Plastome evolution at the edge: structural rearrangements, IR expansion, and gene flux in Hypericaceae (Malpighiales)

Abstract

Background

The evolutionary history of the Hypericaceae Juss. family remains poorly understood despite previous phylogenomic efforts. A prior study on Hypericum ascyron revealed exceptional plastome rearrangements and gene loss events, prompting questions about whether such genomic patterns are unique to Hypericum or reflect broader evolutionary trends within the family.

Results

To explore plastome evolution across Hypericaceae, we sequenced 12 complete chloroplast genomes representing seven genera from the three major tribes, Hypericeae, Vismieae, and Cratoxyleae, and two outgroup plastomes from Clusiaceae. Comparative analysis of 281 Malpighiales plastomes showed that Hypericaceae differ significantly in total plastome size and SSC region length, while LSC and IR regions showed no significant differences.. Plastome sizes in Hypericaceae ranged from 138 to 176 kb, reflecting extensive structural variation, multiple inversions, IR expansions, and lineage-specific rearrangements. Cratoxyleae exhibited a relatively conserved plastome structure with one major inversion and minimal gene loss, representing the most stable lineage within the family. Species within Vismieae showed markedly expanded IR regions, while Hypericeae exhibited frequent gene and intron losses associated with structural instability. In Hypericum, genes such as matK and accD were relocated into or near the IR regions, accompanied by lineage-specific ORFs likely formed through repeat-mediated recombination. Several genes (rpl23, rpl32, rps7, rps16, infA, ycf1, ycf2) showed independent losses across the family. Across the Hypericaceae, the protein-coding genes matK, accD and clpP also showed domain-disrupting expansions, potentially impacting their functional roles.

Conclusions

Our results demonstrate that plastome evolution in Hypericaceae is highly dynamic, characterized by substantial structural plasticity, gene loss, and lineage-specific innovation. These findings provide new insights into plastome diversification across the family and lay the groundwork for further phylogenomic and evolutionary studies within Malpighiales.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12870-025-07888-7.

Introduction

The order Malpighiales Juss ex Bercht & J. Presl, one of the most taxonomically and ecologically diverse clades of flowering plants, was first resolved through molecular phylogenetic studies by Chase et al. [6]. This reclassification reshaped the understanding of angiosperm relationships and highlighted the potential of molecular tools in resolving deep evolutionary lineages [6, 59, 62]. Among its major subclades, the clusioid clade comprising five families: Bonnetiaceae L. Beauvis Ex Nakai, Calophyllaceae J.Agardh, Clusiaceae Lindl., Hypericaceae Juss, and Podostemaceae Rich., represents a morphologically and biogeographically complex lineage with over 1,900 species spread across 94 genera [56, 63, 72]. Members of this clade are predominantly tropical, though certain genera extend into temperate and high-altitude regions, often adapting to extreme or specialized ecological niches such as montane forests and fast-flowing aquatic systems [10, 57]. Within the clusioid clade, Hypericaceae stands out due to its taxonomic complexity and unevenly studied genera. Historically treated as a subfamily within Clusiaceae, Hypericaceae was elevated to family rank based on distinct morphological and molecular features [24]. It currently comprises approximately 518–700 species across nine genera and is taxonomically divided into three tribes: Hypericeae Choisy, Vismieae Vandelli, and Cratoxyleae Engl. APG III, 2009 [8]. The genus Hypericum, the most species-rich lineage with a cosmopolitan distribution, accounts for nearly 80% of the family’s diversity. In contrast, the other genera, many of which are confined to tropical regions, have remained poorly sampled in molecular studies, leaving tribal and intergeneric relationships largely unresolved [36, 64].

Previous phylogenetic efforts in Hypericaceae have focused heavily on Hypericum, employing morphological data [46–54], ITS markers [34, 42], and plastid regions (e.g., trnL-trnF, psbA-trnH). These studies provided key insights into biogeographic patterns, such as the disjunction between New World and Old World clades, and proposed evolutionary scenarios involving ancient Gondwanan dispersal and multiple niche shifts [29, 30, 37]. However, these single- or low-copy gene datasets offered limited resolution at deeper nodes and often failed to resolve intertribal relationships, especially for Vismieae and Cratoxyleae.

Plastid genomes (plastomes) have emerged as powerful tools for resolving angiosperm relationships at both deep and shallow phylogenetic levels [20]. Their quadripartite structure, gene content, and substitution dynamics offer evolutionary signals that are often inaccessible through nuclear markers alone. Furthermore, increasing evidence suggests that plastome structural variation, such as inversions, IR boundary shifts, and gene loss, is not just phylogenetically informative but may also reflect lineage-specific ecological adaptations and genome evolution under selection [18, 69]. In this context, H. ascyron. was recently shown to possess one of the most highly rearranged plastomes among angiosperms, with large inversions, pseudogenization, and accelerated substitution rates [7]. Whether this pattern is unique to H. ascyron or indicative of a broader trend across Hypericaceae remains unclear.

To address this question and improve resolution of phylogenetic relationships within Hypericaceae, we sequenced and analyzed complete chloroplast genomes from 14 species across all three tribes: Hypericeae (including Hypericum, Triadenum Raf., and Thornea Breedlove & E.M.McClint), Vismieae (Harungana Lam., Vismia Vand.), and Cratoxyleae (Cratoxylum Blume., Eliea (Lam.) Cambess.). Two species from Clusiaceae were included as outgroups. By combining comparative plastome analysis with phylogenomic reconstruction based on 79 plastid protein-coding genes, we aimed to assess the extent and pattern of structural plastome variation across the family, Investigate gene loss, pseudogenization, and possible functional relocation events and, resolve tribal and infrageneric relationships in light of previous nuclear and morphological findings.

Our results provide new insight into plastome evolution in Hypericaceae and reveal lineage-specific trends in genome reorganization that align with ecological transitions and biogeographic structuring. These findings not only refine the phylogenetic backbone of the family but also shed light on how structural plastome dynamics may correlate with environmental adaptation and diversification.

Results

Variation in plastome structure and size across hypericaceae tribes

Plastome sizes in the tribe Hypericeae ranged from 138,163 bp in Hypericum patulum to 167,482 bp in Triadenum fauriei (Table 1). The large single-copy (LSC) region exhibited substantial variation, spanning from 103,150 bp in H. patulum to 85,948 bp in T. calcicola. The inverted repeat (IR) region reached its maximum length in H. laxum (32,823 bp) and its minimum in H. patulum (12,032 bp). The small single-copy (SSC) region showed a narrower range, from 10,423 bp in H. laxum to 12,421 bp in T. calcicola. In the Vismieae tribe, plastome sizes were generally larger, ranging from 166,590 bp in Vismia baccifera to 176,170 bp in Harungana madagascariensis (Fig. 1). The LSC region measured 67,585 bp in V. baccifera and 57,619 bp in Harungana madagascariensis, while the IR region was largest in Harungana madagascariensis (52,124 bp) and smallest in V. baccifera (42,583 bp). The SSC region in both species remained within the size range observed in Hypericeae. For the Cratoxyleae tribe, plastome sizes were more conserved, with a median length of approximately 156–157 kb. The LSC region measured 85,588 bp in Eliea articulata and 86,024 bp in Cratoxylum maingayi. The SSC regions ranged from 17 to 18 kb, and the IR regions ranged from 25 to 26 kb. These values are comparable to those found in the outgroup plastomes of Moronobea riparia and Rheedia edulis. Across all three tribes, plastome sizes varied from 138,163 bp (H. patulum) to 176,170 bp (Harungana madagascariensis), reflecting both lineage-specific expansions and reductions in chloroplast genome architecture (Table 1).

Table 1.

Summary of plastome features for 13 Hypericaceae and two Clusiaceae species used in this study for phylogenetic analysis. The table includes total plastome size, lengths of the large single- copy (LSC), inverted repeat (IR), and small single-copy (SSC) regions, GC content, and the number of protein-coding sequences (CDS) and tRNA genes

Family	Species	Total	LSC	IR	SSC	GC %	CDS	tRNA
Hypericaceae	Hypericum ascyron	162,286	97,542	26,846	11,052	37.40%	74	34
	Hypericum erectum	143,167	99,272	16,370	11,155	37.10%	72	34
	Hypericum chejuense	142,029	98,131	16,375	11,148	37.20%	73	34
	Hypericum patulum	138,163	103,150	12,032	10,949	38.10%	72	34
	Hypericum laxum	165,288	89,219	32,823	10,423	38.10%	73	34
	Hypericum japonicum	165,190	89,123	32,805	10,457	32.00%	75	34
	Triadenum fauriei	167,482	92,077	31,916	11,191	37.40%	75	34
	Thornea calcicola	150,935	85,948	26,283	12,421	37.90%	75	34
	Harungana madagascariensis	176,170	57,619	52,124	14,303	36.90%	75	35
	Harungana rubescens	174,661	57,746	51,809	13,297	37.10%	75	35
	Vismia baccifera	166,590	67,585	42,583	13,839	36.90%	75	35
	Eliea articulata	156,123	85,588	25,640	19,254	36.40%	77	35
	Cratoxylum maingayi	157,279	86,024	26,155	18,945	36.30%	77	35
Clusiaceae	Moronobea riparia	156,826	84,914	27,057	17,798	36.00%	77	35
	Rheedia edulis	158,160	86,384	27,010	17,756	35.80%	78	35

Fig. 1.

Chloroplast genome map of Vismia and Harungana (Vismieae) showing an expanded inverted repeat (IR). The circular plastome structure of Vismieae tribe displays a notable expansion of the IR region to ~ 50 kb, leading to extensive duplication of genes beyond the typical IR boundary (duplicated genes shown with grey dot). Genes are color-coded by functional category, with arrows indicating transcriptional direction. LSC, SSC, and IR regions are labeled, and the inner grey circle indicates GC content. The red arrow marks the direction of IR expansion toward the LSC, while the blue arrow denotes contraction toward the SSC

Plastome size variation in hypericaceae and related malpighiales lineages

To evaluate patterns of plastome size evolution within Hypericaceae and across the broader order Malpighiales, we compared plastome sizes among 281 species, including representatives from both focal and outgroup families (Fig. 2). The distribution of plastome sizes revealed substantial inter- and intra-lineage variability. Species within.

Fig. 2.

Comparison of plastome sizes across 281 species from Hypericaceae (red) and other families within Malpighiales (blue). Each panel represents a distinct plastome region (Total, LSC, SSC, IR). Boxplots summarize length distributions with mean values (black diamonds). Statistical significance between Hypericaceae and other Malpighiales was evaluated using Wilcoxon rank-sum tests. Black diamond indicates the Mean (average) of plastome length for that family

Hypericum, including H. ascyron, H. laxum, and H. patulum, clustered within a relatively narrow and conserved size range. The genera Triadenum and Eliea, also members of Hypericaceae, showed a similar pattern. In contrast, taxa from the Vismieae tribe exhibited significantly larger plastome sizes due to IR expansion and contraction of SSC, appearing as outliers relative to the general distribution of Hypericaceae and most Malpighiales families. These enlarged plastomes may reflect lineage-specific structural expansions or accumulation of non-coding sequences. At the lower end of the spectrum, H. chejuense and H. erectum displayed comparatively reduced plastome sizes within Hypericaceae, suggesting possible genome contraction events. Comparative analyses of plastome size and genomic regions between Malpighiales and Hypericaceae were performed, and Wilcoxon rank-sum tests were applied to evaluate pairwise differences between the two groups. Significant differences were detected for Total plastome length (p = 0.0023) and SSC region (p = 0.0138), while LSC and IR did not differ significantly (p > 0.05) (Fig. 2). Plastome size comparisons showed that Hypericaceae and related families occupy a distinct range relative to other Malpighiales.

Distribution of tandem repeats across hypericaceae plastomes

The comparative analysis of tandem repeat content revealed notable variability across Hypericaceae plastomes and outgroup species from Podostemaceae, Calophyllaceae, and Clusiaceae (Fig. 3). Species within Hypericaceae, particularly T. fauriei, H. ascyron, and H. hookerianum, exhibited a markedly higher number of tandem repeats, with over 90 repeats detected in some cases. These were predominantly within the 100–199 bp and 200–499 bp length classes. In contrast, outgroup taxa such as M. riparia and R. edulis displayed relatively low repeat abundance, primarily in the 50–199 bp range. Notably, several Hypericaceae species harbored long repeats (≥ 500 bp), which were either rare or absent in outgroup plastomes.

Fig. 3.

Distribution of tandem repeats in plastomes of Hypericaceae species compared to outgroup species from Podostemaceae, Calophyllaceae, and Clusiaceae. The bar plot shows the number of tandem repeats grouped by repeat length categories. Colors indicate different repeat size ranges, with Hypericaceae species generally exhibiting a higher abundance and broader size range of repeats than outgroups

Structural evolution of the hypericaceae plastome

Hypericeae tribe

Comparison of plastomes from the Hypericeae tribe with M. ferrea used as a reference for the ancestral plastome structure revealed dynamic structural rearrangements, including multiple inversions, segmental rearrangements, and inverted repeat (IR) boundary shifts (Supplementary Fig. 1). Mauve alignment analysis identified 25 locally collinear blocks (LCBs), suggesting at least seven inversion events involving 21 breakpoints, including: trnH–psbA, psbA–trnK, rps16–psbL, atpA–rps2, rpoC2–rpoC1, rpoC1–rpoB, petN–psbM, rps7–psbD, psbD–ndhC, ndhC–ndhJ, atpE–rbcL, accD, psaI–petA, clpP–psbJ, psbB–rps19, rpl2, petG, ndhG, ndhF, rpl32–rps15, and ycf1. In total, we identified ten separate inversion events across the Hypericeae plastomes within LCB regions 2–11 and 17 (Supplementary Fig. 1). Specifically, in H. ascyron exhibited eight inversions, H. chejuense and H. erectum showed eleven inversions, T. calcicola displayed ten inversions, H. patulum showed seven inversions, and H. laxum and T. fauriei had six inversions each. The largest inversion identified across Hypericeae was LCB 8 (~ 14 kb), present in all species, and LCB 18 (~ 15 kb) in H. ascyron, H. chejuense, and H. patulum (Supplementary Fig. 4). Notably, one inversion located in the IR region was associated with IR expansion. A prominent structural rearrangement involved the relocation of the matK gene from the LSC region into the IR region, consistently observed across all species in the tribe. This rearrangement was accompanied by the apparent loss of the trnK-UUU gene, leaving only a 655 bp remnant of the 5′ intron with 73.1% sequence identity (Supplementary Fig. 5).

Vismieae tribe

The plastomes of the Vismieae tribe also demonstrated dynamic structural variation compared to Mesua ferrea, including five inversion events, IR boundary shifts, and a notably large inversion spanning 54 kb. Mauve alignment revealed 9 LCBs corresponding to six inversions and ten breakpoints, including: trnH–psbA, trnK–rps16, trnG–trnQ, rbcL–trnR, petA–accD, psbJ–rps18, clpP–rpl20, and psbB–trnN (Supplementary Fig. 2). Despite these rearrangements, a high degree of synteny was maintained among Vismieae plastomes. The IR regions were markedly expanded (~ 54 kb), distinct from other Hypericaceae tribes. As a consequence of this expansion, a partial duplication of the ycf1 gene was detected only in Vismieae plastomes.

Cratoxyleae tribe

Plastomes from the Cratoxyleae tribe exhibited a single large inversion spanning ~ 65 kb compared to Mesua ferrea. Mauve alignment identified four LCBs, suggesting two inversion events with three major breakpoints: trnH–psbA, rbcL–trnK and accD–trnN, and (Supplementary Fig. 3). The two representatives species, Cratoxylum maingayi and Eliea articulata, displayed identical syntenic block structures, indicating structural conservation within the tribe.

Ancestral plastome structural evolution in hypericaceae

Phylogenomic analysis suggests that the Cratoxyleae tribe, comprising the genera Eliea and Cratoxylum, represents the most ancestral lineage within the Hypericaceae family. A defining event in this lineage is a single large inversion spanning approximately 65 kb, representing the earliest major plastome structural rearrangement (Fig. 4). This inversion distinctly positions the matK gene within the large single-copy (LSC) region and retains the accD gene near the LSC–IR boundary in Cratoxyleae, a configuration differing from the Hypericeae tribe, where a unique rearrangement consistently relocates matK into the IR region and positions accD fully within the LSC. This pattern, characterized by a matK relocation to the IR, is specific to the Hypericeae tribe and is not observed in the Cratoxyleae tribe, which retains a more ancestral gene order, or in the structurally intermediate Vismieae tribe. In Cratoxyleae plastomes, this arrangement contrasts with the canonical organization found in the core Hypericum group. Notably, 22 plastomes across Hypericaceae show shifts in gene order near the LSC–IR boundaries, particularly involving the repositioning of trnH and rbcL, often replacing the ancestral location of matK. This rearranged configuration appears to be a defining structural signature of the Hypericeae lineage.

Fig. 4.

Ancestral reconstruction of plastome structural variation across Hypericaceae with Calophyllaceae as the outgroup. Major events, including IR shifts, inversions, and gene relocations, are shown, with blue dashed line boxes marking the three key plastome breakage points within and among Hypericaceae family members. Red (expansion) and blue (contraction) arrows indicate IR boundary shifts, respectively; pseudogenes/partial genes are denoted by the pseudogene symbol, and translocated genes are shown in rectangle black boxes. Lineage-specific changes in accD, matK, clpP, and ycf regions are highlighted across Cratoxyleae, Vismieae, Hypericeae, and the Hypericum core lineages

Another ancestral structural event occurred in the Vismieae tribe, which exhibits a ~ 54 kb inversion encompassing two large- locally collinear blocks (LCBs) (Fig. 4). This inversion appears to represent a secondary rearrangement relative to the Cratoxyleae event, with breakpoints near the accD and trnK-UUU genes. The Vismieae plastomes also display a pronounced IR expansion, extending into the ycf1 region, resulting in a partial duplication of the ycf1 gene in the opposing IR copy. In contrast, plastomes from the Hypericeae tribe reveal extensive structural complexity, characterized by a greater number of LCBs and at least 15 breakpoints localized primarily within the LSC region (Supplementary Figs. 1–3). These rearrangements reflect multiple independent inversion events and IR boundary shifts, marking Hypericeae as the most structurally dynamic lineage among the three tribes.

IR expansion and contraction in the hypericaceae family

Structural variation in the IR boundaries of Hypericaceae plastomes, relative to the ancestral plastome of Mesua ferrea, reveals dynamic patterns of expansion and contraction, resulting in gene duplication and relocation events across lineages (Fig. 4; Supplementary Fig. 4). Within the Hypericeae tribe, significant IR contraction was observed at the LSC/IRA junction in Hypericum ascyron and H. chejuense, reducing the IR by approximately 11–13 kb and excluding a block of eight genes (from rpl2 to rps12). An even more pronounced contraction (~ 17 kb) was detected in H. erectum, affecting the rps11–rps7 region at the LSC/IRA/SSC boundaries. In contrast, species such as H. laxum and T. fauriei within Hypericeae, as well as all examined members of the Vismieae tribe, showed no evidence of IR contraction. Across the family, IR expansions at both the IRB/SSC and IRA/SSC boundaries were more consistently observed. These expansions led to the duplication of ndhF in most species across Hypericeae, Vismieae, and Cratoxyleae tribes. The only exception was Thornea calcicola, which retained a full copy and a partial copy of ndhF near the IRB/SSC boundary. In addition, a complete duplication of ycf1 was detected only in H. ascyron and T. fauriei, suggesting lineage-specific expansions in these taxa (Fig. 4). A particularly notable case of large-scale IR expansion occurred in the Vismieae tribe, where the IR region extended from IRB/LSC to the rpl2–psaJ region, expanding to approximately 56 kb—nearly double the IR size compared to other tribes. This boundary shift resulted in the duplication of up to 30 genes, with no contraction events observed in this tribe. Similarly, the Cratoxyleae tribe showed no evidence of contraction but exhibited an expansion at the IRB/SSC junction, leading to the duplication of the entire ycf1 gene on one side. A partial copy (~ 981 bp) of ycf1 was found at the opposite IR boundary near the LSC/IRA junction (Fig. 4). The IR boundaries showed dynamic patterns of expansion, contraction, and gene duplication across Hypericaceae, demonstrating extensive plastome structural variation among the lineages.

Gene Relocation and ORF Emergence in Hypericum Plastomes

Comparative analysis of plastid genomes among five Hypericum species (H. ascyron, H. chejuense, H. erectum, H. laxum, and H. patulum) revealed lineage-specific structural rearrangements associated with the relocation of accD and matK genes and the emergence of novel open reading frames (ORFs) (Supplementary Fig. 5). In H. ascyron, both matK and accD are translocated into or near the inverted repeat (IR) regions. This rearrangement is accompanied by the presence of lineage-specific ORFs, including ORF13 (between accD and matK) and ORF15 (downstream of matK). These novel ORFs are absent in other species, suggesting they may have originated from IR-mediated recombination or structural rearrangement at the IR junctions. In H. chejuense and H. erectum, similar patterns are observed. While matK remains conserved, ORF66/67 (H. chejuense) and ORF8/9 (H. erectum) are located downstream, suggesting independent ORF formation following structural rearrangement events. In H. laxum, accD is placed upstream of ORF1, again implying that the disruption of plastome structure near accD may promote ORF emergence. In contrast, H. patulum shows a markedly different organization. Both accD and matK are located adjacently within the large single-copy (LSC) region, and no novel ORFs are detected in their vicinity. A more conserved structure was observed near these genes, likely due to the absence of recombination-prone IR junctions in these regions. This pattern suggests that relocation of coding genes into the IR and disruption of neighboring intergenic regions may have contributed to the formation of novel ORFs in specific Hypericum lineages.

Parallel structural placement of matK and accD genes

An unusual ~ 65 kb inversion in the basal clade of the Cratoxyleae tribe positions the trnH–rbcL region at the beginning of the large single-copy (LSC) region, placing trnH and matK in a canonical configuration typically observed in chloroplast genomes (Fig. 4; Supplementary Fig. 9). Structural comparisons across Cratoxyleae, Vismieae, and Hypericeae tribes revealed conserved adjacency of the matK on the same strand, likely originating from a single ancestral inversion. Subsequent lineage-specific rearrangements may have altered their order but retained this adjacency, indicating shared structural ancestry rather than independent parallel events. In Cratoxyleae, the matK and accD genes are positioned adjacently in the LSC region (matK:accD), a configuration inferred to represent the ancestral state. In contrast, a second large inversion (~ 54 kb) in the Vismieae tribe results in the reversal of this order to accD:matK, positioned adjacent to the inverted repeat (IR) region, along with a substantial IR expansion. This modified accD:matK orientation is also retained in the basal genus of the Hypericeae tribe (Triadenum), accompanied by the incorporation of ycf1 into the IR, further supporting a shared structural signature between Vismieae and early diverging Hypericeae plastomes. Interestingly, Thornea and H. patulum retain the ancestral matK:accD configuration, suggesting structural conservation with the Cratoxyleae lineage. However, pronounced divergence is evident in the Hypericum core clade, particularly the Trigynobrathys section (H. laxum, H. japonicum), where multiple inversion events appear to have disrupted the original gene order. In this group, the repositioning of accD into the LSC and matK into the IR region is observed, except in H. patulum, which uniquely retains the ancestral Cratoxyleae-like arrangement (Fig. 4). In Hypericaceae, the presence of a free-standing matK gene may represent a frequent ancestral breakage point, suggesting an early structural transition in the family. Additionally, matK was translocated from its typical position within the trnK-UUU intron into the IR region, a rare event among angiosperm plastomes (Supplementary Fig. 9). Comparative analysis of the accD gene across Hypericaceae plastomes revealed the presence of internal insertions comprising tandem repeat sequences, with notable variation in both length and position (Supplementary Fig. 10). Two species, H. erectum and H. chejuense, exhibited exceptionally long insertions of 1,468 and 1,484 bp, respectively, disrupting the conserved coding region of accD. In contrast, most other species contained shorter insertions ranging from 79 to 459 bp.

Selective retention of matK–trnK-UUU

In plastomes of the Hypericeae tribe, substantial restructuring occurred around the trnK-UUU region. The trnK-UUU gene was lost in all examined species, and matK was retained as a free-standing gene, uncoupled from its usual intronic context. Despite the loss of intron, matK was preserved across all species, indicating strong functional conservation (Supplementary Fig. 5 and 6). The Hypericeae plastomes retained only five introns overall, with both the cis-spliced rps12 and trnK-UUU introns being absent. Selection analysis showed dN/dS ratios > 1 for matK along the branches leading to Hypericaceae and related Clusioid lineages. Likelihood ratio tests (LRTs) supported the hypothesis that matK underwent positive selection, likely driven by structural reorganization and the functional compensation for the loss of trnK-UUU (Supplementary Fig. 6). Furthermore, although most plastid genes showed conservation, some rpo genes (e.g., rpoC1) displayed intron loss and signs of positive selection. These genes encode the RNA polymerase type I enzyme, which is essential for plastid tRNA and mRNA synthesis.

Gene and intron loss patterns in hypericaceae and related lineages

The plastomes of Hypericeae species contained 69–75 protein-coding genes, 29 tRNA genes, and four rRNA genes (Table 1). Notably, the translation initiation factor (infA), ribosomal protein S16 (rps16), and tRNA-Lys (trnK-UUU) genes were completely absent across all examined Hypericeae plastomes (Supplementary Fig. 6 & 7). Additionally, rps7 was found to be pesduogenized only in H. ascyron, H. erectum, and H. patulum, while rpl23 was pesduogenized in all Hypericeae species. The rpl32 gene also appeared to be pesduogenized in all Hypericeae members, likely due to frameshift mutations and internal stop codons. Complete or near- loss of ycf1 and ycf2 was detected in H. erectum, H. chejuense, H. patulum, and T. calcicola. Intron loss was also widespread in Hypericeae plastomes. Specifically, all species lacked both introns of the clpP gene, the second (cis-spliced) intron of rps12, the rpoC1 intron, and the second intron of ycf3. In the Cratoxyleae tribe, gene loss was limited to infA and rps16, while intron losses mirrored those in Hypericeae, with absence of introns in rps12, ycf3, and the second intron of clpP. The Vismieae tribe also exhibited complete loss of infA and rps16, and pseudogenization of rpl23, rpl32, and rps7, as well as ycf1 and ycf2 in all sampled species of Harungana and Vismia (Supplementary Fig. 7). Intron losses in Vismieae included rps12, ycf3, clpP, rpoC1, and atpF, suggesting extensive structural reduction across the plastomes of this lineage. In the related Clusiaceae family, plastome gene losses were more limited: infA was absent in both Moronobea riparia and Rheedia edulis, while rps16 was lost only in M. riparia. Both species exhibited intron loss in ycf3 but retained other intron-containing genes intact.

Phylogenetic reconstruction of hypericaceae and related malpighiales

To reconstruct evolutionary relationships within the Hypericaceae family and among related lineages in the order Malpighiales, a maximum likelihood (ML) approach was applied using a concatenated alignment of 79 plastid protein-coding genes (supermatrix length: 46,929 bp) across 29 genera, including 25 Hypericaceae species and 4 outgroup taxa (Fig. 5). The outgroup comprised representatives from the clusioid clade, including H. monogynum and C. cochinchinense (Hypericaceae), T. trifaria (Podostemaceae), M. ferrea (Calophyllaceae), and Clusiaceae species (Fig. 5). The ML analysis yielded an optimal tree topology with a log-likelihood of –226,352.109, while both ML and BI trees displayed highly congruent topologies. All major branches were strongly supported with bootstrap support (BS) = 100 and Bayesian posterior probability (PP) = 1.00 (Fig. 5). Phylogenetic analyses recovered the Cratoxyleae tribe as a monophyletic group and sister to a clade composed of Vismieae and Hypericeae, confirming tribe-level relationships within Hypericaceae. Within this framework, Vismieae was recovered as sister to Hypericeae, indicating a closer evolutionary affinity between these two tribes.

Fig. 5.

Maximum likelihood phylogeny of Hypericaceae based on 79 plastid protein-coding genes. The tree was reconstructed using concatenated nucleotide sequences of 79 plastid protein-coding genes from 29 Hypericaceae species and one outgroup. The tree resolves three major tribes of Hypericaceae: Hypericeae, Vismieae, and Cratoxyleae, each indicated with colored blocks and corresponding floral images. Species of Hypericum are further classified into “old” and “new” world lineages. The genus Hypericum exhibits multiple distinct clades, including H. ascyron, which forms a separate lineage within the new world clade

Among the Hypericum species: H. laxum and H. japonicum grouped closely with Triadenum fauriei and T. fauriei, forming a well-supported clade corresponding to the Trigynobrathys section (BS/PP = 100/1.00). Thornea species also clustered within this clade, confirming their close phylogenetic relationship to Triadenum and the Trigynobrathys section. H. patulum and H. monogynum formed a clade within the Ascyreia s.l. + campylosporous (Hypericum core sect.) section, which was resolved as sister to the Roscyna section represented by H. ascyron (BS/PP = 100/1.00). In the core Hypericum, H. erectum and H. chejuense formed a distinct clade (BS/PP = 100/1.00), which was recovered as sister to the combined Ascyreia s.l., Roscyna, and campylosporous lineages. Within Vismieae, Harungana and Vismia species were recovered as a well-supported monophyletic group. In Cratoxyleae, Cratoxylum and Eliea formed a strongly supported sister clade to the Hypericeae–Vismieae clade (BS/PP = 100/1.00), further confirming tribal boundaries. Additional ML phylogenies constructed using individual plastid gene sets (e.g., atp, ndh, pet, psa, psb, rpl, rps, rpo, ycf) yielded consistent topologies, though variation was observed in certain marker-based trees (ccsA, clpP, pet, psb, ycf), which showed slight topological shifts in relationships among tribes. Analysis of nuclear ITS sequences provided complementary evidence, particularly supporting the separation of H. laxum and H. japonicum from core Hypericum, instead of grouping them with Triadenum, thereby corroborating plastome-based results (Supplementary Fig. 8). Notably, the relationships among Thornea, Triadenum, and the Trigynobrathys section appeared controversial in certain gene trees, particularly those constructed using rps, pet, and ycf genes. Collectively, both coalescent-based and concatenated tree methods support a scenario in which are the New World Hypericum clades (Thornea, Triadenum, Trigynobrathys) diverge early, While both New and Old-World clades (Androsaemum, Ascyria, Roscyna, and the Hypericum core) form a monophyletic group sister to Vismieae. These results provide robust phylogenomic evidence for major taxonomic realignments within Hypericaceae and reinforce plastome structure and gene order as useful phylogenetic markers for resolving complex tribal and sectional relationships.

Elevated substitution rates in hypericaceae

Analysis of substitution rates showed that plastid genes in Hypericeae experienced significantly elevated rates of molecular evolution in specific bordered disturbed genes among the tribe. Selection analysis using RELAX detected intensified selection across several plastid genes in Hypericaceae, with clpP (k = 10.05), rpoC1 (k = 6.81), accD (k = 2.75), matK (k = 2.75), and rps3 (k > 1) all showing k > 1 relative to the outgroups. Corresponding dN/dS ratios based on ASTRAL comparisons were also elevated (rps3 = 54.01, accD = 47.83, rpoC1 = 32.1, clpP = 4.17, matK = 2.73; p < 0.000) (Supplementary Fig. 10–16). Among these, accD, clpP, rpoC1, and matK show the strongest and most consistent evidence of positive or intensified selection, supported by both analyses. Although ω values were likely inflated by low synonymous substitution rates, the agreement between RELAX and dN/dS results indicates these genes have undergone non-neutral and potentially adaptive evolution in Hypericaceae. Branch-site models and RELAX analyses further revealed that several other genes in Hypericaceae underwent episodic positive selection at species or genus level (Supplementary Table 2–3).

Discussion

In a previous study, we reported that the plastid genome of H. ascyron exhibited extensive structural disruption, including multiple large inversions, gene and intron loss, and significantly elevated substitution rates [7]. These features positioned H. ascyron as one of the most highly rearranged plastomes among angiosperms and raised questions about whether this instability reflected a family-wide patterns or species-specific feature. The present study addresses this by expanding the taxonomic scope to include representatives from all three tribes of Hypericaceae (Hypericeae, Vismieae, and Cratoxyleae) and conducting comparative analyses with outgroups from Clusiaceae Our results confirm that these features are shared across Hypericaceae rather than unique to H. ascyron.

Plastome structural dynamics evolution in Hypericaceae

Plastome structural variation was extensive and lineage-specific. Plastome sizes ranged from 138,163 bp (H. patulum) to 176,170 bp (Harungana madagascariensis), with much of this variation attributable to IR expansions and contractions. Overall plastome size comparison highlights both the evolutionary conservation and structural divergence across Malpighiales, with Hypericaceae and its close relatives occupying a distinct position within this landscape. Members of the Vismieae tribe exhibited dramatic IR expansions exceeding 50 kb, leading to the duplication of over 20 genes. Conversely, significant IR contraction in Hypericeae resulted in the loss of multiple boundary-associated genes. A large ∼65-kb inversion linking the rbcL–psbA and matK/trnK-UUU–accD regions appears to be shared across all three tribes of Hypericaceae, indicating an ancestral structural signature predating their divergence. Subsequent tribe-specific inversions, particularly in Hypericeae and Vismieae, further reshaped the plastome architecture. Comparable rearrangements in Passifloraceae and Podostemaceae [2, 4]suggest that plastome instability is an emergent trait of Malpighiales lineages undergoing ecological specialization. Gene loss and pseudogenization were most pronounced in the Hypericeae tribe, where infA, rpl23, rpl32, and rps16 were absent in all species, while rps7, ycf1, and ycf2 were pseudogenized in several lineages. These losses likely reflect functional transfer to the nuclear genome, a pattern documented in Ranunculaceae and Passifloraceae [41, 61]. Examples from Populus and Geraniaceae [65–67] demonstrate that plastid gene loss is often coupled with nuclear compensation rather than random decay.

Gene evolution and adaptive plastome remodeling in hypericaceae

The consistent relocation of matK into the IR region and its decoupling from the trnK-UUU intron further suggest a functional repurposing under lineage-specific selective pressures. This inference is supported by our dN/dS analysis, which revealed elevated substitution rates in matK across multiple branches in the Hypericeae tribe. Notably, in several species where matK has translocated, we observed the presence of adjacent partial or large open reading frames (ORFs), a feature absent in closely related species where matK remains embedded within the trnK-UUU intron. These ORFs may represent truncated pseudogenes, novel gene fusions, or co-evolving loci, and their persistence suggests a possible structural or regulatory role, although functional validation is required. The relocation of matK or accD into IR regions, as seen in Hypericum species, may lead to structural disruption of neighboring intergenic regions and generation of lineage-specific ORFs, a phenomenon documented in other taxa [17, 69, 75]. The presence of palindromic and inverted repeats at IR junctions likely facilitated these events through homologous or illegitimate recombination [11, 23, 38]. Additional processes such as micro-recombination or slipped-strand mispairing [28, 58] may also have contributed. The retention of certain ORFs across species may result from homogenizing recombination between IR copies, preserving their sequence integrity [14, 39, 43]. Comparable IR-mediated rearrangements in other angiosperms [68] support the likelihood of such mechanisms in Hypericum [68]. Evidence from previous studies reinforces the idea that matK is subject to adaptive evolution. [19] showed positive selection in functional domain of matK [25, 69] demonstrated its ability to function as a freestanding gene; Recent work [45, 71] links plastome rearrangements with ecological transitions. Together, these studies and our results suggest that matK relocation is not random but may represent a potentially adaptive structural adjustment. In addition to matK, several functional genes including accD, clpP, and rpoC1, which showed intron loss and elevated substitution rates. These genes are hotspots for adaptive evolution linked to plastid performance under environmental stress, as seen in Silene and Geraniaceae [13, 68].

Phylogenetic relationships and tribal evolution

Phylogenomic reconstruction using 79 plastid protein-coding genes yielded a well-resolved tree supporting the monophyly of Hypericeae, Vismieae, and Cratoxyleae. Cratoxyleae was recovered as the earliest diverging lineage and sister to a clade comprising Vismieae and Hypericeae. This topology clarifies deep relationships that were previously ambiguous [34, 35, 42]. Cratoxyleae, with its conserved structure and Southeast Asian distribution, likely retains the ancestral plastome configuration. By contrast, Hypericeae and Vismieae exhibit greater rearrangements and IR expansions associated with derived lineages and ecological diversification [34].Within Hypericeae, H. laxum and H. japonicum form a strongly supported clade with Triadenum and Thornea, supporting their inclusion within a redefined Hypericum. This pattern corresponds to previous nuclear and morphological evidence [30, 33] and suggests paraphyly within core Hypericum. Morphological similarities may reflect convergent or retained ancestral traits rather than deep taxonomic splits [36]. The H. laxum–Triadenum clade corresponds to New World sections (e.g., Trigynobrathys), whereas H. erectum and H. patulum align with Old World clades (Ascyreia, Roscyna). This East–West divergence matches Oligocene–Miocene climatic oscillations that promoted montane colonization and speciation [29, 37]. Our plastome-scale phylogeny indicates that these biogeographic patterns may be associated with genome remodeling.

Evolutionary and ecological implications

Environmental shifts likely acted as major drivers of plastome dynamism. Cold-tolerant Hypericum species radiated rapidly in montane regions such as the Andes and East Asia [37]. Structural genome evolution—including gene relocation, intron loss, and inversion—may reflect adaptive responses to these pressures, as also seen in Geranium and Silene [13, 40]. Taken together, these findings suggest that plastome restructuring in Hypericaceae represents a family-wide adaptive strategy linked to ecological diversification.

Conclusion

This study expands plastome analyses across all three tribes of Hypericaceae, revealing extensive structural rearrangements, IR expansions, and recurrent gene losses across the family. The Hypericeae tribe showed extensive plastome rearrangements and a pronounced abundance of tandem repeats. The family exhibited recurrent loss of genes and introns, along with elevated substitution rates compared with Podostemaceae, indicating accelerated plastome evolution. Overall, Hypericaceae plastomes display high structural and evolutionary flexibility, offering new insights into plastid diversification within Malpighiales.

Materials and methods

Plant material and Next generation sequencing

This study analyzed the plastid genomes of 14 species across the families Hypericaceae and Clusiaceae, including eight species from the tribe Hypericeae, three from Vismieae, and two from Cratoxyleae. Additionally, two species from Clusiaceae served as outgroups. Plant material used in this study was obtained either as leaf tissue or DNA from the Kew DNA Bank. Wild-collected samples from Korea were collected from common, non-protected species and therefore did not require special permits under Korean biodiversity regulations. Formal identification of the species was performed by SeonJoo Park, and voucher information is provided in Supplementary Table 1. Additional specimens were obtained from established DNA and herbaria, including the Missouri Botanical Garden, the Kew DNA Bank, and the University of Texas Herbarium, with voucher details available in Supplementary Table 1.This study did not involve the collection of any species listed in the Convention on the International Trade in Endangered Species of Wild Fauna and Flora (CITES). Fresh leaves were used for several species, including H. erectum Thunb., H. patulum Thunb., H. chejuense S.J.Park & K.J. Kim, H. laxum Blume., H. japonicum Thunp., and Triadenum fauriei R.Keller, with genomic DNA extracted using the GeneAll Plant SV Mini Kit (GeneAll Biotechnology, Seoul, Korea). Herbarium specimens leaves such as Harungana madagascariensis Lam. Ex Poir., Vismia baccifera (L.) Triana & Planch., Eliea articulata (Lam.), Thornea calcicola Standl. & Steyerm., Rheedia edulis (Seem.) Planch. & Triana., and Moronobea riparia Planch. & Triana. were processed using a modified 3 × CTAB protocol with a 3-h incubation, as described by Doyle and Doyle DNA extraction method [1]. For Cratoxylum maingayi and Harungana rubescens (Oliv.) Byng & Christenh., DNA was sourced from the Kew DNA Bank. High-throughput sequencing was carried out on the Illumina HiSeq 2500 platform (Illumina Inc., San Diego, CA), generating approximately 6 Gb of 150 bp paired-end reads per sample from 550 bp insert libraries.

Assembly and genome size comparison

Chloroplast genome assembly was conducted using GetOrganelle v1.7.5.3 and Velvet v1.2.10, with k-mer sizes ranging from 97 to 147 optimized for genome coverage and accuracy [21, 74]. Assembly quality was evaluated by mapping paired-end reads to the assembled plastomes using Bowtie2 v2.2.6 [26]. The results were visually inspected and verified in Geneious R11.0.5 (https://www.geneious.com). To analyze genome size variation, 281 complete chloroplast genomes from the order Malpighiales were downloaded from the NCBI nucleotide database as of February 28, 2022. Genome size statistics and visualizations were generated in R v4.0.4 using the ggplot2 package [70]. Summary statistics (mean ± standard deviation and median) were calculated for each region and family group. Differences in plastome length distributions between Hypericaceae and other Malpighiales were evaluated using two-sided non-parametric Wilcoxon rank-sum tests implemented in R (v4.x) using the ggpubr, dplyr, and ggplot2 packages. Adjusted p-values were computed using the Benjamini–Hochberg false discovery rate (FDR) correction to account for multiple comparisons. All newly assembled plastome sequences have been deposited in the NCBI GenBank database under accession numbers [PQ010624–PQ010637].

Plastid gene annotation and Genome rearrangement

Gene annotation was initially performed in Geneious R11.0.5 using Nicotiana tabacum as the reference genome [60]. Gene boundaries were verified by BLASTN searches using NCBI-BLAST + v2.7.1 [3]. Transfer RNA genes were annotated with tRNAscan-SE v2.0.3 [5] and ARAGORN v1.2.38 [27], and circular genome maps were constructed using OGDraw v1.3.1. [15]. Structural rearrangements were investigated by comparing species from each tribe of Hypericaceae with an outgroup species (Mesua ferrea L., Calophyllaceae) using the progressive Mauve algorithm implemented in Geneious [9], Geneious).

Phylogenetic and substitution rate estimation

Phylogenetic reconstruction was based on 79 plastid protein-coding genes extracted from 29 plastomes. Each gene was aligned using MAFFT v7.450 with the G-INS-i strategy and concatenated into a supermatrix [22]. Phylogenetic inference was carried out using maximum likelihood in IQ-TREE v1.6.2 under the GTR + GAMMA + I model with 1,000 ultrafast bootstrap replicates [31], and Bayesian inference using MrBayes v3.3.7a [55]. Coalescent-based species trees were generated using ASTRAL v5.7.8 with 79 gene trees derived from IQ-TREE, incorporating models selected by ModelFinder [32]. Additionally, nuclear ITS regions were retrieved using GetOrganelle and aligned with MUSCLE [12]. ITS based phylogenies were reconstructed using IQ-TREE with the GTR + GAMMA + I model.

Selection analysis

To detect selection, nonsynonymous (dN) and synonymous (dS) substitution rates were estimated for each plastid gene using the CODEML program within the PAML v4.8 package under the F3 × 4 codon frequency model [73]. Branch-site positive selection was tested using the adaptive Branch-Site Random Effects Likelihood (absREL) model in HyPhy v2.5 via the Datamonkey server, with Holm-Bonferroni correction for multiple testing [44]. A total of 70 protein-coding genes were aligned using the Translation Align option in MAFFT, and likelihood ratio tests (LRTs) were conducted to evaluate variation in dN/dS ratios across branches. Visualization of substitution rates was performed in R v4.0.4 using ggplot2 [70]. To further explore divergence in the inverted repeat (IR) regions, the ycf1 and ycf2 genes from 11 genera of Malpighiales, including a known pseudogene in Passiflora edulis, were aligned using MAFFT. Phylogenetic trees were constructed using PhyML with 100 bootstrap replicates to assess evolutionary patterns associated with IR-specific gene dynamics [16].

Authors’ contributions

JCS performed the experiments, generated datasets and figures, and wrote the first draft of the manuscript. KTP contributed to the data assembled, performed analyses, and read/edited the manuscript. SJP contributed to project design and read/edited the manuscript. All authors read and approved the final draft of the manuscript.

Funding

This research was supported by the Gyeongsangbuk-do RISE (Regional Innovation System & Education) project [2025-RISE-B0080529002321].

Data availability

All newly generated plastome sequences have been deposited in GenBank under accession numbers [PQ010624–PQ010637]. Supporting data, including alignment files, gene order maps, and phylogenetic trees, are available in Supplementary Figures and Tables.

Declarations

Ethics approval and consent to participate

Experimental study on the plant, including collection of the material, comply with institutional, national, and international guidelines.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.