from 0.0597 to 0.76267. Among them, cox1 and nad4L had the lowest ω values (ω < 0.15) and showed very low mean pairwise ω. Notably, the nad3 gene was an exception, with an overall ω value of 1.12595. It also exhibited the highest transition/transversion rate ratio (kappa = 3.71196) among all genes. Pairwise comparisons showed ω > 1 for genes like nad3, atp8, and nad5 in specific distant species pairs. Site-specific positive selection was tested using LRTs, followed by FDR correction for multiple comparisons. Seven genes showed statistically significant signals of positive selection at an FDR threshold of < 0.05. Among them, matR and nad7 were highly significant (FDR < 0.01). atp1, ccmFn, nad2, nad3, and nad4 were also significant (0.01 ≤ FDR < 0.05). The BEB analysis identified 2 to 6 positively selected sites (posterior probability BEB ≥ 0.95) within each of these seven genes. Site-model comparisons (M8 vs. M7 and M2a vs. M1a) congruently identified a set of genes (ccmFn, matR, nad3, nad4 and nad7) under significant site-specific positive selection. A branch-site model test detected no significant positive selection specific to the H. beccarii lineage (LRT_branchsite = 0、pvalue = 1、FDR = 1). The nad3 gene showed a distinct pattern, being the only gene for which both the gene-wide ω was greater than 1 and site-specific positive selection was statistically supported (FDR < 0.05), with two positively selected sites identified (BEB ≥ 0.95). However, in pairwise comparisons between closely related species (within the same genus) and among intraspecific samples, the ω values for nad3 were substantially below 1 (Table S16).

Phylogenetic relationships and RNA editing

A phylogenetic tree constructed based on six single-copy mitochondrial genes (atp8, ccmB, cob, mttB, nad1, and nad2) showed that the Halophila genus formed a monophyletic group, in which H. beccarii samples clustered into a clade and constituted a sister group with the freshwater species S. aloides (Fig. 8a).

In total, 193, 193, and 197 RNA editing sites were found across 26 PCGs in H. beccarii (PP936079–PP936106), H. beccarii (PV999277-PV999304), and H. ovalis, respectively (Fig. 8b and c; Table S17-S19). The RNA editing results of the mitochondria of the three Halophila samples investigated showed that all identified potential RNA editing sites were C-to-T(U) edits, with the matR gene having the highest number of editing events among all mitochondrial genes. Furthermore, RNA editing analysis revealed specific deletions: nad3 and nad6 in H. beccarii completely lacked RNA editing, whereas cox1, nad3, and nad6 in H. ovalis showed the same editing absence. RNA editing analysis results were validated using RNA-seq data. A total of 31 and 65 RNA editing sites were identified in H. beccarii and H. ovalis, respectively. Additionally, it was consistently found that the nad3 and nad6 genes lacked editing sites in both species (Table S20, S21).

Discussion

This study presents the first systematic revelation of the adaptive evolutionary mechanisms in mitochondrial genomes of the marine plants H. ovalis and H. beccarii. Through comparative genomics and evolutionary analyses, we demonstrate that seagrasses employ multidimensional strategies—including structural remodeling, molecular evolutionary innovations, and regulatory simplification—to adapt to high-salinity environments.

Structural dynamics and functional evolution of mitochondrial genomes in halophila

Compared with terrestrial plants, aquatic plants exhibit a massive loss of RPGs, with the Halophila genus retaining only rps12 and rpl10 (Fig. 5). It is important to note that RPG loss is an established trend across diverse angiosperm lineages [49]. However, non-essential gene simplification has been reported to potentially be associated with adaptation during the transition from terrestrial to aquatic environments [14], analogous to gene family contraction observed in terrestrial plants evolving toward saltmarsh habitats. For instance, Thellungiella parvula showed a reduced number of signal transduction-related and disease resistance genes, while the number of stress-adaptive genes expanded [50]. In the adaptation of Aegiceras corniculatum to intertidal saltmarsh environments, 3,096 gene family members were reduced compared with closely related terrestrial species [51]. Adams et al. [52] similarly reported frequent losses of RPGs and sdh genes, together with high levels of conservation of core respiratory genes across angiosperms. These cases suggest that gene family contraction might streamline stress response by eliminating genetic redundancy, though this link requires further functional validation. Similarly, our finding of massive RPGs loss in seagrasses presents a genomic profile characterized by the simplification of non-essential genes concurrent with the high conservation of core respiratory chain genes (e.g., cob, cox3, and multiple nad genes; Pi < 0.05). Its prevalence in marine species may be associated with environments having elevated energy demands, though the adaptive significance of this correlation remains to be tested.

This study compared the mitochondrial genome characteristics of 11 aquatic plant species (Table S9-S11). The mitochondrial genomes of most species exhibit a single-chromosome structure (Table S9-S11). However, H. beccarii (28 chromosomes) and H. ovalis (12 chromosomes) show distinct multi-chromosome characteristics, suggesting that mitochondrial genomes in Halophila may have undergone frequent structural rearrangement events. Notably, repeat sequence distribution further reflects species-specific adaptation: H. beccarii accumulates the highest total SSR length and number among all studied species, whereas it has the smallest number and total length of dispersed repeats compared to other species in this study. Thus, the mitochondrial genome of H. beccarii presents a notable novel case challenging traditional views: it exhibits an extreme multi-circular architecture (28 chromosomes) despite a remarkably low content of large dispersed repeats. This observation challenges the prevailing paradigm that attributes extensive recombination in plant mitochondria primarily to homologous recombination between large repeated sequences [53–56]. The co-occurrence of high SSR density and low dispersed repeat content in this highly fragmented genome invites speculation about a potential link. Given that SSRs are recognized hotspots for DNA strand slippage and double-strand breaks (DSBs) and are potential substrates for microhomology-mediated end joining (MMEJ) in plant mitochondria [31, 57–64], one might cautiously consider whether prolific SSRs could be linked to enhanced genomic plasticity. We speculate that a genome rich in SSRs might be more susceptible to rearrangement events through mechanisms like MMEJ. Whether this susceptibility plays a role in the observed fragmentation, and if so, to what extent compared to other unknown factors, remains an open question. This speculative model, which posits a potential association between SSR abundance and genome fragmentation, provides a conceptual framework for future research to explore the diverse evolutionary strategies shaping mitochondrial genome architecture. Furthermore, plastid-to-mitochondrion gene transfer events may contribute to adaptive evolution: ten and four chloroplast-derived sequences were identified in H. beccarii and H. ovalis, respectively. This phenomenon aligns with the general rule of inter-organellar genetic integration in plants [65, 66], where intact gene transfer enhances adaptability through functional compensation.

Molecular evolutionary drivers and regulatory elements

The third base of high-frequency codons showed a significant bias toward A/U (RSCU > 1), with the arginine AGA codon as the most dominant type (RSCU = 1.59–1.70), a pattern consistent with the natural selection-driven codon optimization theory. In various organisms, such preferences have been shown to enhance translation efficiency by matching abundant tRNAs [67, 68]; however, the specific mechanism in Halophila requires further validation via tRNA omics. ENC-GC3 correlation analysis indicated that ca. 80% of genes of Halophila were dominated by natural selection, while the regression slope (0.0639–0.0703) from the neutrality plots excluded mutational pressure as the primary driver (Fig. 4; Table S7). This pattern is widespread in eukaryotic mitochondria: bacterial and insect mitochondrial codon biases enhance adaptability by improving translation efficiency [67, 69], whereas plant mitochondria similarly exhibit selection-driven conservation [70]. Notably, high-RSCU codons (e.g., AGA) in genes associated with energy metabolism can significantly accelerate protein synthesis [68].

The evidence for selection on codon usage prompted us to examine the potential for selection at the level of protein function. Our analyses revealed a consistent signal of elevated relative substitution rates (Ka/Ks) for the nad3 gene across both pairwise comparisons and phylogeny-aware branch model analyses (gene-wide ω > 1). Further site-specific selection tests identified statistically significant positive selection in nad3 and six other genes (atp1, ccmFn, matR, nad2, nad4, nad7), with nad3 being unique in displaying both gene-wide and site-specific signals. While pairwise Ka/Ks analysis alone cannot definitively infer selection mode, the convergence of branch and site-model results for nad3 strengthens the case for its involvement in adaptive evolution, although the observed pattern could also be consistent with a prolonged period of relaxed purifying selection or, in lineages with historically small effective population sizes, the fixation of mutations by genetic drift. This finding aligns with previous studies linking nad3 to stress responses, marking it as a candidate gene worthy of targeted investigation for its potential role in environmental adaptation [71–73]. In contrast, genes like atp1 and cox1 showed low ω values, underscoring their high evolutionary conservation, which is typical for fundamental components of the oxidative phosphorylation machinery.

Phylogenetic analysis supported the monophyly of Halophila and its sister group relationship with the freshwater species S. aloides, providing evidence for the ecological transition of monocots from freshwater to marine environments [14]. This aligns with multiple previous studies on seagrass phylogeny, where analyses based on different molecular markers and genomic data support the origin of seagrasses from land ancestors followed by marine expansion [74]. At the gene expression regulation level, we acquired 193, 193, 197, RNA editing sites within all the PCGs in the mitochondrial genome of H. beccarii (PP936079–PP936106), H. beccarii, and H. ovalis, respectively (Fig. 8); such number was significantly less than that of other terrestrial monophyllous plants or other angiosperms [75], such as Oryza granulata(488 editing sites [76]), Indocalamus longiauritus (602 editing sites [77]), Cocos nucifera (734 editing sites [78]), Phoenix dactylifera (600 editing sites [79]), and B. umbellatus (557 editing sites [75]). RNA editing deficiencies were particularly prominent: nad3 and nad6 in H. beccarii, and cox1, nad3, and nad6 in H. ovalis completely lacked RNA editing. This suggests a trend toward transcriptional simplification in Halophila mitochondria. We propose that this could represent an adaptive strategy for streamlining gene expression in the marine environment. However, non-adaptive explanations, such as genetic drift or relaxed selective constraint, must also be considered as equally plausible mechanisms driving this pattern. This hypothesis that merits future functional investigation.

Limitations of the study

It is important to note that our analyses are based on mitochondrial genome sequences from a single individual per species. While this approach is standard for de novo organellar genome assembly and provides a robust foundation for interspecific comparisons, it inherently limits our ability to assess intraspecific genetic variation or to make population-level inferences about evolutionary dynamics. Future studies incorporating multiple individuals per species will be valuable to explore the diversity within these taxa.

Authors’ contributions

J.S.W. and C.S.Q. designed the research, J.S.W., Z.J., and W.Y. conducted field investigation and sample collection, J.S.W., S.J., C.Z.F and W.Y. conducted the data analysis, J.S.W. wrote the draft, C.S.Q. improved the manuscript. All authors contributed to the interpretation of the results and approved the final manuscript.

Funding

This study was supported by the Hainan provincial Natural Science Foundation of China (ZDYF2024SHFZ146, 323RC554), Department budget projects of Hainan provincial in 2026 (Operation and Maintenance Project of Tropical Typical Marine Ecosystem Research Center), National Natural Science Foundation of China (42306169, 42166006).

Data availability

The raw sequencing data generated in this study have been deposited in the NCBI Sequence Read Archive (SRA) under the accession numbers PRJNA1303958 and PRJNA1303959.The assembled mitochondrial genome sequences of *H. beccarii* and *H. ovalis* have been deposited in GenBank under the accession numbers PV999277- PV999304 and PV999242- PV999253, respectively.The assembled chloroplast genome sequences of the two species have been deposited in GenBank under the accession numbers PV993969 and PV993971, respectively.The multiple sequence alignments and individual gene sequences used in the phylogenetic analyses have been included in the Supplementary Dataset of this article Tables S22-S28.

Declarations

Ethics approval and consent to participate

This study did not involve human subjects, animal experiments, protected species, or biological samples requiring ethical approval. Thus, no ethics approval was necessary.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.