Bipartite mitogenome of Fargesia angustissima reveals repeat-mediated structural dynamics, inter-organellar DNA transfers, and phylogenetic discordance

Abstract

Background

As an endemic bamboo and a critical dietary staple for the endangered giant panda, Fargesia angustissima T. P. Yi is integral to the ecological stability of Western Sichuan’s forest ecosystems. Nevertheless, the absence of a complete mitogenome has constrained investigations into its complex genomic architecture, evolutionary history, and the genetic basis underlying the montane adaptation.

Results

Using a hybrid sequencing strategy (BGI short reads and PacBio HiFi long reads), we de novo assembled and characterized the bipartite mitogenome of F. angustissima (429,279 bp, 44.18% GC content). The mitogenome encodes a conserved repertoire of 37 unique protein-coding genes (PCGs), exhibiting a pronounced A/U-ending codon bias driven by directional mutation pressure. Long-read mapping validated that a 7,368 bp palindromic repeat (R1) maintained major bipartite isomers (45.45% recombinant frequency), while 39 tetrameric Simple Sequence Repeats (SSRs) and 298 dispersed repeat pairs mediate localized structural variations. To counteract sequence-level mutational drift, 488 C-to-U RNA-editing sites impose strict functional stringency. These modifications predominantly convert hydrophilic Ser/Pro to hydrophobic Leu (42.0%) to ensure the proper membrane integration of respiratory complexes, while also generating obligate stop codons in atp6, atp9, and ccmFC. Moreover, we identified 27 plastid-derived sequences (MTPTs) spanning 20.151 kb, demonstrating an asymmetric intracellular DNA migration pattern; the plastid Large Single Copy (LSC) region (e.g., MTPT27) serves as an evolutionary hotspot for sequence transfer, contrasting with the structural conservatism of the Inverted Repeat (IR) regions (MTPT1, MTPT11, and MTPT12). Despite high sequence identity (> 99.8%) across the Bambusoideae, synteny analyses revealed extensive structural rearrangements. Subsequent phylogenetic reconstruction uncovered significant inter-organellar topological conflicts, providing genomic evidence for historical reticulate evolution and incomplete lineage sorting (ILS) during the rapid diversification of the Fargesia genus.

Conclusions

The F. angustissima mitogenome exemplifies a paradigm of structural plasticity, post-transcriptional constraint, and phylogenetic discordance. By overcoming the low nucleotide diversity barrier inherent to temperate bamboos, a structurally and functionally derived marker framework was established. The genomic reference and the associated candidate loci provide a molecular blueprint for non-destructive population genetics and data-driven conservation within giant panda habitats.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12870-026-09566-8.

Introduction

Plant mitochondria are essential, semi-autonomous, and maternally inherited organelles that regulate cellular respiration and energy metabolism, thereby fundamentally impacting plant growth and stress acclimation [1]. In contrast to the highly conserved plastome, plant mitochondrial genomes (mitogenomes) display extraordinary structural complexity and dynamic features, underpinning evolutionary fitness and genetic stability [2, 3]. They frequently exhibit size fluctuations, extensive genomic rearrangements, incorporation of foreign DNA sequences (e.g., Mitochondrial Plastid DNAs/MTPTs), and prevalent RNA-editing events [4]. Consequently, rather than existing as simple, single circular molecules, they often present complex physical conformations, including multi-branched, linear, and multi-chromosomal circular structures [5–8]. Beyond single-species characterization, comparative phylogenetic analyses of mitogenomic structural variation provide novel insights into species diversification. The advent of long-read sequencing technologies (e.g., PacBio HiFi) coupled with specialized assembly methods (e.g., PMAT) has significantly accelerated the exploration of these structural dynamics across diverse evolutionary lineages [9, 10]. Thus, deciphering the complete architecture of a mitogenome is critical for understanding the molecular dynamics of intracellular gene transfer, energy regulation, and the genetic basis of environmental adaptation in terrestrial plants [11, 12].

Fargesia, comprising over 90 recognized species endemic to East Asia, is the largest genus of temperate montane bamboo within the tribe Arundinarieae (Bambusoideae: Poaceae). These bamboos are functional to regional forest ecosystems, providing primary forage and habitat for endangered wildlife, most notably the giant panda (Ailuropoda melanoleuca) [13, 14]. Fargesia represents a taxonomically recalcitrant and climate-sensitive lineage, widely known for dominating subalpine environments [15]. Its recent explosive radiation, combined with protracted vegetative cycles and restricted seed dispersal, has led to shallow interspecific divergence, compromising the effectiveness of standard DNA barcoding [16, 17]. Moreover, pervasive historical hybridization has driven a highly reticulate evolutionary fate, blurring species boundaries and generating profound inter-organellar phylogenetic discordance [18]. Among these taxa, Fargesia angustissima T. P. Yi (commonly known as You bamboo) holds paramount ecological and conservation significance. Naturally distributed across the Qionglai Mountains in western Sichuan Province, China, it thrives at mid-to-low elevations (700-1,500 m) in montane environments, rather than strictly subalpine zones [19]. Within this specific niche, they form dense understory populations and serve as a critical staple food source for the giant panda [20]. The ecological persistence of F. angustissima in these montane habitats directly underpins the foraging stability and survival of endangered fauna.

Despite its vital ecological profile, genomic research on F. angustissima has been predominantly confined to the plastome [20]. The critical absence of mitogenomic data severely limits our understanding of its structural dynamics and inter-organellar phylogenetic incongruence, key genomic indicators necessary for resolving environmental adaptation, ancient introgression, and incomplete lineage sorting (ILS) within the genus Fargesia [21]. Deciphering the F. angustissima mitogenome is therefore a fundamental prerequisite for unraveling the complex and reticulate evolutionary history of temperate woody bamboos.

To address this gap, we employed an integrated hybrid sequencing strategy (BGI short reads and PacBio HiFi long reads) to de novo assemble the complete mitogenome of F. angustissima. In this study, we systematically profile its bipartite chromosomal architecture, repetitive sequence landscape, asymmetric MTPT migration patterns, and functional RNA-editing events. By reconstructing dual-organellar phylogenies, we elucidate the profound topological discordance and independent evolutionary trajectories of the mitogenome and plastome. An integrated organelle-level perspective yields structural and post-transcriptional mitogenomic markers, establishing a genetic foundation for monitoring population genetics and guiding the conservation of this critical giant panda dietary staple under shifting environmental conditions.

Materials and methods

Plant material, DNA extraction, and high-throughput sequencing

Fresh leaves of F. angustissima were collected from the eastern slopes of the Sichuan Wolong National Nature Reserve (elevation: 1200–1800 m; Coordinates: 103° 22’ 22.4” E, 31° 03’ 57.1” N), within the Qionglai Mountains, China. Following formal identification by Wenbin Yang, a voucher specimen was deposited in the herbarium of Nanjing Forestry University (ID: ST2023111005). Upon collection, 100 mg of leaf tissue was immediately flash-frozen in liquid nitrogen and stored at -80 °C until further processing. Total genomic DNA was extracted via the modified cetyltrimethylammonium bromide (CTAB) method [22]. DNA integrity, purity, and concentration were assessed using a Nanodrop 2000c UV/Vis spectrophotometer (Thermo Fisher Scientific, USA) and 1% agarose gel electrophoresis. The specific parameters for the library construction were based on the previous report [22]. Hybrid sequencing was performed by Wuhan Benagen Technology Co., Ltd. (Wuhan, China) using the DNBSEQ-T7 (BGI, China) and PacBio Sequel II (Pacific Biosciences, USA) platforms.

De novo mitogenome assembly and gene annotation

The de novo assembly of the F. angustissima mitogenome was performed using a hybrid sequencing strategy. Initial unitig assembly graphs were generated in GFA format using GetOrganelle v1.7.7.1 for the BGI short reads (150 bp) and OATK for the PacBio HiFi long reads (10–25 kb) [23]. Candidate mitochondrial contigs were isolated via BLASTn v2.16.0 alignments [24]. To resolve repetitive regions and multi-branched structures, the PacBio HiFi long reads were mapped back to the candidate contigs with BWA v0.7.1 [25]. The resulting assembly graph was manually curated in Bandage v0.9.0 to trace the optimal long-read-supported paths, finalizing the physical conformation of the mitogenome [26]. Structural and functional annotations were conducted using GeSeq v2.03 [27]. Specific ribosomal RNA (rRNA) and transfer RNA (tRNA) genes were verified using BLASTn and tRNAscan-SE v2.0.12 [28]. Automated annotation outputs were imported into Apollo v2.7.0 for manual inspection and boundary correction [29]. Finally, the circular and linear genomic maps were rendered using OGDRAW v1.3.1 [30].

Codon usage, mutation pressure, and repeat analyses

All protein-coding genes (PCGs) were extracted from the annotated mitogenome using PhyloSuite v1.2.3 [31]. To evaluate codon usage bias, the relative synonymous codon usage (RSCU) values were calculated in MEGA 12.1 [32]. To elucidate mutational dynamics, positional GC contents across the first (GC1), second (GC2), and third (GC3) codon positions were profiled using the EMBOSS cusp tool [33]. The Effective Number of Codons (ENC) for each PCG was quantified with CodonW v1.4.4 (https://sourceforge.net/projects/codonw/files/) to assess the absolute magnitude of codon bias and distinguish directional mutation pressure from natural selection impediments. Simple sequence repeats (SSRs) were mined using MISA-web v2.1, applying minimum repeat thresholds of 10, 5, 4, 3, 3, and 3 for mono-, di-, tri-, tetra-, penta-, and hexanucleotide motifs, respectively [34]. Tandem repeats were detected using the Tandem Repeats Finder program v4.09 [35]. Dispersed repeats (palindromic and forward repeats ≥ 30 bp) were characterized with REPuter v1.0 [36]. The genomic distribution of these repetitive elements was mapped using the RCircos package v1.2.2 [37].

Identification of MTPTs and RNA-editing sites

The complete F. angustissima plastome was de novo assembled from BGI short reads using GetOrganelle v1.7.7.1 [38]. The plastome was subsequently annotated and manually curated using CPGAVAS2 and CPGView pipelines [39, 40]. Homologous sequence alignments were performed between the assembled mitogenome and plastome using BLASTn (E-value cutoff ≤ 1 × 10^{− 5}) to identify integrated plastid fragments [24]. The resulting inter-organellar DNA transfers were visualized as a synteny chord diagram within TBtools [41].

Potential C-to-U RNA-editing events across the mitochondrial PCGs were predicted using a convolutional neural network-based tool, Deepred-mt, retaining only high-confidence sites with a probability score ≥ 0.9 [42]. BEDTools v2.31.1 was deployed to filter and validate predicted sites mapped within coding sequences (CDS) [43]. To distinguish genuine RNA-editing events from stochastic sequencing noise, strict quantitative thresholds were applied: a minimum sequencing depth (coverage > 5) to prevent false positives from low-coverage stochasticity, a baseline editing threshold (variant frequency ≥ 0.1) to filter background sequencing errors, and a statistical significance filter (p ≤ 0.05), consistent with established benchmarking protocols [11].

Synteny and phylogenetic analyses

The conserved syntenic blocks (≥ 500 bp) across related bamboo taxa were identified using BLASTn [24]. Synteny analyses and the corresponding visualizations were generated using MCScanX [44]. For phylogenetic reconstructions, complete mitogenomes (47 species) and plastomes (45 species) were extracted and curated using PhyloSuite v1.1.16 [31]. Following multiple sequence alignments with MAFFT v7.525, we inferred maximum-likelihood (ML) phylogenies with IQ-TREE v2.3 using the best-fit substitution model and 1000 bootstrap replicates [31, 45]. The resulting phylogenetic topologies were visualized and annotated using the interactive Tree Of Life (iTOL) v7 web server [46].

Results

Mitogenome assembly, recombination dynamics, and bipartite architecture

We de novo assembled the F. angustissima mitogenome using PacBio HiFi long reads (N50 = 16,827 bp) to resolve complex repetitive regions (Table S1). The initial unitig assembly graph displayed a multi-branched topology consisting of six interconnected contigs, with lengths ranging from 17,198 bp to 180,261 bp and an average sequencing depth of 72.5 × (Table S2). The graph is characterized by multiple interconnected nodes (contigs) and edges (read-supported linkages), distinctly departing from a canonical unipartite circular map (Fig. 1A and Table S3). Mapped PacBio HiFi reads unambiguously bridged the five repetitive nodes (ctg2-ctg5-ctg3-ctg4-ctg1), providing path evidence to order and orient them into the major conformation (Chromosome 1). Specifically, ctg6 was assembled as a distinct, standalone circular molecule, designated as Chromosome 2 (Fig. 1B). The decoded bipartite mitogenome totals 429,279 bp in length (44.18% GC), comprising the principal Chromosome 1 (412,081 bp, 44.27% GC) and the subsidiary Chromosome 2 (17,198 bp, 41.96% GC) (Table S4).

Fig. 1.

Assembly graph and physical map of the bipartite F. angustissima mitogenome. A Initial assembly graph displaying six principal contigs (ctg1–ctg6) with their respective sequence lengths and depths. B Resolved bipartite architecture. PacBio HiFi long-read mapping disambiguated overlapping repetitive paths, yielding a major chromosome (Chromosome 1: ctg2-ctg5-ctg3-ctg4-ctg1) and an autonomous sub-chromosome (Chromosome 2: ctg6). C Linear physical maps detailing gene distribution across the two mitochondrial chromosomes. Color-coded blocks indicate distinct functional gene categories

To untangle the multipartite assembly graph and validate the structural dynamics in vivo, we performed stoichiometric remapping of PacBio HiFi long reads across predicted repeat elements. The quantitative approach identified 21 dispersed repeat pairs acting as active homologous recombination hotspots (Fig. S1 and Table S5). Consistent with length-dependent dynamics, the 7,368-bp palindromic repeat (R1) drove high-frequency events, maintaining a near-equilibrium stoichiometric ratio of major to recombinant conformations (54.55% vs. 45.45%). Notably, a 169-bp forward repeat (R11) uniquely functioned as an inter-chromosomal hinge mediating the dynamic excision of Chromosome 2. In addition, read mapping from orthogonal sequencing platforms exhibited continuous, gap-free coverage across all contig boundaries, verifying the structural completeness of the assembly (Fig. S2). Collectively, the integration of congruent coverage plots with quantitative read-spanning data confirms that the observed topological complexity represents a dynamic in vivo population of recombinant isomers.

Functional annotation identified 64 genes (61 unique), including 37 core PCGs, three rRNA genes (rrn5, rrn18, and rrn26), and 21 tRNA genes (Fig. 1C; Table 1). Among PCGs, 24 are involved in oxidative phosphorylation: nine NADH dehydrogenase genes (nad1, nad2, nad3, nad4, nad4L, nad5, nad6, nad7, and nad9), five ATP synthase genes (atp1, atp4, atp6, atp8, and atp9), three cytochrome c oxidase genes (cox1, cox2, and cox3), four cytochrome c biogenesis genes (ccmB, ccmC, ccmFC, and ccmFN), one cytochrome b gene (cob), one succinate dehydrogenase gene (sdh4), and one protein transport subunit gene (mttB). Furthermore, the mitogenome contains 12 ribosomal protein genes (large subunits: rpl2, rpl5, and rpl16; small subunits: rps1, rpl2, rpl3, rpl4, rpl7, rpl11, rpl13, rpl14, and rpl19) alongside one maturase gene (matR). Physical mapping revealed that, except for atp4 and atp9, which are located exclusively on Chromosome 2, all other unique genes are distributed across Chromosome 1. Structural annotation identified cis- or trans-splicing introns within several specific genes, including nad1, nad2, nad4, nad5, nad7, ccmFC, and rps3. Additionally, accounting for genomic duplication, three tRNAs (trnM-CAU, trnN-GUU, and trnP-UGG) were found to be multi-copy.

Table 1.

The list of gene annotations in the F. angustissima mitogenome

Gene Group	Gene Name
ATP synthase	atp1, atp4, atp6, atp8, atp9
NADH dehydrogenase	nad1, nad2, nad3, nad4, nad4L, nad5, nad6, nad7, nad9
Cytochrome b	cob
Cytochrome c biogenesis	ccmB, ccmC, ccmFC, ccmFN
Cytochrome c oxidase	cox1, cox2, cox3
Maturases	matR
Protein transport subunit	mttB
Succinate dehydrogenase	sdh4
Ribosomal large subunit	rpl2, rpl5, rpl16
Ribosomal small subunit	rps1, rps2, rps3, rps4, rps7, rps11, rps13, rps14, rps19
Ribosome RNA	rrn5, rrn18, rrn26
Transfer RNA	trnC-GCA, trnD-GUC, trnE-UUC, trnF-GAA, trnfM-CAU, trnH-GUG, trnI-CAU, trnK-UUU, trnL-CAA, trnM-CAU (×2), trnN-GUU (×2), trnP-UGG (×2), trnQ-UUG, trnS-CGA, trnS-GCU, trnS-GGA, trnS-UGA, trnT-UGU, trnV-UAC, trnW-CCA, trnY-GUA

The (×2) represents the gene’s copy number

Codon usage bias and mutational signatures

We analyzed the RSCU across all 37 PCGs to evaluate translational efficiency and evolutionary bias in F. angustissima. These concatenated CDS comprise 11,092 codons (including termination codons), revealing a pronounced genome-wide codon usage bias. The amino acid composition is dominated by leucine (Leu) (6 codons, 10.49%), whereas Tryptophan (Trp) is the rarest (1 codon, 1.56%) (Fig. 2A and Table S6). Excluding the invariant AUG and UGG codons (RSCU = 1.0), 30 codons displayed strong usage preference (RSCU > 1.0). Notably, 27 of these preferred codons (90%) terminate in A or U, reflecting a strong third-position AT bias characteristic of plant mitogenomes. Within this biased repertoire, Alanine (Ala) exhibited the strongest preference for the GCU codon (RSCU = 1.57), closely followed by Glutamine (Gln) for the CAA codon (RSCU = 1.56). Furthermore, translation termination is particularly governed by the UAA stop codon.

Fig. 2.

Codon usage patterns and evolutionary drivers of 37 PCGs in the F. angustissima mitogenome. A RSCU profiles. Different colors within the stacked bars distinguish synonymous codons for respective amino acids. An RSCU > 1.0 signifies positive codon usage bias. B ENC-GC3s scatter plot. The solid black curve delineates the expected ENC under pure mutational pressure at GC3s. Red dots denote the observed ENC values of individual PCGs

To elucidate the evolutionary drivers of this bias, we evaluated the ENC and positional GC content. The PCGs exhibited an average ENC of 53.70 (range: 38.65–61.00) (Table S7). This elevated value (> 45) denotes a relatively weak overall codon bias, suggesting the absence of stringent natural selection. Furthermore, positional nucleotide analysis revealed a distinct, descending GC content gradient across three codon positions: GC1 (47.60%) > GC2 (42.75%) > GC3 (39.45%). Pronounced AT accumulation at the third wobble position (low GC3) corroborates the observed A/U-ending preference. An ENC-GC3 scatter plot showed that most PCGs cluster tightly below the theoretical expectation curve, indicating that ENC values are determined by mutational pressure (Fig. 2B and Table S8). Notably, atp9 emerged as a distinct outlier with the lowest ENC value (38.65), suggesting strong functional selective constraints. Collectively, these metrics provide genomic evidence that directional mutation pressure, rather than natural selection, primarily dictates codon usage in the F. angustissima mitogenome.

Repeat sequences landscape and structural plasticity

The mitogenome harbors abundant repetitive sequences, including tandem and dispersed repeats, which fundamentally drive structural variation through intra- and inter-molecular recombination [47]. We profiled these elements across the bipartite mitogenome of F. angustissima. SSRs (or microsatellites), defined as short tandem motifs of 1–6 bp, were extensively characterized. A total of 102 SSRs were identified, exhibiting an asymmetric distribution pattern: all 102 loci are exclusively localized to Chromosome 1 (Fig. 3A and Table S9). Tetrameric motifs accounted for the highest proportion of these SSRs (38.24%, 39 occurrences) and imply elevated rates of polymerase slippage during mitogenomic replication, providing a mechanistic basis for their utility as high-resolution polymorphic markers. Among the mononucleotide repeats, poly-T tracts maintained absolute dominance (61.29%). Consistent with this asymmetric distribution, a systematic survey identified an additional 23 larger tandem repeats, with unit lengths ranging from 11 to 41 bp, which are also strictly confined to Chromosome 1 (Table S10).

Fig. 3.

Distribution and classification of repetitive sequences within the bipartite F. angustissima mitogenome. A Frequency of SSRs. The bar chart details six SSR motif types, highlighting their exclusive localization on Chromosome 1. B Circular map illustrating the spatial distribution of repetitive elements. The inner intersecting arcs denote dispersed repeats, classified as palindromic (purple) or forward (green) repeats. The concentric tracks indicate the positions of tandem repeats (inner track, black lines) and SSRs (outer track, black lines)

Dispersed repeats, recognized as the primary catalysts for large-scale genomic rearrangements, were highly abundant. We detected 298 pairs of dispersed repeats (length ≥ 30 bp) across the mitogenome, comprising 124 palindromic and 174 forward repeats, with mean GC contents of 48.95% and 51.86%, respectively (Table S11). The maximum repeat unit reached 7,368 bp, corresponding to the primary palindromic repeat (R1) that mediates the previously described high-frequency structural isomerization (Table S5 and S12). Topological mapping of dispersed elements revealed a distinct structural asymmetry between the two chromosomes (Fig. 3B). Chromosome 1 displays a dense internal repeat network, underscoring its structural plasticity and the physical basis for recombination. Notably, numerous inner intersecting arcs demonstrate extensive inter-chromosomal linkages, established by both palindromic (purple) and forward (green) repeats.

Asymmetric intracellular DNA migration from plastid to mitochondrion

To investigate intracellular DNA migration, we de novo assembled the complete plastome of F. angustissima, which is 139,658 bp in length with an overall GC content of 38.87% (Fig. S3). Whole-sequence homologous alignments between the organellar genomes confirmed the frequent occurrence of MTPTs. We identified 27 MTPTs, of which 7 exceeded 1 kb. The maximum fragment length for a single transfer event reached 3,145 bp, which is strictly localized on Chromosome 2 (Table S13). Mapping these MTPTs against the reference plastome revealed a highly asymmetric distribution across the plastid structural domains (Fig. 4). The vast majority of these horizontal transfers originate in the Large Single Copy (LSC) region, making it an evolutionary hotspot for DNA migration and contributions to extensive insertions (e.g., MTPT27 (3,145 bp), MTPT13 (2,603 bp), and MTPT9 (1,680 bp)). In contrast, transfers originating from the Small Single Copy (SSC) region were comparatively rare, with MTPT14 (1,629 bp) as the sole representative, which harbors ndh genes. Notably, due to the structural symmetry of the plastid Inverted Repeat (IR) regions, MTPTs derived from these domains (e.g., MTPT1, MTPT11, and MTPT12) exhibit expected dual coordinate mapping across both the IRA and IRB domains.

Fig. 4.

Genomic visualization of MTPTs in the F. angustissima mitogenome. The outer tracks represent the target mitochondrial chromosomes (blue-grey arcs) and the source plastome (cpDNA; light green arc). Yellow interior ribbons indicate homologous DNA fragments transferred from the plastome to the bipartite mitogenome. MTPT, mitochondrial plastid DNA; cpDNA, chloroplast DNA

Functional annotation of these 27 MTPTs revealed 10 PCGs and 10 tRNA genes (Table S13). During this massive cross-organellar sequence migration, 14 intact plastid-derived genes were captured and integrated by the mitochondria, comprising six PCGs (atpE, ndhI, ndhJ, petB, psbH, and psbN) and eight tRNA genes (trnN-GUU-2X, complete trnV-UAC, trnM-CAU-2X, trnH-GUG-2X, trnF-GAA, trnW-CCA, trnC-GCA, and trnP-GGG). Additionally, the transfer analysis identified specific intergenic spacers (IGS), including IGS (trnL-CAA-trnI-CAU), IGS (trnI-CAU-trnL-CAA), and IGS (ndhC-trnV-UAC).

Prediction of RNA-editing sites and functional implications

Post-transcriptional RNA editing is a critical mechanism that compensates for sequence-level mutational drift, ensuring the correct translation and functional folding of mitochondrial proteins [48]. In-depth genome-wide prediction identified 488 high-confidence C-to-U RNA-editing sites across 35 unique PCGs (Fig. 5A and Table S14). Editing frequencies varied significantly across gene clusters. Among these, ccmC showed the highest frequency of editing events (36 sites), closely followed by ccmB (32 sites), ccmFN (32 sites), nad2 (32 sites), and nad7 (30 sites).

Fig. 5.

Predicted RNA-editing sites in PCGs of the F. angustissima mitogenome. A Distribution of predicted RNA editing sites across 35 identified PCGs. B Frequency of amino acid conversions induced by RNA editing. The x-axis indicates conversion types, including non-synonymous, synonymous, and premature stop codons (End)

Mechanistically, large numbers of C-to-U conversions exhibited a positional bias toward the first and second codon positions, driving non-synonymous amino acid alterations (Fig. 5B). The editing landscape was dominated by conversions of Ser to Leu (105 sites, 21.5%), Pro to Leu (100 sites, 20.5%), and Ser to Phe (69 sites, 14.1%) (Table S14). The combined Ser/Pro-to-Leu conversions account for 42.0% of all observed events. The targeted and systematic transition significantly enhances polypeptide hydrophobicity, an essential biochemical prerequisite for proper folding and efficient membrane anchoring of oxidative phosphorylation complexes [49]. Three distinct editing events generated obligatory translation termination codons: atp6 (CAA to UAA), atp9 (CGA to UGA), and ccmFC (CGA to UGA). These editing-mediated stop codons represent an indispensable post-transcriptional regulatory mechanism, dictating the precise truncation required for structurally viable and mature functional subunits.

Phylogenetic discordance and loss of mitogenomic collinearity

To elucidate evolutionary dynamics of the Fargesia lineage, maximum-likelihood (ML) trees were reconstructed from 32 conserved PCGs across 47 mitogenomes and 45 corresponding plastomes (Fig. 6). Both organellar phylogenies recapitulated the deep-node topologies of the five major Poaceae subfamilies (Pooideae, Bambusoideae, Oryzoideae, Panicoideae, and Chloridoideae), using species from Cyperaceae as outgroups. However, a significant inter-organellar phylogenetic discordance was observed within the Bambusoideae subclade. In the mitogenomic tree, F. angustissima exhibited strong topological conflict, clustering outside its genus and forming a well-supported sister clade (Bootstrap value = 87%) with Ferrocalamus rimosivaginus (Fig. 6A). Concurrently, its congener, F. qinlingensis, clustered into a distinct, distant clade with Pseudosasa japonica and Indocalamus species. Conversely, the plastomic phylogeny grouped F. angustissima tightly into a strictly monophyletic Fargesia clade (Bootstrap value = 90%), completely aligning with traditional morphological taxonomy (Fig. 6B).

Fig. 6.

Phylogenies resolving the evolutionary position of F. angustissima. A Tree inferred from the complete mitogenomes of 47 representative species. B Phylogenetic tree inferred from complete plastomes of 45 representative species. Numbers at the nodes indicate bootstrap support values based on 1,000 replicates using maximum-likelihood (ML) trees. Colored background blocks demarcate distinct taxonomic subfamilies within Poaceae and the outgroup (Cyperoideae)

To evaluate the structural conservation of the F. angustissima mitogenome, we performed a comparative synteny analysis against seven genetically related species, including six bamboos and one rice outgroup. The resulting collinearity plot visually uncovers structural polymorphism across bamboo mitogenomes. Although a substantial proportion of homologous sequences is densely retained between F. angustissima and its closely related congener F. qinlingensis (> 99.8% identity), the alignment is fragmented into dozens of distinct syntenic blocks (Fig. 7). The complex web of intersecting syntenic links in the plot shows that these conserved blocks have undergone spatial rearrangements, predominantly involving sequence inversions and translocations, as indicated by 407 syntenic blocks and micro-synteny mapping (Fig. S4). This profound loss of strict collinearity and complete disruption of gene order, even among closely related congeneric species, directly underscores the lack of physical conservation in bamboo mitogenomic architectures.

Fig. 7.

Mitogenomic synteny and structural rearrangements between F. angustissima and seven Poaceae species. Horizontal bars represent the mitochondrial genomes of the respective taxa. Gray ribbons denote homologous syntenic blocks in the same orientation, while pink ribbons indicate structural inversions. Collinear blocks shorter than 0.5 kb were excluded to minimize background noise

Discussion

Bipartite architecture and repeat-mediated structural evolution

Recent advancements in high-throughput sequencing have facilitated the precise assembly of complex plant mitogenomes, providing essential genomic resources for evolutionary, taxonomic, and conservation studies [50]. Beyond their canonical roles in bioenergetics and cellular metabolism, mitochondria function as central regulatory hubs orchestrating nutrient sensing, programmed cell death, and reactive oxygen species (ROS) homeostasis [49]. Unlike plastomes and animal mitogenomes, plant mitogenomes exhibit extraordinary structural complexity and conformational plasticity, primarily driven by frequent sequence rearrangements and genomic gain or loss events [51]. Although mitochondria are strictly maternally inherited in most angiosperms, occasional paternal leakage can induce heteroplasmy and mediate recombination between divergent mitogenomes [52]. While these active genomic dynamics drive immense structural diversity, the primary nucleotide sequences often experience relaxed selection and sequence-level drift. To ameliorate this, mitochondrial transcripts must undergo pervasive post-transcriptional modifications (e.g., RNA editing) to maintain functionality, though their precise physiological and adaptive significance often remains under-characterized [1, 48]. Understanding these intertwined structural and functional dynamics provides a crucial baseline for comparative genomics [3].

Within the subfamily Bambusoideae, Fargesia species are predominantly endemic to montane forests, where F. angustissima serves as an irreplaceable forage resource and habitat foundation for endangered sympatric wildlife, particularly the giant panda [53]. The intricate historical polyphyly of the Bambusoideae has long constrained the systematic characterization of mitochondrial evolution across this lineage [54]. In particular, the exact roles of structural variation in environmental stress acclimation, as well as the use of mitogenomic markers to resolve cryptic species classification and phylogenetic discordance, remain unresolved in F. angustissima. The complete mitogenome totals 429,279 bp in length (GC content: 44.18%) and contains 37 unique PCGs, reflecting a conserved coding capacity typical of bamboo lineages [22, 55]. Significantly, continuous long-read mapping explicitly resolved the mitogenome into a bipartite architecture comprising Chromosome 1 (412,081 bp) and Chromosome 2 (17,198 bp). Our structural analysis confirms that this complex multi-chromosomal state is affected by high-frequency homologous recombination. A R1 acts as a definitive recombination hotspot, maintaining a quantifiable in vivo near-equilibrium mixture of structural isomers. Moreover, the absolute absence of inter-chromosomal bridging reads between the two sub-genomes physically demonstrates that Chromosome 2 exists and replicates autonomously.

Consistent with the architectural complexity and size divergence pervasive across the plant kingdom, multi-chromosomal architectures and complex isomeric networks are frequently observed in diverse taxa [56–59]. Within the bamboo lineage, recent genomic studies have characterized similar multi-chromosomal or linear-circular combinations in related species, including F. qinlingensis and I. longiauritus [22, 60]. Aligning with these evolutionary footprints, our physical mapping data demonstrate that the F. angustissima mitogenome does not exist as a single static circle, but rather operates as a highly dynamic, multi-branched conformational network comprising two autonomous chromosomes. Ultimately, uncovering this intricate structural dynamism establishes an essential genetic framework for unraveling the complex evolutionary patterns of the Bambusoideae, facilitating the future exploration of montane climate adaptation and informing targeted conservation strategies.

Genomic conservation, mutational dynamics, and post-transcriptional constraints

Originating from a prolonged endosymbiotic process, plant mitochondria have lost substantial portions of their ancestral DNA through intracellular migration to the nucleus, retaining only critical functional elements [61]. While plant mitogenomes generally exhibit lower nucleotide substitution rates compared to their animal counterparts, they undergo extensive structural rearrangements driven by repetitive elements and exogenous sequence insertions [62, 63]. Despite this highly dynamic structural environment, our annotation of the F. angustissima mitogenome identified a highly conserved coding capacity of 37 unique PCGs. Compared with the 41 standard PCGs typically found in ancestral angiosperms, this retention suggests a stable evolutionary stringency of the core bioenergetic machinery within this bamboo lineage. This conservation starkly contrasts with the heavily truncated or highly fragmented mitogenomes observed in certain woody species such as Cinnamomum camphora, which exhibit extreme size expansion and gene loss [64].

At the molecular level, the transmission of genetic information in F. angustissima is profoundly shaped by directional mutation pressure. Because synonymous codon selection does not alter the resulting amino acid sequence, it improves translation accuracy and regulates the expression levels of specific proteins [65]. The identification of 30 preferred synonymous codons reveals a pronounced bias toward A/U-ending triplets, a pattern consistent with other Poaceae species such as Agrostis stolonifera [66]. Quantitative metrics (ENC-GC3 plot) indicate that the codon usage bias is driven by directional mutation pressure rather than translational selection. The continuous clustering of most PCGs along the expected distribution further corroborates mutation pressure as the dominant evolutionary force. These mutational signatures reflect relaxation constraints on the primary sequence, providing a theoretical framework for designing codon-optimization strategies to enhance mitochondrial protein expression in F. angustissima.

To physiologically rectify the sequence-level effects induced by this mutational pressure, RNA editing operates as an indispensable post-transcriptional checkpoint in F. angustissima. The 488 high-confidence C-to-U editing sites are comparable in frequency to those in model angiosperms, including Arabidopsis (441 sites) and rice [67, 68]. Interestingly, these editing events predominantly drive a systematic shift from hydrophilic to hydrophobic amino acid residues, most notably the conversion of Ser/Pro to Leu, which accounts for 42.0% of all observed events. The targeted biochemical transition is a strict prerequisite for the proper topological folding and inner-membrane anchoring of respiratory chain complexes [69]. Notably, ccm genes undergo hyper-efficient RNA editing, a conserved post-transcriptional signature mirroring that of the subalpine congener F. qinlingensis [22]. In native mid-elevation understories, chilling stress intrinsically impairs membrane fluidity and exacerbates respiratory electron leakage [1, 49]. By structurally securing cytochrome c maturation, this targeted modification functions as a crucial buffer to sustain oxidative phosphorylation and mitigate ROS accumulation, representing a key evolutionary strategy for environmental adaptation [15]. Besides, the high density of editing in nad4 signifies its role as a critical regulatory node in mitochondrial Complex I assembly [70]. By contrast, highly conserved genes (e.g., rpl2 and rps11) exhibited only a single RNA-editing site, indicating that strict sequence conservation was retained during mitogenome evolution [58, 71]. The reliance on specific editing events to generate obligate translation termination codons (e.g., atp6 and atp9) further highlights an evolutionary dependency on post-transcriptional repair in most plants [22, 72]. These phenomena underscore an evolutionary trade-off: while the F. angustissima mitogenome experiences relaxed selective pressures at the primary nucleotide level, RNA editing imposes a rigorous functional checkpoint to ensure respiratory efficacy and cellular homeostasis [73].

Inter-organellar discordance in bambusoideae and reticulate evolution

Complementing these localized mutational signatures, comparative collinearity analyses reveal a loss of strict synteny across the bamboo lineage. While F. angustissima conserved genomic blocks with closely related congeners, it lacks broad structural collinearity with other species across the Pooideae and Bambusoideae. As demonstrated in our structural analyses, this spatial variation and loss of collinearity are highly likely driven by frequent repeat-mediated homologous recombination, which continuously shapes evolutionary divergence, a mechanism similarly noted in other complex taxa [22, 55, 74]. The intense structural shuffling constitutes the physical foundation for the intricate diversification history of these lineages, which is mirrored by clear topological incongruences between the organellar phylogenies.

In contrast to the plastome phylogeny, the mitogenomic tree exhibits strong phylogenetic conflict, placing F. angustissima in a strongly supported sister clade with F. rimosivaginus, a taxon occupying a distant basal position in the plastome topology. This inter-organellar discordance extends across multiple groups: the mitochondrial tree supports the monophyly of Indocalamus, whereas the plastome tree renders it paraphyletic by nesting I. longiauritus as a sister group of F. angustissima. Similar topological variations between the organellar trees are also prevalent in the Pooideae subfamily (Triticum, Aegilops, and Agropyron), closely aligning with conflicts identified in other plant families (e.g., Rosaceae) where highly dynamic past hybridization events have obscured definitive phylogenetic relationships [7, 60].

The pronounced topological discordance observed between these obligately maternally inherited organellar genomes provides compelling genomic evidence that the speciation of temperate woody bamboos (Arundinarieae) was not purely bifurcating. Instead, this phenomenon can be primarily attributed to two classical mechanisms: historical reticulate evolution and ILS [75]. Recent nuclear phylogenomic frameworks for the Arundinarieae corroborate these dynamics and provide a critical reference to distinguish their respective contributions [76, 77]. Given that bamboos frequently undergo interspecific hybridization, ancient crossing events followed by continuous introgression likely facilitated organelle capture [76]. In this scenario, a descendant species essentially captures and fixes the organellar genome of a divergent lineage, directly generating the topological clashes observed in independent phylogenetic reconstructions. Concurrently, ILS serves as another critical driver of this inter-organellar discordance [77]. Cross-referencing these independent nuclear datasets clarifies the relative contributions of these mechanisms. While ILS broadly accounts for shallow, species-level incongruences within the Bambusoideae, the deep-node and trans-generic clustering of the F. angustissima mitogenome with Ferrocalamus strongly indicates ancient introgression and subsequent organelle capture.

The Fargesia genus has experienced a rapid evolutionary radiation across the montane ecosystems of the Hengduan Mountains and the western Sichuan Basin [78]. During these rapid speciation events within a compressed geological timeframe, abundant ancestral polymorphic alleles were randomly fixed across descendant taxa. Crucially, the inherently lower nucleotide substitution rate of plant mitogenomes compared to plastomes prolongs the retention of ancestral polymorphisms, rendering mitogenomes particularly susceptible to ILS [79, 80]. Previous studies have documented significant taxonomic difficulties in resolving species identification when relying solely on complete plastomes or nuclear ribosomal DNA sequences, primarily due to the long generation times and complex ancestral history of the Fargesia [18, 53]. Taken together, by uncovering the deep inter-organellar discordance driven by reticulate evolution and ILS, this study exposes the severe limitations of single-organelle phylogenies. It demonstrates the need to incorporate mitogenomic data to reconstruct the trajectories of montane bamboo lineages accurately.

Intracellular DNA migration and potential conservation markers

Beyond internal homologous recombination, the F. angustissima mitogenome is continuously reshaped by exogenous introgression via intracellular horizontal DNA transfer [81]. The MTPT fragments in the mitochondrial genome represent a pervasive evolutionary phenomenon across angiosperms [82]. In F. angustissima, the integration of 27 MTPTs (comprising 20.151 kb, or 4.48% of the mitogenome) underscores a substantial history of cross-organellar sequence capture. Detailed spatial mapping confirms that this migration is highly asymmetric. The overwhelming majority of these transfers (22 out of 27) originate from the LSC region, suggesting that the distinct bias establishes the LSC as an evolutionary hotspot, highly susceptible to structural instability and sequence excision. Conversely, the plastid IR and SSC regions exhibit extreme conservatism, contributing only three and two transfer events, respectively. This stark spatial contrast supports the canonical hypothesis that IRs function as strict structural anchors, effectively suppressing widespread genomic excision and maintaining overall plastome integrity [83].

The biased distribution of these MTPTs demonstrates that intracellular DNA transfer is not a stochastic spillover, but rather a non-random, region-specific dynamic rigorously filtered by post-transfer selection [84]. Over evolutionary time, integrated plastid sequences are frequently purged due to functional redundancy or energetic cost [85]. The limited integration of SSC-derived fragments, confined to specific ndh segments, implies stringent functional checkpoints that prevent arbitrary accumulation of sequences. The evolutionary trajectories of acquired plastid genes differ according to their functional classifications. Although captured PCGs (e.g., atpE, ndhJ, and psbH) maintain intact open reading frames, they lack mitochondrial promoters and essential cis-regulatory elements. Thus, under relaxed mitogenomic selective pressure, these sequences act as neutrally evolving relics destined for progressive pseudogenization [5]. By contrast, translocated tRNAs preserve self-sufficient secondary structures. The host translation machinery frequently recruits these functional transcripts to compensate for native tRNA degradation—a recognized molecular rescue mechanism [82]. While ongoing organellar evolution is frequently accompanied by mitochondrial structural gene loss across angiosperms, F. angustissima demonstrates a relatively conserved ancestral coding repertoire. For instance, whereas the essential rps11 has been sequentially lost in the closely related subalpine congener F. qinlingensis and broader monocot lineages, post-transfer MTPT degradation dynamics also exhibit marked specificity [22, 67]. The asymmetric accumulation of MTPTs and the differential retention of transferred genes demonstrate that the availability of source sequences and functional requirements constrains plastid integration in Fargesia.

The decoupling of mitogenome sequence stasis (> 99.8% identity) and structural fluidity between F. angustissima and F. qinlingensis renders phylogenetic markers ineffective for resolving Bambusoideae speciation. Consequently, high-resolution species delimitation was postulated to exploit the localized volatility at repeat boundaries, MTPT integrations, and microsatellite loci [61, 86]. Driven by polymerase slippage, the tetrameric loci (e.g., SSR nr. 19) function as specific diagnostic markers for delineating clonal boundaries (Table S15). Specific recombination breakpoints can act as structural variant (SV) fingerprints to differentiate cryptic taxa that possess identical nucleotide sequences [72, 87]. Although structural markers exhibit theoretical resolution, empirical validation is currently hindered by converging ecological limitations. In particular, the extensive field-sampling restriction is compounded by the severe scarcity of intraspecific mitogenomes in public repositories, which precludes robust in silico screening. Overcoming these dual constraints to confirm the practical use of these loci will thus require non-destructive and population-scale genetic surveys.

Conclusion

The de novo assembly of the F. angustissima mitogenome resolves a bipartite chromosomal architecture driven by high-frequency, repeat-mediated recombination. Functionally, extensive C-to-U RNA editing operates as a rigorous post-transcriptional buffering to compensate for relaxed sequence-level mutational pressure. Profound inter-organellar phylogenetic discordance provides direct genomic evidence for a reticulate evolutionary history shaped by ancient introgression and ILS. In summary, our study delivers molecular markers that could potentially overcome the low-nucleotide-diversity bottleneck in temperate bamboos, laying the groundwork for cryptic species identification, population genetic monitoring, and targeted habitat restoration of giant panda dietary resources.

Authors’ contributions

T.S., Y.D., and F.C. contributed to the study conception and design. Material preparation was performed by Y.D., L.Y., and H.H. Experiments and data analyses were performed by M.H., J.L., and C.W. Data collection and interpretation were performed by M.H., J.L., C.W., and L.Y. M.H., J.L., and C.W. wrote the first draft of the manuscript. All authors read and approved the final manuscript.

Funding

This research was financially supported by the National Key Technology Research and Development Program (Grant No. 2021YFD2200503), the National Natural Science Foundation of China (Grant No. 31870589 and Grant No. 31700525), and the Postgraduate Research & Practice Innovation Program of Jiangsu Province (Grant No. KYCX24_1383).

Data availability

All data supporting the findings of this study are available within the paper and its Supplementary Information. The assembled bipartite mitogenome and complete plastome sequences of F. angustissima have been deposited in the NCBI GenBank database under the accession numbers PZ188667 (Chromosome 1) and PZ188668 (Chromosome 2). The raw sequencing data have been submitted to the NCBI Sequence Read Archive (SRA) database under the accession numbers SRR38184827 (BGI short reads) and SRR38184828 (PacBio HiFi long reads).

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.