Sliding‐window phylogenetic analyses uncover complex interplastomic recombination in the tropical Asian–American disjunct plant genus <i>Hedyosmum</i> (Chloranthaceae)

The chloroplast genomic (i.e. plastomic) sequences have long been used for inferring phylogenetic relationships of green plants. Current major plant classifications (e.g. Angiosperm Phylogeny Group classification, APG IV, 2016; Pteridophyte Phylogeny Group classification, PPG I, 2016) are predominantly based on the plastid phylogenies (Stull et al., 2023). Due to the rapid progress in DNA sequencing technologies along with decreasing costs, phylogenetic analyses using whole plastomes have become a routine practice (Wang et al., 2024). Plastomes have been presumed to be single double-stranded circular DNA molecules that are inherited uniparentally, with maternal inheritance in most angiosperms and paternal inheritance in gymnosperms (Birky, 1995; Dong et al., 2012; Greiner et al., 2015). These characteristics led to the general belief that plastomes are free from or less likely to undergo intermolecular recombination (Walker et al., 2019). Therefore, different plastomic genes or regions, which are assumed to share the same evolutionary trajectory, are often concatenated directly for phylogenetic analyses (Jansen et al., 2007; Moore et al., 2010; Li et al., 2021). Despite the widespread use of plastomes in phylogenetics, both biparental inheritance and recombination of plastomes – processes that could inadvertently affect inference – have been increasingly detected. The mechanisms that maintain uniparental inheritance, including elimination or degradation of the organelle during male gametophyte development or after pollen mitosis or fertilization, may break down and lead to biparental inheritance (Nagata, 2010). Biparental inheritance of plastomes has been reported in some plant groups, such as Passiflora (Passifloraceae; Hansen et al., 2007; Shrestha et al., 2021), Cicer arietinum (Fabaceae; Kumari et al., 2011), and Actinidia (Actinidiaceae; Li et al., 2013). It is believed that heteroplasmy, that is the mixture of different organelle genomes within a cell or individual, is widespread in both animals and plants (Nagata, 2010; Ramsey & Mandel, 2019; Camus et al., 2022), and c. 20% of angiosperm genera may have undergone biparental inheritance (Zhang & Sodmergen., 2010; Sakamoto & Takami, 2024). The biparental inheritance allows the coexistence of both maternal and paternal plastids in the same offspring cell, creating opportunities for interplastomic recombination. Interspecific plastomic recombination has been created and detected in experimental studies (Medgyesy et al., 1985). However, unlike in the intraplastomic recombination, in which causes (e.g. mediated by short repeat sequences) and consequences (e.g. generating structural variations, such as insertions, deletions, isomers, and concatemers) have been extensively studied (Kolodner & Tewari, 1979; Palmer, 1983; Ogihara et al., 1988; Bendich, 2004; Oldenburg & Bendich, 2004, 2016; Kobayashi et al., 2017), interplastomic recombination, especially historical or natural ones, has rarely been explored, with only a few cases reported (Sullivan et al., 2017; Sancho et al., 2018; Zhu et al., 2018). If interplastomic recombination happened, particularly from divergent parental plastomes, the traditional viewpoint of a plastome as a single nonsegregated locus should be overthrown (Gonçalves et al., 2019; Doyle, 2022). Different regions of the recombined plastome undergo distinct evolutionary trajectories, which are expected to be more closely related to their respective parental (i.e. paternal or maternal) lineages (Fig. 1; more in the Materials and Methods section). If that is the case, they should not be combined in phylogenetic analyses. However, in most phylogenetic studies, the plastid regions are routinely concatenated into a supermatrix, which is then used for phylogenetic inference. In recent years, an increasing number of studies have detected phylogenetic conflicts within plastomes (Walker et al., 2019; Xiao et al., 2020; Zhang et al., 2020; Liu et al., 2022; Thureborn et al., 2024). Conflicts have been demonstrated among different plastomic regions, genes, and functional gene groups (Liu et al., 2012; Walker et al., 2014; Saarela et al., 2018; Zhang et al., 2020; Yang et al., 2021). These conflicts have usually been attributed to stochastic or systematic errors, such as from methodological deficiencies, molecular evolutionary complexities, or as artifacts of incomplete taxon sampling (Burleigh & Mathews, 2007; Liu et al., 2012; Xue et al., 2024). For example, nonfunctional regions, such as intergenic spacers, evolve faster, which are prone to reach saturation, leading to phylogenetic trees that are inconsistent with those derived from coding regions (Liu et al., 2022; Wang et al., 2022). Additionally, rate heterogeneity among different plastid genes or functional groups and horizontal gene transfer between plastomes and mitogenomes or nuclear genomes are also possible factors associated with intraplastomic phylogenetic discordance (Smith, 2014; Zhang et al., 2020; Daniell et al., 2021). Almost no intraplastomic phylogenetic conflict was definitely or exclusively ascribed to interplastomic recombination (but see Sullivan et al., 2017; Sancho et al., 2018; Zhu et al., 2018). Particularly, it has not been elucidated why and how intraplastomic phylogenetic conflict can be reliably attributed to interplastomic recombination. Neither a clear rationale nor a proper statistical framework has been proposed for detecting or assessing the confidence of interplastomic recombination through phylogenetic analysis. Chloranthaceae is an early diverged angiosperm family comprising four extant genera – Ascarina, Chloranthus, Hedyosmum, and Sarcandra – and over 70 species (Todzia, 1988; Zhang et al., 2011). Chloranthus and Sarcandra have bisexual flowers, while Ascarina and Hedyosmum have unisexual flowers and are usually dioecious (Todzia, 1988; Kong et al., 2002). Hedyosmum possesses over 40 species, which are distributed in tropical America, except for H. orientale isolated in Southeast Asia (Todzia, 1988). Traditionally, Hedyosmum is classified into two subgenera, that is, Hedyosmum and Tafalla. Previous phylogenetic studies utilizing several plastid fragments and nuclear ribosomal internal transcribed spacer (ITS) sequences demonstrated widespread phylogenetic conflicts in Hedyosmum (Zhang et al., 2015). These conflicts involved even the deepest splits within the genus, in which H. orientale was suggested to be either the sister of all other Hedyosmum species or nested within them, being the sister to subgenus Tafalla (Zhang & Renner, 2003; Antonelli & Sanmartín, 2011; Zhang et al., 2011, 2015). These conflicts were detected only between the plastid and nuclear ITS sequences and are, at least partially, ascribed to possible rampant hybridization or reticulated evolution in Hedyosmum (Zhang et al., 2015). However, whether the phylogenetic conflicts exist within the plastome or between plastid fragments has not yet been explored. In this study, we first propose a rationale and then present a novel statistical framework for detecting interplastomic recombination based on sliding-window phylogenetic analysis (SWPA). Then, based on 22 plastomic sequences of Chloranthaceae, we investigate whether there are any intraplastomic conflicts within Chloranthaceae, particularly in the largest genus Hedyosmum, which experienced complex reticulate evolution. For any such conflict, we also examine whether it can be attributed to interplastomic recombination and what abiotic settings might have been associated with it. Using our methodology, we indeed find a striking phylogenetic conflict in Hedyosmum that is reliably ascribed to interplastomic recombination; that the complex interplastomic recombination with two or more instances of template shifts happened around the Oligocene–Miocene boundary; and that the involved lineages are now disjunct in tropical Asia and America, possibly as a result of global climate cooling or long-distance dispersal. This study highlights the necessity of examining and separately using different intraplastomic fragments with alternative evolutionary trajectories in phylogenetic analyses. The methodology could be easily applied or extended to detect intermolecular recombination elsewhere, such as in nuclear genomes. When interplastomic recombination is absent (Scenario I in Fig. 1), each plastome is inherited as a single coherent unit or locus. All fragments or sites within a plastome are expected to have the same evolutionary trajectory and relationship. In this scenario, the derived phylogenetic trees from each sliding window are anticipated to be congruent and accurately reflect the true relationships, provided the phylogenetic inferences are accurate. Nevertheless, because of insufficient informative sites and/or imperfections of the phylogenetic reconstruction methods (Nei & Kumar, 2000), spurious relationships could be stochastically inferred. Consequently, the true relationship would be recovered as the dominant topology, with a minority of inferred alternative false topologies. Taking a four-taxa (sub)tree as an example, as depicted in Fig. 1 Scenario I, the true relationship of (((A, B), C), outgroup) is expected to be the most frequently recovered in the sliding-window phylogenies. By contrast, the two alternative false topologies (((A, C), B), outgroup) and (((B, C), A), outgroup) are anticipated to occur occasionally and, p equally probably. Furthermore, both of them would generally exhibit features of low support, infrequency, and improbability of occurring in multiple consecutive windows. When interplastomic recombination occurs (Scenario II in Fig. 1), different evolutionary trajectories would be combined into the same plastome. In a simplified recombination scenario where plastome G is derived from the recombination of plastomes E and F, different regions or fragments in G would surely exhibit closer relationships with those from their respective donor plastome lineages. If the plastomes are divided into smaller fragments as in the SWPA, it is anticipated that two primary phylogenetic relationships will be uncovered: (((E, G), F), outgroup) and (((F, G), E), outgroup). Similarly, the informative sites are not always dominant and the complexity of the data cannot always be accurately modeled by the phylogenetic methods, the incorrect relationship, (((E, F), G), outgroup), may be occasionally recovered. The false minority topology should also have features of poor support, rarity, and discontinuity. The presence of two (or more) alternative dominant trees is crucial for validating the occurrence of interplastomic recombination. If two topologies, for example T1 and T2, are both supported by significantly considerable or consecutive windows, then both topologies should represent true evolutionary trajectories since they are very unlikely to be obtained by chance. Assuming the context of no or random phylogenetic information where no true phylogenetic signals govern and T1, T2, and T3 are equally probably obtained just by chance (null hypothesis), the probability to acquire a specific tree, for example T1, in a given or more quantity, can be formulated as P = ∑ x = N T 1 N T N T ! x ! N T − x ! 2 N T − x / 3 N T = ∑ x = N T 1 N T C N T x 2 N T − x / 3 N T $$ $$ in N T 1 $$ $$ the number of T1 and N T $$ $$ the of T1, T2, and If the are N T 1 $$ $$ can represent the of T1 in while N T $$ $$ can to the of T1, T2, and the probability of a topology that not represent the true evolutionary trajectory is supported by a number of consecutive by chance can be as P = 1 3 N $$ $$ the in each of the consecutive sliding is the probability can be as P = = 1 N ∑ x = N ! x ! − x ! 2 − x / 3 = = 1 N ∑ x = N C x 2 − x / 3 $$ $$ in N $$ $$ the in the consecutive sliding plastomes in this study were obtained from for and comprising 22 Chloranthaceae species Ascarina, Chloranthus, Hedyosmum, and with 2 species as analysis that the gene of the 22 Chloranthaceae plastomes is and no occurs (Fig. The only is that repeat occurs in Hedyosmum species, H. H. H. H. and H. in the of Therefore, the whole plastome sequences were generally in Chloranthaceae using & 2013). of template and in interplastomic recombination and the of phylogenetic as a of sliding-window two of sliding-window were The first was of with a of and the with a of 2 regions and sites for each were by using the = and in (Zhang et al., 2023). were for each window using and the is less frequently in phylogenetic studies, it has a distinct with the Therefore, their can be used as for detecting the phylogenetic conflicts among sliding and possible systematic to each inferences were with The was in the with inferences were in a with of random and a The inferences all equally and were as was through using the same The phylogenetic relationships were by to their respective sliding along the whole plastomic with the of the more the recombination break particularly whether they are in genes or coding we the phylogenetic signals along the chloroplast using as in Zhu et for sliding = and = and each were the topologies, which from each other by the of H. orientale and Then, the supported was for each window and by the among the topologies. The of the were as and the traditional plastome particularly in Sullivan et and Zhu et in which the phylogenetic conflicts between such data were as from interplastomic recombination, we and trees based on the data coding regions and the single single and whether any phylogenetic conflict could be between data particularly for the intraplastomic phylogenetic conflict that was detected and attributed to interplastomic recombination by using our methodology based on the number of recombination methods in the et al., were also as the with a window of & 2000), the with a of the with a window of a of a random number of and a of 70 et al., the with a window of (Smith, the with a window of & the with a window of and a of 40 et al., 2000), and the with the settings et al., 2018). The recombination detected by more methods were to Chloranthaceae and related have been reported and et al., 2024). them, two and were the most frequently used in the analyses of Chloranthaceae (Zhang & Renner, 2003; Antonelli & Sanmartín, 2011; Zhang et al., 2011). the phylogenetic of (Zhang et al., 2011), we the taxon as early as c. in our analyses to the of Hedyosmum, as it was to represent a of Hedyosmum (Zhang et al., 2011, et al., 2024). In the between and Chloranthaceae Moore et al., 2010; Li et al., 2019; et al., et al., was also used as a Due to the that two major evolutionary trajectories were detected from regions within the plastomes and we concatenated the regions with the same trajectory and them were the distributed in et al., with a and I G The of the most recent of and Chloranthaceae, that is, the of the phylogenetic tree, was with a = = and the of the of Hedyosmum was with a = = = to Zhang et were for with and trees The of the and of the two were using et al., that the for all et al., was to the after the first 20% of trees as The sliding-window phylogenetic analyses and topologies (Fig. in which H. orientale is either sister to other subgenus Hedyosmum species or subgenus Tafalla or all other Hedyosmum species The and methods generally trees with the same topology in each of the sliding windows, except of T1 is the most frequently recovered topology, supported by of and of window In the T1 is supported by and two consecutive windows. In the T1 is supported by and instances of two and consecutive windows. is the most frequently recovered topology, which is supported by of the and of window In both and is supported by two genomic regions, that is, of consecutive and the other of an of T3 is the least supported topology, which is supported by of the window and and of window phylogenies and between T1, supported by most and consecutive with to support, is the dominant topology and the true evolutionary trajectory of the plastomic is less T1, it an alternative major topology and is supported by and two consecutive with to in the and (Fig. Therefore, they should most likely represent two distinct evolutionary trajectories within the plastome of H. which a relationship either to other subgenus Hedyosmum species or subgenus Tafalla in considerable and consecutive windows. This of the sliding-window phylogenies with the recombination as in Fig. the we that no recombination has happened and the most frequently recovered T1 is the only true evolutionary trajectory, then we have to that is not true and just from stochastic and/or systematic in phylogenetic stochastic or systematic would not any specific topologies, they have the same chance to T1, T2, or T3 probability for each of the Then, the probability of consecutive window phylogenies to the same assumed false topology is If the for are into the probability of in consecutive by chance would be to Therefore, we that phylogenetic signals to indeed particularly in the consecutive of the two consecutive in both and they might also reflect their true evolutionary trajectory with less confidence with those with more consecutive the same T3 is supported by a of four consecutive in the inference and two in the inference with low T3 an alternative true evolutionary trajectory or was just obtained by for example to the of the is that the plastome of H. orientale from a recombination of two diverged is the plastome of a comprising two other species in subgenus Hedyosmum, and the other is the plastome of a comprising the species of the subgenus Tafalla. occurs to template in the plastome & Bendich, Therefore, the recombination related to H. orientale experienced at least two template shifts during plastome with in the and the other in the If T2, derived from and based on both and methods, also a true evolutionary trajectory, we two more template around to and the other around to The inferred interplastomic recombination that hybridization have happened between two divergent Hedyosmum and the nuclear genes of H. orientale might be to the biparental of subgenus Hedyosmum and subgenus However, H. orientale has with either the nuclear from might have been such the inferred interplastomic recombination may be the that the hybridization In of the two striking shifts detected is in regions such as at the of or and Therefore, we that template shifts may in interplastomic recombination. The with more with the and from the analyses (Fig. The some inconsistent relationships between and For example, for window in the SWPA, T1 and T2, the and trees from the same Therefore, for the in the analyses should be an as the window in the the phylogenies more topologies and of and and of and of the and window = which likely reflect more stochastic to the in phylogenetic particularly for the relationships among the lineages involved in the interplastomic recombination and = are more sliding window and phylogenetic analyses are generally with the from the and (Fig. The of the and sites T1, as in Fig. The of and the sites (Fig. T3 is supported by a plastomic c. (Fig. more phylogenetic and their more are detected the and These phylogenetic might represent shifts of true evolutionary trajectories or false phylogenetic of and 40 and and to be in the intergenic regions and as intergenic regions multiple distinct topologies with different phylogenetic signals usually both and intergenic In no phylogenetic or template can be to coding the possible of sequencing and from short and particularly possible (but not in the species H. we an plastome of H. orientale that was from long then an and that the plastome of H. orientale used in the analyses from the by only These could be attributed to and are not to the and The phylogenetic analyses that concatenated coding regions or genes concatenated regions, or the regions and separately to detect any supported phylogenetic conflict in Chloranthaceae and are to any interplastomic recombination The coding and trees are in topology (i.e. in which H. orientale is sister to the comprising the other two Hedyosmum species = and Similarly, the coding and trees have the same topology T2, in which H. orientale is sister to subgenus Tafalla = and Fig. The trees based on the and regions also have the same topology T1 = and trees based on the and regions topologies and among the they are not supported in the and the = and it more likely to have from inference true evolutionary trajectories from interplastomic recombination (Fig. The to detect phylogenetic conflict is because for example, the and regions, each distinct evolutionary trajectories, which could the phylogenetic conflicts between routine may have to detect phylogenetic conflicts in reliably conflicts to interplastomic recombination is not that to the of a clear and statistical as proposed in this Furthermore, interplastomic recombination traditional phylogenetic analyses may only the true evolutionary relationships or even result in Therefore, we that phylogenetic studies should into intraplastomic phylogenetic conflicts from interplastomic recombination. The recombination methods in different recombination in Chloranthaceae The plastomes of a of species, including Chloranthus, two Hedyosmum, and Sarcandra species, are suggested to be interplastomic recombination by at least four of them has been by our methodology, as no striking phylogenetic conflict is involved with the H. orientale by our methodology has not been detected by Nevertheless, we the of our methodology with a clear rationale and a statistical It is very unlikely or that the interplastomic recombination happened in species, with some even between isolated as suggested by the methods in the methodology we propose should be at least an to the recombination methods, and two different evolutionary trajectories we the of the recombination by inferring the of H. orientale T1 and T2, on the T1, the of H. orientale was to be while it was to based on the (Fig. These two with the that the fragments T1 and in H. orientale diverged from the two and recombined into plastome inherited by H. The is the Oligocene–Miocene which is with the based on the nuclear ITS and the combined data of ITS and two plastid fragments & Sanmartín, 2011; Zhang et al., 2015). the evolution of Hedyosmum as a whole has been extensively studied and & Sanmartín, 2011), the which might have been associated with the recombination and lineages is The climate c. Fig. might have supported such as Hedyosmum now disjunct in tropical Asia and America, to in the no or which were distributed in all the in the et al., have been from that The climate in the might have Hedyosmum to to low to the present in tropical Asia and However, the disjunct in was ascribed to a by (Wang et al., long-distance has also been used to the disjunct in et al., et al., and et al., 2023). that the of Hedyosmum are usually and is to be by a long-distance by cannot be yet for the disjunct of Hedyosmum in tropical Asia and In this study, we the rationale and a statistical framework based on for detecting interplastomic recombination. Using our methodology, we a supported interplastomic recombination in This could not be detected by traditional phylogenetic analyses and recombination The plastome is to result from two or more along two distinct parental DNA in with in the and the other in the The the regions of the with different evolutionary trajectories diverged from their two at the Oligocene–Miocene being with the of recombination. The and the two parental lineages are disjunct in tropical Asia and America, as an of climate cooling or long-distance dispersal. for of Hedyosmum and This study was supported by the of and the of and of of of from the the for and the of this and the led the data analyses with from and the of the led the with from and and equally to this The data the of this are in the Fig. plastome of 22 Chloranthaceae Fig. topologies T2, and from the sliding-window phylogenetic analyses and their to each of the along the plastome. Fig. sliding window and phylogenetic signals in the Chloranthaceae Fig. The inferred phylogenies based on coding and Fig. The inferred phylogenies based on single single and repeat and in this of the of the sliding = = the topologies is not for the or of any by the should be to the The is not for the or of any information by the should be to the for the The with to in and in any