Codon usage bias and selective constraints in Gentianales mitogenomes

Abstract

Mitochondrial genome evolution(MGE) in flowering plants is quasi-intertwined-dynamic. MGE is driven via mutational pressures, translational selection, and functional constraints. However, unveiling the intra- and inter-genomic interplay governing evolutionary drives remains challenging. We investigate MGE-dynamic across twelve Gentianales species, revealing distinct codon usage patterns influenced by opposing evolutionary forces. While the first and second codon positions are highly conserved, the third codon positions show significant variability (27.7% – 45.8%), reflecting diverse selective pressures. Multi-dimensional analyses, including ENc-GC3s plots, neutrality plots, and PR2 bias, indicate that natural selection predominantly governs codon usage, outweighing mutational biases. Key findings include, non-significant correlations between GC12 and GC3 (R² ≤ 0.21), suggesting minimal mutational impact on genome composition; ENc-GC12 analysis showing codon optimization results from both selection and mutation; and PR2-plot analysis highlighting a preference for T- and G-ending codons, indicative of translational efficiency constraints. Gene-specific analyses of substitution rates (dN, dS, and dN/dS ) uncover heterogeneous selective landscapes, with genes such as atp, ccm, nad, and rps exhibiting signatures of positive selection. Substantial mutually offsetting dynamics between T3s and C3s (r = − 0.73), coupled with strong correlations between G3s and translational-efficiency indices (CAI: r = 0.69; CBI: r = 0.65), underscore that third-codon biases optimize translation. Evolutionary rates (dS and dN/dS ) show positive correlations with GC3 content (r = 0.45 and r = 0.33, respectively), indicating the influence of nucleotide composition on synonymous substitutions. Thus, these results reveal the interplay of the mutation–selection balance in non-recombining genomes and offer new perspectives on mitochondrial diversity in flowering plants.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12864-026-12583-4.

Introduction

Due to intra- and inter-genomic interplay, the mitochondrial genomes of angiosperms exhibit evolutionarily constrained, reflecting homeostasis between mutation-driven change and codon-level translational selection. This dynamic equilibrium underlies the selective retention of essential functional elements required for organellar and organismal homeostasis. For example, a striking paradox in evolutionary dynamics is that, among the widely studied chloroplast, nuclear, and mitochondrial genomes, the mitochondrial genome evolves at the slowest rate of sequence substitution [1]. The observed unparalleled structural fluidity, coupled with the relatively low synonymous substitution (dS) rates typically around 1.5–5 × 10⁻¹¹ substitutions per site per year is frequently less than one-sixth of the rate of nuclear DNA [2, 3]. Such robustness is associated with DNA repair efficiency and the strong functional constraints acting on core mitochondrial genes. Predominated by mutational bias, translational selection, and genetic drift, mitochondrial genomes are shaped by these forces, while genome rearrangements and gene deletions remain crucial contributors. Although insertions play a vital role in larger-scale changes, they are treated separately. Herein, we focus primarily on codon usage bias and selective constraints. Key challenges include disentangling the comparative contributions of slowly and rapidly evolving mitogenomes. Additionally, lineage-specific shifts in codon usage may reflect either neutral processes or adaptive optimization of mitochondrial function. Moreover, research has predominantly focused on model species and crops, leaving non-model plants particularly those from biodiversity hotspots underrepresented [4, 5].

Addressing gaps in our understanding of plant mitogenome evolution, adaptive codon usage, and phylogenetic relationships requires broader taxonomic sampling, standardized methodologies, and integrative analyses. Codon usage bias (CUB) in plant mitogenomes results from a complex interplay of mutational pressure, natural selection, and evolutionary constraints, with the pronounced AT-richness of these genomes being important feature [6, 7]. Unlike nuclear and chloroplast genomes, plant mitogenomes exhibit unique evolutionary characteristics, namely suppressed synonymous substitutions, widespread RNA editing, and frequent horizontal gene transfer. These factors collectively diminish CUB and promote the use of A/T-ending codons [8, 9]. A plant mitogenomes has confirmed low substitution rates compared to animal mitochondria [7–9], that is likely driven by efficient DNA repair mechanisms and strong purifying selection on oxidative phosphorylation genes. Such constraints may limit the adaptive potential of plant mitogenomes; for instance, the slow evolution of mitochondrial-encoded respiratory subunits could reduce metabolic plasticity in response to environmental stress. CUB-related translational inefficiencies may further impair fitness during rapid climate shifts. Understanding how lineage-specific variation in CUB (e.g., between crops and wild relatives) affects protein folding and energy metabolism remains a critical research priority. Overcoming challenges such as limited mitogenome sampling and structural complexity [10] is essential for revealing the role of mitochondrial genomes in plant adaptation under emerging selective pressures.

Advances in high-throughput sequencing and computational biology have enabled deeper insights into codon usage bias (CUB) in plant mitochondrial genomes, uncovering unique evolutionary patterns across angiosperms, gymnosperms, and early-diverging lineages [11]. Core respiratory genes (e.g., nad1, cox1) exhibit strong CUB at first codon positions, linked to translational efficiency and protein stability [12–14], while fern mitogenomes (e.g., Ophioglossum vulgatum L.) demonstrate correlations between CUB, expression levels, and functional constraints, particularly in ribosomal proteins like rps3 [15]. Adenosine triphosphate (ATP) synthase genes (e.g., atp6) show heterogeneous evolutionary rates reflecting lineage-specific selective pressures [15], yet critical gaps remain in understanding how CUB patterns correlate with gene function, especially in understudied clades like Gentianales where drivers of substitution rate variation and CUB remain unresolved.

To address these evolutionary pressures, resolving the paradox of mitochondrial genome evolutionary rates in Gentianales is imperative. Whether this lineage evolves more slowly or rapidly than other angiosperms remain unresolved, despite its substantial biotechnological implications. Investigating codon usage bias (CUB) and substitution rate variation offers a powerful lens through which to examine adaptive responses to abiotic stress and the underlying molecular evolutionary trajectories. This report links mitochondrial genome evolution to ecological adaptation and diversification in Gentianales. Analysis of CUB and substitution rates across selected dozen of species allows categorization of functional constraints and patterns of codon optimization associated with stress resilience. Thus, our results uncover the mutation selection balance in non-recombining genomes and offer a genomic foundation for crop improvement and the conservation of underexplored lineages.

Materials and methods

Sampling, mitochondrial DNA extraction, and genome sequencing

Fresh leaf samples of taxonomically identified Rubia cordifolia L. (Rubiaceae) were collected from Gulele Botanical Garden in Addis Ababa, Ethiopia, as part of a comprehensive study of mitogenomic diversity within Gentianales. Our analysis included 12 representative species spanning four key families (Rubiaceae, Gentianaceae, Apocynaceae, and Loganiaceae), incorporating 11 publicly available mitogenomes from GenBank alongside our novel sequencing of R. cordifolia to address a significant genomic gap in Rubiaceae. Genomic DNA was extracted using a commercial plant DNA extraction kit following the manufacturer’s protocol. For Illumina sequencing, we prepared libraries using the NEBNext Ultra II DNA Library Prep Kit (New England Biolabs) with ~ 350 bp insert sizes, followed by paired-end sequencing (2 × 150 bp) on an Illumina NovaSeq 6000 platform with dual-index adapters. Complementary long-read sequencing was performed using Oxford Nanopore Technology’s SQK-LSK109 kit on FLO-MIN106 flow cells by Benagen Tech Solutions Co., Ltd (Wuhan, China). The inclusion of this pharmacologically significant species provides crucial insights into structural variation and evolutionary patterns within Gentianales mitogenomes.

Mitogenome assembly and annotation

The Rubia cordifolia mitogenome was assembled from Illumina short-read data using GetOrganelle v1.7.5 with parameters -R 20 -k 21,45,65,85,105 -P 1,000,000 -F embplant-mt [16]The assembled reads were visualized and filtered in Bandage v0.80 [17] to generate the complete mitogenome. Protein-coding genes (PCGs) were annotated using GeSeq v2.03 [18] with Coffea arabica (LO788988) as a reference. We used the Find ORFs function in Geneious to confirm the protein-coding genes (PCG) annotations, while tRNA genes were identified using the tRNAscan-SE web service [19]. The newly assembled mitogenome was deposited in NCBI GenBank (accession: PX866621-PX866622). To determine mitochondrial synteny blocks among the 12 representative Gentianales species, all the mitogenomes were aligned using Mauve v2.3.1 [20] with default parameters, including automatic seed weight calculation (15) and a minimum local collinear block (LCB) score of 30,000.

Codon usage analysis

We scrutinized codon usage of 12 Gentianales species and 26 mitochondrial PCGs. During the gene extraction process in Phylosuite package [21], sequences < 300 bp were excluded [22], and s top codons were excluded from all analyses to improve the accuracy of the downstream analyses. To begin with, we calculated third-position nucleotides (A3s, G3s, C3s, and T3s), Codon Adaptation Index—CAI [23], Codon Bias Index—CBI [24], Frequency of Optimal Codons Fop [25], Effective Number of Codons—ENc [22], GC content at codon positions (GC1 ,GC2, and GC3), and Relative Synonymous Codon Usage—RSCU [26] using CodonW [27]. Codons with RSCU > 1 were considered high-frequency [28]. Next is the Neutrality analysis where the GC content was calculated using CodonW [27] for complete mitochondrial genes and individual codon positions (GC1, GC2, GC3). Using OriginPro, we generated 2D scatter plots with to examine correlations between the codon positions [15, 29]. An additional aspect involved the calculation of Effective Number of Codons ( ENc) which was performed in CodonW software and the Codon Usage Software Program (CUSP) program. GC_3S and ENc values were plotted against each other to investigate the influence of mitochondrial genome base composition on CUB. Simultaneously, a standard curve was constructed to illustrate the functional correlation between ENc and GC3s under mutational pressure [22], and the theoretical ENc values were computed using the formula . Finally, Parity Rule 2 (PR2) plot was used to explore CUB from mutation or selection and nucleotide compositional shifts in gene evolution in the mitogenomes of Gentianales. The PR2-plot analysis, where  horizontal(X- axis) and  vertical (Y-axis) coordinates [30, 31]. Base content (A = T, C = G) near the plot center (0.5, 0.5) indicates no codon bias. If the coding sequnce sites are close to the central point, mutation is thought to be the main cause of codon bias at the third position, while a divergence from the center supports natural selection.

Mitochondrial gene evolution: dN, dS, and dN/dS analyses

To study mitochondrial gene evolution via synonymous (dS), non-synonymous (dN), and the dN/dS ratio analyses, we used PhyloSuite software to extract conserved coding sequences from 12 representative Gentianales species (Table S2). Subsequently, the sequences were aligned with manual correction using MEGA’s “Muscle” module (which is codon-based) (Version 7.0) (Additional file 2) [32]. With Solanum aethiopicum L. (Solanaceae, NC 050335.1) as the outgroup, a maximum likelihood (ML) phylogenetic tree was built using the IQtree plugin (v1.6.8) (Additional file 1) [33]. In the codeml program of PAML (Version 4.9), dN, dS, and were computed using the F3 × 4 model [34, 35]. Subsequently, descriptive statistics and the Kruskal-Walli’s rank sum test were employed concurrently to evaluate the significance of group differences. Moreover, these results were visualized in boxplots using ggplot2 package in R software [36]. We conducted a correlation analysis to scrutinize the relationship between codon usage bias and evolutionary rates in mitochondrial genes across 12 representative Gentianales species. Our analysis mainly focused on various codon usage indices ( CAI, CBI, Fop, ENc, T3s, C3s, A3s, G3s, GC1, GC2, GC3) [22] and their correlation with dN, dS, and dN/dS values. Spearman’s rank correlation coefficients and P-values were calculated to assess the strength and statistical significance of these relationships [37]. The analysis was conducted using R software [36].

Results and discussion

Characteristic of Gentianales mitogenomes

A multi-metric comparative approach reveals significant structural divergence in mitochondrial genomes across the 12 representative Gentianales species (Table S1). Mitogenome sizes ranged from 175,011 bp in R. cordifolia to 741,811 bp in Hoya speciosa Decne, with GC content remaining relatively stable between 43.4% and 44.8%. The number of protein-coding genes varied from 27 in Uncaria rhynchophylla (Miq.) Miq. to 49 in Psychotria viridis Ruiz. & Pav., while exon counts ranged from 33 in Hoya lithophytica Kidyoo to 72 in Rhazya stricta Decne. Gene numbers varied from 24 in R. stricta to 83 in P. viridis. All species contained three rRNA genes, except U. rhynchophylla, which had six. The number of tRNA genes ranged from 12 in U. rhynchophylla to 30 in Gentiana straminea Maxim. These results highlight considerable structural variation in Gentianales mitogenomes, particularly in genome size, exon composition, and tRNA content, while GC content and rRNA gene numbers remained relatively conserved. These patterns underscore the structural fluidity and functional plasticity of Gentianales mitogenomes, reflecting the balance between genomic expansion, gene retention, and evolutionary constraints. Comparative mitogenomic analysis revealed strict conservation of core respiratory genes (atp1/4/8, ccmB/C/FC/FN, cox1/3, mttB, nad1/3/4/6) alongside dynamic evolution of accessory elements. Notably, photosynthesis-related genes (petA/B, psaA, psbA, rbcL) were exclusively retained in U. rhynchophylla, suggesting plastid-derived gene transfers. Additionally, lineage-specific losses of atp6, cox2, and nad5 highlight ongoing mitogenome streamlining (Figure S1). Furthermore, comparative synteny analysis revealed extensive mitochondrial genome rearrangements across the Gentianales species (Figure S2).

Codon bias and nucleotide flux across mitogenomes

Comparative mitogenomic analysis of 12 Gentianales species revealed both conserved and lineage-specific patterns. Most species exhibited the GC1 > GC2 > GC3 gradient, except I. jasminoides and U. rhynchophylla (GC2 < GC3), indicating species-specific selection (Fig. 1A). Overall GC content was conserved ( 44%), with U. rhynchophylla slightly higher (GC3 45.3%) and A. syriaca/R. stricta lower (< 43.6%). At the gene level, GC1 was consistently high ( 38–59%) and GC2 constrained ( 36–47%), whereas GC3 was highly variable ( 27–57%), particularly in nad and rps genes, reflecting adaptive dynamics. In contrast, mttB and sdh were more balanced. These results suggest that GC1 and GC2 are functionally constrained, while GC3 largely drives codon usage bias through mutation and selection (Tables S3, S4).

Fig. 1.

Variation in GC content and nucleotide frequency protein coding genes across 12 species of Gentianales. A, GC content at the first (GC1), second (GC2), and third (GC3) codon positions. B, Nucleotide composition at the third codon position, specifically T3s, A3s, G3s, and C3s, across the same species.The box represent the mean percentage of each nucleotide. Notable variation is observed in GC3 content and the dominance of T3s, signifying a potential codon usage bias linked to lineage-specific evolutionary pressures and adaptation strategies

The third codon positions of mitochondrial genes showed a consistent pattern across species, with most following the canonical order T3s > A3s > G3s > C3s (Fig. 1B). A clear A/T bias was observed, with T3s (0.21–0.54) and A3s (0.28–0.48) generally higher than C3s (0.2–0.4) and G3s (0.14–0.41). Gene-level analysis confirmed higher T3s/A3s ( 0.29–0.54) and lower C3s/G3s ( 0.11–0.37), indicating an overall A/T preference. GC3 was highly variable, particularly in nad and rps genes, whereas mttB and sdh were more balanced, reflecting lineage-specific codon usage and functional constraints (Tables S3, S4).

Evolution of codon usage in Gentianales mitochondrial genomes

A multi-metric codon usage analysis confirmed conserved translational efficiency across species (Fop: 0.36–0.39; CAI: 0.16–0.20; ENc: 57.3–59.25), with A. syriaca emerging as a notable outlier. Gene-level patterns indicated cox genes with strongest adaptation (CAI = 0.20), mttB with optimal usage (Fop = 0.39; ENc = 51.17), and matR exhibiting conflicting signals (high Fop = 0.42; ENc = 58.64) (Tables S3, S4). We found that U. rhynchophylla displayed the strongest codon bias, preferentially using UUU (Phe; RSCU = 1.83) over UUG (Leu; RSCU = 0.38), when stop codon usage varied between D. indicum (UGA preference; RSCU = 1.69) and Gentiana straminea (UAA preference; RSCU = 1.09) (Figure S3). The persistent underrepresentation of GC-ending codons suggests mutational bias or tRNA availability constraints, underscoring evolutionary trade-offs between selection and mutation in shaping mitochondrial codon usage, with potential implications for plant genome engineering.

Neutrality plot analysis of GC content signatures in Gentianales

We investigated the GC content fingerprint, defined by the relationship between GC3 and GC12, across twelve Gentianales species, revealing weak to moderate correlations (Fig. 2). The neutrality plots showed 0.01 ≤ R² ≤ 0.21, with varying statistical significance across species. Notably, R. cordifolia exhibited the strongest correlation (R² = 0.21, p < 0.05), potentially indicating selective pressures acting on GC content. In contrast, most species, such as D. indicus (R² = 0.01, p = 0.55) and A. syriaca (R² = 0.06, p = 0.23), showed no significant association, suggesting minimal mutational bias effects (Fig. 2, Table S5). Gene-specific variation was also evident, with atp and cox genes displaying distinct trends compared to nad and sdh, highlighting functional or structural constraints. Regression coefficient analysis (via t-test) confirmed no significant universal trend in base composition across codon positions. These findings suggest that natural selection predominantly shapes codon usage bias in Gentianales mitochondrial genomes, with mutation playing a secondary role, as shown in Fig. 2 and Table S5. The absence of a consistent GC correlation underscores the complexity of evolutionary forces acting on organelle genomes.

Fig. 2.

Correlation between GC3 and GC12 content across the twelve plant species of Gentianales. Each plot represents a single species, illustrating the correlation between GC3 content (x-axis) and GC12 content (y-axis) for the genes atp, ccm, cox, matR, mttB, nad, rps, and sdh. Trend lines represent the direction and strength of the correlation, with R²- and p-values provided to quantify the significance of each correlation. Data points are color-coded according to gene type, allowing for a visual distinction of gene-specific patterns

Mutation–selection balance through ENc–GC12 analysis

Analysis of codon usage reveals a spectrum of evolutionary strategies, reflecting different balances between mutational and selective forces (Fig. 3, Table S6). We identified three distinct regimes based on the correlation between GC content at the second codon position (GC12) and codon usage efficiency (ENc). In the first mutation-driven regime (P. viridis, T. jasminoides, R. stricta), a strong positive correlation (R = 0.60–0.65, p < 0.01) coupled with high GC12 (> 50%) and low ENc (< 45) indicates that mutational pressure is the primary architect of codon bias. A second, transitional regime (G. straminea, H. lithophytica) is characterized by a moderate correlation (R = 0.56–0.58), suggesting a balanced interplay between mutation and selection. Conversely, the selection-driven regime (S. hydrophyllacea, U. rhynchophylla, A. syriaca) shows a weak GC12-ENc dependence (R = 0.41–0.43), indicating natural selection is the dominant force overriding mutational biases. The species C. arabica (R = 0.14) emerges as a notable outlier, consistent with a neutral evolution model. Together, these species form a continuum from tightly coupled optimization (e.g., P. viridis, R = 0.91) to a largely decoupled state (e.g., C. arabica, R = 0.02). While the shared ENc distribution (40–60) suggests a common constraint, the underlying strategies diverge, with high-GC3 species potentially favoring mutational robustness and low-GC3 species prioritizing translational efficiency. These findings demonstrate the diverse ways evolutionary pressures shape mitogenome architecture within this important plant order.

Fig. 3.

Codon usage bias in Gentianales mitogenomes. A, ENc-GC3s plots for 12 representative species of Gentianales. Each scatter plot depicts the relationship between the ENc and GC content at the third codon position for individual genes. The dashed line represents the expected ENc under neutral conditions (i.e., influenced solely by mutation pressure). Deviations from this line suggest the influence of natural selection on codon usage bias. B, Pearson correlation coefficients (r) between ENc and GC3s for each species. Bars are color-coded according to panel A, and error bars represent the standard deviation. Species with higher R-values indicate stronger mutation bias in codon usage, whereas lower R-values suggest a greater role for selective constraints

PR2 plot analysis of selection-driven codon bias in Gentianales mitochondria

Figure 4 presents a schematic illustration of the PR2-plot analysis, which highlights the unequal usage of A/T and G/C at third codon positions across the sampled Gentianales mitogenomes. The substantial deviations of mitochondrial gene points from the plot’s center suggest strong selective pressures shaping codon usage bias (CUB), with variations observed across species and genes. Specifically, the data points predominantly cluster in the lower and lower-right quadrants, indicating a systematic preference for T- and G-ending codons over A- and C-ending ones. This bias is particularly evident in genes encoding oxidative phosphorylation (OXPHOS) subunits, such as cox1 and nad5, suggesting functional constraints tied to energy metabolism. Several outlier genes displayed extreme nucleotide preferences at third codon positions, with near-exclusive usage of T or G, which may reflect lineage-specific adaptations or specialized translational optimization (Fig. 4, Table S7). Overall, the observed T/G bias contrasts with the typical AT-rich composition of animal mitochondrial genomes, suggesting that selection, rather than mutation, plays a dominant role in shaping codon usage to preserve mitochondrial function.

Fig. 4.

Correlation between GC and AT biases in mitochondrial genes across different plant species. The scatter plots illustrate the relationship between GC bias G₃/ (G₃ + C₃) and AT bias A₃/ (A₃ + T₃) for selected mitochondrial genes across the studied species. Gene categories are color-coded, and species names are displayed above each plot. The colored quadrants in the background demarcate bias ranges for GC and AT content, providing a visual reference for bias intensity

Four species exhibited higher AT bias, skewing towards the right of the plot along the axis. These species A. syriaca, D. indicus, R. cordifolia, and G. straminea favor adenine at the third codon position (Fig. 4, Table S4). In contrast, five species, including C. arabica, P. viridis, S. hydrophyllacea, T. jasminoides,and U. rhynchophylla, exhibited lower AT bias, favoring thymine (Fig. 4, Table S7). A similar pattern was observed for GC bias, skewing towards the left of the plot along the axis. While D. indicus and R. stricta, along with two of the AT-biased species (A. syriaca and G. straminea), show a greater inclination toward guanine, the remaining species (C. arabica, P. viridis, S. hydrophyllacea, T. jasminoides, and U. rhynchophylla) tend to favor cytosine. These patterns reflect species-specific influences of mutational pressures or selection on mitochondrial gene evolution.

Adaptive dynamics of mitochondrial genes in Gentianales

Mitogenomic analysis uncovered distinct patterns of molecular evolution across functional gene categories. Analysis of nonsynonymous (dN) and synonymous (dS) substitution rates revealed varying selective pressures. We observed significant variation in both nonsynonymous (dN) and synonymous (dS) substitution rates. Elevated dN values in the nad gene of C. arabica (0.9) and the rps gene of D. indicus (0.88) suggested relaxed functional constraints, contrasting sharply with the extremely low dN in the nad gene of R. cordifolia (0.004), which indicated strong purifying selection (Fig. 5A). A similar pattern was seen for dS rates, which were elevated in nad (D. indicus, 2.41) and ccm (G. straminea, 2.3), suggesting relaxed constraints, whereas lower dS values in nad and rps genes of C. arabica (0.14–0.27) pointed to ongoing purifying selection (Fig. 5B). Formal selection analyses using the dN/dS ratio provided a clear confirmation of these dynamics, identifying positive selection in atp (U. rhynchophylla, dN/dS ≈ 1.45) and nad (H. lithophytica, dN/dS = 1.42), and intense purifying selection in mttB (G. straminea, dN/dS = 0.01) (Fig. 5C).

Fig. 5.

Comparative analysis of evolutionary rates across mitochondrial gene groups in diverse land plants. A, Boxplots illustrate the distribution of non-synonymous substitution rates (dN). B, synonymous substitution rates (dN). C, dN/dS ratios across twelve representative land plant species. Each dot represents an individual gene, color-coded by functional category: atp (ATP synthase), ccm (cytochrome c maturation), cox (cytochrome c oxidase), matR (maturase R), mttB (membrane targeting and translocation protein B), nad (NADH dehydrogenase), rps (ribosomal proteins), and sdh (succinate dehydrogenase). The horizontal line within each box indicates the median, with whiskers extending to 1.5× the interquartile range. Variation in evolutionary rates among gene groups and species reflects differential functional constraints and selective pressures acting on mitochondrial protein-coding genes.

A significant variability with evolutionary pressures across mitochondrial genes observed. The nucleotide substitution rates (dN, dS, and dN/dS ) across eight mitochondrial genes evaluated. The ccm gene exhibited the highest nonsynonymous substitution rate (dN = 0.42), while mttB showed the lowest (dN = 0.33) (Table S8). For synonymous substitution rates (dS), mttB had the highest mean (dS = 1.15), and atp the lowest (dS = 0.29) (Table S8). The ccm gene also displayed the highest dN/dS ratio (0.4), indicating stronger purifying selection, while mttB had the lowest dN/dS ratio (0.22), reflecting weaker selective pressure. Other genes exhibited intermediate values for both dN/dS ratios and substitution rates.

Evolutionary shaping of codon bias

A comprehensive correlation analysis of studied Gentianales mitogenomes shows unique evolutionary trends govern via interplay among nucleotide composition, codon usage bias (CUB), and selection pressures. Within nucleotide composition, T3s and C3s are strongly negatively correlated (r = − 0.73), while G3s shows positive associations with CUB indices (CAI, CBI, Fop; r = 0.65–0.70), indicating that GC-rich codons contribute to stronger codon bias. CUB measures themselves are highly intercorrelated (r = 0.85–0.95) but negatively related to ENc (r = − 0.38 to − 0.45), consistent with ENc’s inverse relationship to bias strength. GC content, particularly GC3s and GC3 (r = 1.00), strongly modulates CUB, with GC3s correlating positively with G3s (r = 0.62) and negatively with ENc (r = − 0.47). Selection parameters (dN. dS, dN/dS) exhibit weaker correlations with nucleotide composition and CUB, suggesting that evolutionary rates are partly independent of compositional biases. Notably, dS and show modest positive correlations with GC3 (r = 0.45 and r = 0.33, respectively), indicating that synonymous substitution rates may be modulated by local GC content. Overall, these correlation patterns highlight the complex interplay between nucleotide composition, codon preference, GC content, and selection in shaping mitogenome evolution in Gentianales (Fig. 6).

Fig. 6.

Correlation matrix of nucleotide composition, codon usage bias, GC content, and selection pressure indicators. The heatmap displays pairwise Pearson correlation coefficients among 20 variables, organized into four categories: nucleotide composition (T3s, C3s, A3s, and G3s), codon usage bias indices (CAI, CBI, Fop, and ENc), GC content metrics (GC1, GC2, GC3, GC3_S, overall GC), and selection pressure indicators (dN, dS, dN/dS). The color scale represents the magnitude and direction of the correlations, with red tones indicating strong positive correlations, blue tones denoting strong negative correlations, and white cells representing non-significant correlations (p > 0.05). Colored bars beneath the matrix categorize the variables: red for nucleotide composition, blue for codon usage bias, green for GC content, and orange for selection pressure.

Discussion

Mitochondrial genomes of Gentianales, including economically and medicinally important species, remain limited; sequencing R. cordifolia fills a key research gap in Rubiaceae. Comparative analysis of selected Gentianales mitogenomes provides important insights into their evolution shaped by mutational biases and functional constraints. Most species maintain the expected GC gradient (GC1 > GC2 > GC3) [15], notable exceptions are I. jasminoides and U. rhynchophylla (GC2 < GC3), indicating lineage-specific evolutionary pressures, such as differential DNA repair [38] or GC-biased gene conversion [39]. At the codon level, species diverge into A3s and T3s dominant clusters (e.g., 28.38% A3s in U. rhynchophylla vs. 28.25% T3s in R. stricta) [40] with P. viridis emerging as an outlier through elevated G3s (22.84%), potentially indicating horizontal transfer [41]. Functional analysis shows respiratory genes maintain strong T3s bias (0.456–0.460) and suppressed G3s (0.170–0.185) for structural optimization [42, 43], ribosomal proteins favor A3s (0.438) for translational efficiency [23], whereas the matR intron shows exceptional C3s enrichment (0.391) for RNA structural requirements [44]. These trend demonstrate how mitochondrial genomes balance universal functional constraints with lineage-specific adaptation [45, 46], underscoring the need for further studies on tRNA coevolution and structural impacts of codon bias.

Accordingly, an analysis of selected Gentianales mitogenomes reveals unique evolutionary patterns shaping mitochondrial codon usage. Strong purifying selection was observed to maintain translational efficiency across most species (Fop: 0.3589–0.389; CAI: 0.157–0.170; ENc: 57.3–59.25) [47], although A. syriaca emerged as an outlier, suggesting a potential relaxation of selective pressures in this lineage [48]. Gene-level analyses demonstrate functional specialization, with respiratory genes (cox) showing the strongest translational optimization (CAI = 0.201) [49] while the matR intron exhibits conflicting selection signals (Fop = 0.415; ENc = 58.64) [44], reflecting competing demands between RNA structure and protein function. Species-specific patterns, including Uncaria rhynchophylla’s pronounced UUU preference (RSCU = 1.83) and divergent stop codon usage, may result from coevolution with tRNA pools [50]. or adaptation to physiological demands [51]. The systematic underrepresentation of GC-ending codons supports AT-biased mutation pressure in plant mitochondria [52]. However, tRNA availability could contribute to these patterns [40]. These findings provide fundamental insights into mitogenome evolution and establish a framework for studying mitonuclear coevolution, with important implications for mitochondrial genome engineering [52].

Our multi-feature analysis of GC content fingerprints across representative Gentianales mitogenomes species reveals distinct patterns of codon position correlations, with R² values ranging from 0.01 to 0.21 (Fig. 3). While R. cordifolia showed the strongest GC3-GC12 correlation (R² = 0.21, p < 0.05), suggesting potential selective pressures [29], most species exhibited non-significant associations (D. indicus: R² = 0.01, p = 0.55; A. syriaca: R² = 0.06, p = 0.23), indicating minimal mutational bias effects [53]. Gene-specific patterns emerged, with respiratory genes (atp, cox) displaying distinct GC trends compared to metabolic genes (nad, sdh), likely reflecting functional constraints [54]. Regression coefficient analysis confirmed no universal base composition trend across codon positions (p > 0.05), supporting natural selection as the primary driver of codon usage bias in these mitogenomes [55]. These results underline the complex interplay of evolutionary forces shaping organelle genomes [56], with implications for understanding mitochondrial genome evolution in flowering plants [57].

Deterministic dynamics of codon usage in Gentianales mitogenomes reveal distinct selective pressures across species. Our analysis of selected Gentianales mitogenome identified three evolutionary strategies: Mutation-dominated species (P. viridis, T. jasminoides,and R. stricta) showed strong GC12-ENc coupling (r = 0.60–0.65, p < 0.01) and high GC12 (> 50%), indicating mutational pressure as the primary driver [15]. Transitional species (G. straminea, H. lithophytica) exhibited intermediate patterns (r = 0.56–0.58), suggesting balanced mutational and selective pressures [58]. Selection-driven species (S. hydrophyllacea, U. rhynchophylla, and A. syriaca) displayed weak GC12 dependence (r = 0.41–0.43), with C. arabica (r = 0.14) representing a potential neutral evolution outlier [59]. The triangular ENc distribution (40–60) reflects fundamental constraints on mitogenome architecture [60], demonstrating how distinct evolutionary pressures shape organelle genomes in this ecologically diverse plant order. High-GC3 species appear to favor mutational robustness, while low-GC3 species optimize translational efficiency [60], revealing a continuum of evolutionary strategies from tightly constrained (R² = 0.91 for P. viridis) to effectively uncoupled (R² = 0.02 for C. arabica) optimization patterns [15].

In Fig. 4, the PR2-plot analysis illustrates a significant deviation from neutral expectations in Gentianales mitogenomes [61].An asymmetric nucleotide composition was observed at the third codon positions (A ≠ T, G ≠ C), with systematic clustering in the lower and lower-right quadrants, indicating a strong preference for T- and G-ending codons. This bias is particularly marked in OXPHOS genes (cox1, nad5) [62], signifying functional constraints on energy metabolism [63], while extreme outliers with near-exclusive T/G usage likely reflect lineage-specific translational optimization [64].The observed T/G bias contrasts sharply with animal mitochondrial AT-richness [65], demonstrating that selection dominates over mutational pressure [66] in shaping plant organelle codon usage, with deviations from parity rules [67] confirming strong selective constraints [49].

Mitogenomic analysis revealed a heterogeneous landscape of molecular evolution across functional gene categories, driven by a complex interplay of selective pressures as evidenced by significant variation in substitution rates (dN, dS, and dN/dS); while strong purifying selection was observed in genes like nad in R. cordifolia (dN = 0.004) and mttB in G. straminea (dN/dS= 0.01), signatures of positive selection were detected in atp (dN/dS≈ 1.45) and nad (dN/dS= 1.42), indicating a dynamic balance between constraint and adaptation [68–70], and comparative analyses showed that genes like ccm experienced stronger purifying selection (higher dN/dS) compared to mttB, highlighting that the mitochondrial genome is not a static entity but a mosaic of evolutionary rates where essential functions are maintained under strong constraint [71, 72]. while other loci exhibit greater tolerance for change or undergo adaptive evolution as part of the broader trend of organellar genome evolution [63] .

A multi-indicator correlation analysis of representative Gentianales mitogenomes exhibits fundamental evolutionary dynamics shaped by the relationship of mutational biases and selection pressures, as shown in Fig. 6. We identified: (1) a strong T3s-C3s anticorrelation (r = -0.73, p < 0.001) indicative of thymine-cytosine mutational bias at synonymous sites [72]; (2) striking G3s associations with codon adaptation indices (CAI: r = 0.69; CBI: r = 0.65), demonstrating selection for G-ending codons in highly expressed genes [73]; and (3) a positive correlation between dS and dN/dS(r = 0.45 and r = 0.33, respectively) with GC3 content, revealing pervasive purifying selection, with GC%-dS covariation suggesting influences of GC-biased gene conversion (Fig. 6) [74]. These patterns resolve long-standing debates on the relative contributions of mutation and selection in plant mitogenomes by demonstrating that their evolution follows predictable principles: neutral processes create baseline variation, purifying selection constrains essential functions, and translational efficiency shapes codon usage, with lineage-specific constraints modulating these universal patterns [49, 65]. Our results clarify key aspects of organelle–nuclear coevolution [68] and the mutation–selection balance in non-recombining genomes [75], yielding new insights into the origins of mitochondrial genomic diversity.

Conclusion

In summary, our work uncovers wide-ranging structural and compositional heterogeneity in Gentianales mitochondrial genomes despite conserved core functions, revealing more than fourfold size variation, pervasive rearrangements, and lineage-specific gene losses. Codon-usage patterns follow a conserved GC1 > GC2 > GC3 hierarchy, with GC3 driving lineage-specific biases and translational optimization. Neutrality, ENc–GC12, PR2, and rate-based analyses collectively show that natural selection rather than mutation alone governs codon usage, generating distinct mutation–selection regimes across species. However, this study is limited by the relatively limited sampling, which may underrepresent the full breadth of mitochondrial diversity and codon-usage dynamics across Gentianales. These patterns link CUB to ecological pressures, including xeric adaptation via GC3 enrichment, which supports translational fidelity and reduces mutational load under stress. High CAI values, exemplified by A. syriaca, act as bioindicators of oxidative stress, reflecting strong selective pressure to maintain accurate translation. Our tripartite evolutionary model illustrates how non-recombining genomes balance neutrality and selection, providing tools for conservation genomics to identify lineages vulnerable to mutational meltdown and for synthetic biology to optimize transgene expression. We recommend future studies to incorporate broader phylogenetic (including population-level) sampling to perform functional validation which will be essential to refine these evolutionary inferences and test their generality in angiosperms. Collectively, these insights illuminate a dynamically evolving yet functionally constrained mitochondrial landscape, offering broad implications for angiosperm evolution, ecological adaptation, and conservation.

Authors’ contributions

Professor Wan T. and Dr. Sara G. conceived the idea. Professor supervised the project. Sara G., Liao Y., and Girma E. performed the experiments. Professor Wan T. and Dr. Sara G. Wrote the manuscript. Girma E., Ann W., Oyebanji O., Liao Y., Wan T., and Sara G. analyzed the data. All the authors discussed the results, commented, and revised them.

Funding

This work was financially supported by, the National Natural Science Foundation of China (584763SN:38173276).

Data availability

The datasets generated and/or analyzed during this study are available in GenBank (https:/www.ncbi.nlm.nih.gov/genbank), with accession numbers in Supplementary Table S1.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.