Background

Pueraria montana var. lobata (Willd.) Sanjappa & Pradeep (2n = 2x = 22) is a semi-woody, perennial liana, which belongs to the Leguminosae family. It is native to Southeast Asia and introduced in Africa, America, and Europe [1]. As an economic crop, it contains puerarin and other functional components and is used in the production of both pharmaceuticals and health foods [2]. Its dried rhizome is listed in the Chinese Pharmacopeia as kudzu [3]. Kudzu was introduced from Asia to America about 150 years ago for soil erosion control and cattle feed, but later became a highly invasive species [4].

Due to considerable morphological diversity, complex domestication history, and similar health and cosmetic benefits, Pueraria montana var. lobata and the other two varieties faced a protracted period of taxonomic uncertainty. The dried rhizome of Pueraria montana var. thomsonii (Benth.) Wiersema ex D.B. Ward (common name “Fenge” in Chinese) is usually used as a vegetable and for extracting starch, while Pueraria montana var. montana (Willdenow) Maesen & S. M. Almeida ex Sanjappa & Predeep (common name “Gemamu” in Chinese) is confusedly used medicinally and sometimes as a cover crop and fodder [5]. Some taxonomic questions on P. montana var. lobata and the other two varieties are up for debate because of the phenomena of homonym, unscientific classification, genetic relation confusion, and transitional states of highly variable traits. The synonyms or some combinations of P. montana var. lobata in the different studies complicated the species delimitation on account of variation in leaf shape, inflorescence and flower size, and geography by different authors (Table S1), which may have further hindered the classification of P. montana var. lobata and the other two varieties. Several researchers consider that these taxa of P. montana were one species with three varieties [5,6,7,8], while Ohashi et al. [9] considered that they should be treated as two species. Pueraria montana var. lobata and the other two varieties share the following diagnostic characteristics [5]: robust climbers with tuberous roots, dorsifixed stipules entire to fringed at the base, trifoliolate, and flowers clustered at each node of the rachis (Fig. 1, Fig. S1, Table S2). The previous study focusing on the samples located in Taiwan showed that P. lobata (= P. montana var. lobata) and P. montana (= P. montana var. montana) are single species, while P. lobata subsp. thomsonii (Benth.) Ohashi and Tateishi is considered the subspecies of P. lobata (= P. montana var. lobata) based on the leaflets, the relative length between inflorescence and leaves, and the relative length between the wing and keel-petal [9]. The morphology of leaflets from entire to trilobed had no quantitative standard among these three varieties we observed in the field (Fig. S1), as well as the width and the length. Furthermore, the range of the length of the pedicel, bracteole, flower, and calyx overlapped among these taxa [6, 9]. For instance, the length of the pedicel of P. lobata subsp. lobata (= P. montana var. lobata), P. lobata subsp. thomsonii (= P. montana var. thomsonii) and P. montana (= P. montana var. montana) are 2–6 mm, 3–7 mm, and 1.5-4 mm, respectively [9]. Especially, the intermediate morphology of quantitative traits (i.e. the width and length of apex leaflets, the length of pedicel) between P. montana var. thomsonii and P. montana var. lobata reveal that it may be questionable to classification and rank of P. montana merely based on the morphological characteristics (Table S2) [7, 9]. For example, P. montana var. lobata and P. montana var. thomsonii shared the pedicel with white short spreading hairy [9].

Fig. 1
figure 1

Morphological diversity of Pueraria montana var. lobata and the other two varieties in this study. (A)-(B)Pueraria montana var. lobata (YG_127); (C)-(D)Pueraria montana var. thomsonii (FG_43); (E)-(H)Pueraria montana var. montana (GMM_96 and GMM_94). The code of the individual is consistent with Table 1

Full size image

Molecular phylogenies are often incongruent with morphologically-based classification schemes, leading to difficulty in understanding the evolution of morphological traits in many crops involving domestication. Several molecular markers have been used to reconstruct a phylogenetic frame of Pueraria or legume species, but the phylogenetic relationships between P. montana and closely related species are still disputed due to morphological intermediacy and varying molecular markers [7, 10,11,12,13,14,15,16,17,18,19,20]. In previously published works, Pueraria montana var. lobata and the other two varieties formed a clade and nested within Pueraria sensu stricto based on two cpDNA markers, while they formed a polytomy based on one nuclear marker (AS2) [16]. Pueraria montana var. montana should be considered as a single species, and P. montana var. thomsonii sister to P. montana var. lobata based on nrITS sequences involving four accessions of P. montana and its varieties [11]. Pueraria montana var. thomsonii, P. montana var. lobata, and P. montana var. montana clustered within a monophyletic clade but relationships among these taxa are less resolved based on nrITS sequences of 11 accessions of Pueraria [15]. Therefore, insufficiently informative characters and sparse representative accessions in sampling had been the problems for investigating the phylogeny between P. montana var. lobata and closely related taxa.

Full plastome sequencing is useful in resolving complex evolutionary relationships in closely related species due to more informative sites and lower nucleotide substitution rates [21,22,23,24,25]. Moreover, accumulating studies showed that positive selection may be critical for the evolution of genes in plastomes, even if the structure and sequence of plastomes among closely related taxa are conservative [26,27,28,29,30,31,32,33]. Pueraria montana var. lobata is distributed widely and become an invasive plant due to the breeding system, favorable climatic conditions, and human-mediated movement [34, 35]. Given the many important biological processes in plastids, including photosynthesis, the assimilation of sulfur and nitrogen, and phytohormones, it is to be expected that the plastome would be diversified or adapted to a specific environment due to natural selection [36].

Here, we sequenced twenty-six chloroplast genomes for P. montana var. lobata and its allies to explore plastid diversity in the genetic background of kudzu vine that originated from different geographical regions. We aim to (1) establish a phylogenetic framework for P. montana var. lobata and closely related taxa; and (2) compare the structural variation and positive selection of the chloroplast genomes of P. montana var. lobata and closely related taxa.

Results

General features of Plastomes

We sequenced and assembled twenty-six plastomes from P. montana and its allies (Table 1, Table S3). All newly sequenced genomes showed typical quadripartite structure, ranging from 153,360 bp (P. montana var. montana_GMM_112) to 153,551 bp (P. edulis_SYG_78) (Fig. 2), which was similar to previously published complete chloroplast genomes of P. montana var. lobata and P. montana var. thomsonii [18,19,20]. The GC content of these plastomes was 35.4%, except for P. mirifica A. Shaw and Suvat that was 35.5%. We identified 112 genes comprising four ribosomal RNAs, 30 transfer RNAs, and 78 protein-coding genes. Seven protein-coding genes, four rRNAs, and seven tRNAs are duplicated in the inverted repeat regions, and thus each chloroplast genome harbored 130 genes in total (Table S4). The size of the inverted repeat (IR), large single copy (LSC) and small single copy (SSC) regions ranged from 25,632 bp (P. montana var. montana_GMM_112) to 25,684 bp (P. mirifica_TG_79), 84,077 bp (P. mirifica_TG_79) to 84,248 bp (P. edulis_SYG_78), and 17,971 bp (P. mirifica_TG_79) to 18,005 bp (P. montana var. montana_GMM_112), respectively.

Table 1 General features of plastomes newly sequenced in the present study
Full size table
Fig. 2
figure 2

Gene maps of chloroplast genomes of Pueraria accessions in the present study. Genes outside the circles are transcribed in counter clockwise direction and those inside in counterclockwise direction. The colored bars indicate known protein-coding genes, transfer RNA genes, and ribosomal RNA genes. The dashed, dark grey area in the inner circle denotes GC content, and the light grey area indicates genome AT content. LSC indicates large single copy; SSC, indicates small single copy and IR, indicates inverted repeat

Full size image

The mVISTA analysis revealed that the chloroplast genomes of P. montana var. lobata and closely related taxa were conserved generally with a few variable regions (Fig. S2). The proportion of variable sites was higher in the LSC regions than the SSC and IR regions, and higher in the intergenic spacers than the coding regions. We also found there was low variability of adjacent genes in the IRb/SSC and SSC/IRa boundaries of P. montana var. lobata and the other two varieties (Fig. S3). We did not detect any inversions and large indels either within the varieties of P. montana or in comparison to the P. edulis and P. mirifica (Fig. S4). The range of nucleotide diversity was 0-0.00191, with an average of 0.07839. Three genes had a higher π: ndhJ, ycf1 and rps15, and the corresponding noncoding regions trnK-rbcL, trnG-psbZ, psbD-trnT, trnD-psbM, petN-trnC, psbK-trnQ, rps16-accD, psbB-psbT, rps3-rps19, and ccsA-ndhA (Fig. S5).

Plastome phylogenomics

The multiple sequence alignment of all 47 plastomes (including both newly generated and previously published plastomes) was 154,176 bp after excluding one of the inverted repeat regions. The best-fit model for the chloroplast genomes, SSC, IR, LSC, and coding sequences was listed in Table 2. The aligned sequences of the five sequence datasets were little saturated and were thus useful for phylogenetic analyses (Table S5). The topologies estimated from the 47 complete chloroplast sequences (Fig. 3) were broadly similar to those estimated using the other four datasets (Fig. S6-S9), where two major clades of the accessions of P. montana var. lobata and closely related taxa were identified. The monophyly of the Pueraria montana var. lobata clade including all the individuals of P. montana var. lobata and P. montana var. thomsonii was strongly supported in all phylogenetic analyses. All sampled individuals of P. montana var. montana were monophyletic in LSC, SSC and protein-coding sequences dataset and paraphyletic in IR dataset with low support. Thus, we limited the discussion to phylogenetic trees estimated using chloroplast genomes (Fig. 3).

Table 2 Data set characteristics and best substitution models
Full size table
Fig. 3
figure 3

Phylogenetic tree reconstructed based on the 47 complete plastome sequences. Tree topology was based on BI analysis. Numbers at the nodes represent BI posterior probability (PP) / ML bootstrap (MLB) values. PP or MLB values lower than 0.5 or 50% were indicated by hyphens

Full size image

Pueraria mirifica (PP/MLB = 1.00/80) and P. edulis (PP/MLB = 1.00/100) was sister to the varieties of P. montana with strong support, respectively. Pueraria montana var. lobata and the other two varieties are divided into two clades. The Pueraria montana var. montana clade, containing all accessions of P. montana var. montana, showed high support (PP/MLB = 1.00/100). The Pueraria montana var. lobata clade, which includes all accessions of P. montana var. lobata and P. montana var thomsonii, is a monophyletic lineage with strong support (PP/MLB = 1.00/100), but the resolution among the accessions within this clade was relatively low.

Topology testing

Since the tree topologies of the major clades based on the taxon removal experiments showed no changes when outgroups were removed (Fig. S10), we concluded that the tree topology was not affected by long-branch attraction. The incompletely resolved relationships in the network suggested substantial conflicting signals (Fig. 4). All accessions were sorted into respective clades, and the clades were almost distinct. However, the relationships within the Pueraria montana var. montana clade were less clear.

Fig. 4
figure 4

Neighbor-net analyses of Pueraria accessions based on 27 plastome sequences. Clades revealed are labeled with different colors

Full size image

Positive selection

Pueraria montana var. lobata and the other two varieties from different productive areas were divided into two clades. The branch model detected only one gene (ycf2) with a significantly decelerated rate of evolution (ω = 0.62093, p < 0.001). We detected one gene (ndhB, one site) under positive selection at one amino-acid site (p < 0.001, BEB value > 0.95) in the analyses of the branch-site models, taking the Pueraria montana var. lobata clade as the foreground clade. However, no signals of positive selection were detected in the Pueraria montana var. montana clade in either model. Because chloroplast genes tend to be conservative, we compared the fit of the Clade model C (CmC) with Likelihood Ratio Tests (LRT) to detect subtle differences in specific selective constraints between the Pueraria montana var. lobata clade and Pueraria montana var. montana clade. Our CmC analysis showed that accessions of the Pueraria montana var. lobata clade exhibited significant relaxation in six genes (accD, ndhB, ndhC, rpl2, rpoC2, rps2), compared to accessions of the Pueraria montana var. montana clade (Table S6). 26 amino acid residues were identified as having a sign of positive selection in the analyses of the Site model, which predicted that 36 sites greater than 0.95 and 33 sites greater than 0.99 (Table S7) of all legume taxa.

Discussion

Phylogenetic inference

Previous studies revealed that Pueraria s.l. is a polyphyletic group since several lineages could be treated as five distinct genera, namely Pueraria s.s., Teyleria stricta (Kurz) (A) N. Egan & (B) Pan, Neustanthus Bentham, Haymondia A.N. Egan & B. Pan, and Toxicopueraria A.N. Egan & B. Pan based on several key morphological and phylogenetic studies. The remaining species include Pueraria tuberosa (Roxb. ex Willd.) DC. grouped into Pueraria s.s. [7, 10,11,12,13,14,15,16]. The infrageneric classification of the Pueraria s.s. is still largely in dispute, and it is an intricate challenge for taxonomic species, subspecies, or varieties as many synonyms are reported for a single species. For example, Pueraria edulis is treated as the kudzu vine in the southwestern region of China such as Sichuan and Yunnan provinces [37]. In previous works, the phylogenetic tree based on two cpDNA markers or one nuclear maker (AS2) showed that P. montana var. lobata and the other two varieties formed a subclade and nested within Pueraria s.s. with other closely related Pueraria species [16]. Our plastomic data support the previous findings that Haymondia wallichii diverges from Pueraria taxa and demonstrate P. edulis and P. mirifica are separated from three varieties of P. montana, then further revealing that P. montana var. lobata and the other two varieties were divided into two clades, of which one was comprised of P. montana var. lobata and P. montana var. thomsonii, and the other clade included all sampled P. montana var. montana.

For the most part, our phylogenetic results revealed that the Pueraria montana var. montana clade is resolved as a monophyletic group with strong support except for phylogenetic analysis based on IR regions. The phylogenetic analysis based on nrITS sequences showed that P. montana var. montana formed a monophyletic clade sister to the other two varieties, which suggested that P. montana var. montana should be treated as a single species [11, 15, 17]. Pueraria montana var. lobata and P. montana var. thomsonii were different from P. montana var. montana in aspects of the morphology of leaflets, the relative length between inflorescence and leaves, the length and hairiness of the pedicel, the relative length between the wing and keel-petal (Table S2) [9, 10]. Furthermore, the composition and abundance of secondary metabolites also showed significant differences in P. montana var. montana compared to P. montana var. lobata and P. montana var. thomsonii [38].

All our phylogenetic analyses support the sister relationship of P. montana var. lobata and P. montana var. thomsonii with strong support but the relationship among accessions are less resolved. The uncertain association between P. montana var. lobata and P. montana var. thomsonii was also revealed in previous studies [9, 10]. Although these two varieties were having high medicinal values, there are considerable differences in terms of the metabolites, material basis, and efficacy [38,39,40,41]. Morphological intermediacy, insufficiently informative characters, artificial domestication, species complex and putative hybridization/introgression contribute to the lack of agreement between morphological-based classification schemes and recent molecular phylogenies among P. montana var. lobata and closely related taxa [11, 15,16,17]. Further studies of Pueraria species should focus on other markers to achieve better phylogenetic resolution, such as RAD-seq or transcriptome.

Conservation and adaptation of kudzu plastomes

Our results reveal no obvious expansion or contraction of IR region, indels and inversion in the chloroplast genome between P. montana var. lobata and closely related taxa, even compared to the outgroup species of P. mirifica. The results of the mVISTA analysis showed variable intergenic regions in the LSC region. Lower sequence diversity existed in the SSC and IR regions of 24 newly sequenced three varieties of P. montana. The nine noncoding regions (trnK-rbcL, trnG-psbZ, psbD-trnT, trnD-psbM, petN-trnC, psbK-trnQ, rps16-accD, psbB-psbT, and rps3-rps19) and one gene (ndhJ) were detected as relatively divergent in the LSC region, while two genes (rps15, ycf1) and one noncoding region (ccsA-ndhA) were detected in SSC region. These variable regions identified in the present study could be potential markers for phylogenetic or population genetics studies in P. montana var. lobata and its related taxa.

Pueraria species are delineated as climbing vines. Pueraria montana var. lobata (kudzu) inhabits temperate environments or higher altitudes in the tropics [7] and prefers to occur in high-light forest edge areas, even though it can live under both shade and sun [34]. Thus, Pueraria montana var. lobata has been widely found in its native area, and probably introduced elsewhere [7]. The ecology, climatic considerations, growth and establishment pattern, cultivation and genetic divergence may influence the distribution of P. montana var. lobata, even other Pueraria plants [42]. Interestingly, neither similar environment nor genetics seems to be a major factor in the distribution of kudzu, as is the case in the United States and Switzerland [42]. The ecology and genetics of P. montana var. lobata in Southern Switzerland is similar to that in the South-Eastern USA [42]. Pueraria montana var. lobata is not expected to spread on a large scale in Southern Switzerland [42, 43], while it was officially listed as a Federal Noxious Weed due to widespread ecological and economic impacts in the southeastern USA [34, 44]. For example, its invasion has the potential to raise ozone pollution in the southeastern United States [45]. To clarify the adaptation or differentiation of P. montana var. lobata and its closely related taxa, our data provide novel evidence of positive selection of genes under four models from the perspective of the chloroplast genome, as well as in Panax [27], and Rhodiola [31].

The clade model detected the variation in accD, ndhB, ndhC, rpl2, rpoC2, rps2 as being significantly different from the accessions of the Pueraria montana var. montana clade. The plastid accD gene encodes a key enzyme for fatty acid biosynthesis, namely a beta subunit of acetyl-CoA carboxylase, which is important for leaf development [46,47,48]. The ndh genes in plastomes encode a protein act by adjusting the electron transfer from NADH to photosystem I [31, 48]. The gene rpl2 encodes the ribosomal protein L2 of 80 S ribosomes, which is critical for the peptidyl transferase activity [49]. The gene rpoC2 encodes RNA polymerase subunits and the gene rps2 encodes the ribosomal protein, but it is unclear their specific function in the evolution of the plastome of Pueraria.

The site model detected 26 amino acid residues across all sampled species. Using the Pueraria montana var. lobata clade as the foreground, the branch model revealed only ycf2 may have undergone negative selection. ycf2, as the largest coding sequence in plastome and conserved open reading frames, may exhibit similarities with fstH, such as activity associated with cell division, ATPase, and chaperone function [50, 51]. We also detected one possible positively selected gene (ndhB) using the branch-site model, which may be involved in photosynthesis and temperature sensitivity [52, 53]. However, taking the Pueraria montana var. montana clade as the foreground, both the branch and branch-site models failed to detect the positive genes.

The combination of key traits such as relatively high photosynthetic rates, rapid leaf movements, and the ability to fix N2 make P. montana var. lobata an aggressive competitor [34]. Thus, our results propose that the positive selection of these genes, involving leaf development or photosynthesis, may play an important role in adaptation to different habitats in kudzu or other kudzu species, which may influence their plastid differentiation pattern. This adaptation of kudzu and its related taxa occurred when their common ancestor diverged from other Pueraria species.

Conclusions

In this study, we reported the chloroplast genomes for 26 Pueraria accessions. Through the comparisons of these newly sequenced chloroplast genomes to additional sequences from species of legume species in NCBI, we found that gene content and order of P. montana var. lobata and its closely related taxa are highly conserved. The phylogenetic results based on 47 plastomes, LSC, SSC, IR, and protein-coding sequences supported P. edulis and P. mirifica was sister to remaining Pueraria accessions respectively. The remaining Pueraria accessions were divided into two clades but could not resolve the further division within these two clades. We also detected several genes that may play the important role in the evolution of these two clades. Overall, these results demonstrated the power of chloroplast genomes in solving the phylogenetic relationship, as well as the variation of chloroplast genomes between P. montana var. lobata and its closely related species. Future work with more taxa sampling and markers will help to illustrate gene introgression, ILS, and other biological factors that may play various roles among lineages of P. montana var. lobata and P. montana var. montana.

Methods

Taxon sampling, DNA extraction and sequencing

Three to five fresh leaf samples from each individual were collected in their native habitats and dried in silica gel and stored at room temperature for DNA extraction. Then we selected these accessions from representative productive areas. Twenty-six accessions, including P. montana var. lobata and the other two varieties, were selected for Illumina sequencing. Pueraria mirifica Airy Shaw & Suvat. and Pueraria edulis Pampan. were also sequenced as outgroups. All vouchers were deposited in the Guangxi Academy of Agricultural Sciences (GAAS, Nanning, China) (Table 1). The samples of Pueraria accessions were identified by Hua-Bing Yan and Xiao-Hong Shang in the field. Silica-dried leaf material was sent to the Novogene Bioinformatics Technology Co. Ltd. (Tianjin, China) for library preparation and Illumina sequencing on NovaSeq 6000. A paired-end library (150 × 2) was constructed with an insert size of 350 base pairs (bp). Total genomic DNA was isolated using a DNeasy Plant Mini Kit (Qiagen, CA, United State).

Sequence assembly and annotation

The raw data were first filtered using fastp to remove those paired reads that contained more than 10% of N bases, and more than 50% of bases with low Phred quality (Q ≤ 5), and adapter sequences. The clean data were mapping assembly to create initial reference sequences using MIRA v4.0.2 [54]. These newly created reference sequences were assembled as the complete plastome sequences using MITObim v1.8 [55]. The previously published complete plastome sequences of P. montana var. lobata (GenBank nos., MT818508 and MN180247) and P. montana var. thomsonii (GenBank no., MN515038) were used as seed references. These assemblies were manually remapped and inspected using Geneious v. R9.0.2 [56] under default parameters to check for misassemblies. The tRNAscan-SE ver. 1.21 software [57] was applied to verify the tRNA genes with default parameters. Gene annotations were assigned using DOGMA [58]. Inverted repeat boundaries were determined using the BLAST method [59] (https://blast.ncbi.nlm.nih.gov/Blast.cgi) and verified using Geneious v. R9.0.2. The circular maps of newly sequenced chloroplast plastomes were drawn using the online program OGDRAW v. 1.2 [60].

Identification of the hypervariable regions

To better identify possible rearrangements, five newly sequenced chloroplast genomes of P. montana var. lobata and closely related taxa were compared using the global alignment algorithm Shuffle-LAGAN [61] with mVISTA (https://genome.lbl.gov/vista/mvista/submit.shtml), with plastome of P. mirifica was chosen as the reference genomes to detect the gene content and structural variation between P. montana var. lobata and the other two varieties. To further identify specific patterns of structural variation at the species level, five chloroplast genomes were used as the representatives to align with default parameters using MAUVE plugin [62] in Geneious v R9.0.2. We also detected the contraction and expansion of the boundaries between SC and IR regions using IRscope [63]. The nucleotide diversity (π) analysis of 24 newly sequenced plastomes of three varieties of P. montana was implemented using DnaSP v. 6.12.03 [64] with window length set to 500 bp, and the step size set as 200 bp.

Phylogenetic analyses

To better determine the phylogenetic position of P. montana var. lobata and other varieties, we reconstructed a phylogeny using 47 chloroplast genomes, including the representatives of the major subclades of legume species referring to the previous studies [16, 18,19,20]. Twenty-one previously published plastomes were downloaded from the NCBI GenBank database (https://www.ncbi.nlm.nih.gov/) (Table S8). The phylogenetic analyses were conducted based on the following five datasets: (1) the complete chloroplast genome sequences that excluded one copy of the inverted repeat (IR) region; (2) the large single-copy region (LSC); (3) the small single-copy region (SSC); (4) one inverted repeats region (IR), and (5) 73 protein-coding genes excluding genes infA, rpl33, rps16, ycf1, ycf4 that were lost in several outgroups (Table S8). These regions were aligned in Geneious v R9.0.2 using Mauve progressive using default options. DAMBE v6.4.29 was used to assess substitution saturation [65, 66]. Phylogenetic analyses were inferred using both maximum likelihood (ML) and Bayesian inference (BI) via the CIPRES Science Gateway server (http://www.phylo.org/) [67]. ML analysis using rapid bootstrapping and 1000 replications was performed in RAxML [68] with the GTR-GAMMA model. The appropriate model for each dataset was determined using jModelTest v2.1.4 [69]. For the BI analysis, two independent Markov chain Monte Carlo (MCMC) runs with three heated and one cold chain were executed using MrBayes v3.2.7 [70]. Trees were sampled every 1000 generations. The first 25% of trees were discarded as burn-in and the remaining samples were then summarized and a majority-rule consensus tree was constructed. Two million generations were run for the IR dataset and data-complete in the taxon removal experiments dataset, whereas three million generations were run for the other four datasets (Table 2). The stationarity was considered to be reached when the average standard deviation of split frequencies fell below 0.01. We also determined the convergence of runs using Tracer v 1.6 [71] to assess the effective sample size (ESS) > 200 for all parameters.

Topology hypothesis testing

Taxon removal experiments can aid in the investigation of whether distantly related outgroups have a biased attraction to long branches within the ingroups [72,73,74]. To test if outgroup selection affected the topology of the ingroups, we recovered phylogenetic trees by removing outgroups to include only 30 accessions of P. montana var. lobata and its closely related taxa. Haymondia wallichii (DC.) (A) N. Egan & (B) Pan bis was chosen as the outgroup based on similar morphology and relatedly phylogenetic relationship with Pueraria taxa in our analysis. To better visualize and evaluate the phylogenetic signal conflicts between ingroup taxa, the neighbor-net algorithm based on uncorrected P-distances was implemented in SplitsTree v4.14.4 [75], excluding all outgroup taxa to leave P. montana var. lobata and the other two varieties.

Positive selection analysis

We chose the ML tree based on 47 chloroplast genomes to identify positively selected genes in the EasyCodeML [76] with the likelihood ratio test (LRT) since the topology of ML trees and BI trees were congruent. Because the genes infA, rpl33, rps16, ycf1 and ycf4 were lost in several legume species (Table S8), we selected 73 protein-coding genes for further analysis. The amino acid sequences of the 73 protein-coding genes were aligned using MAFFT and converted nucleotide alignments into PAML format in EasyCodeML according to the manual [76]. Four different models (i.e. site model, branch model, branch-site model, and clade model) were used to test the selection pressure. We set two clades to detect positive evolution at the clade level: one is the Pueraria montana var. lobata clade and the other is the Pueraria montana var. montana clade using the other three model selections, namely branch model, branch-site model and clade model. Firstly, we used a branch model to identify rapidly evolving genes. Genes with p < 0.05 and ω > 1 are considered under positive selection, while others with p < 0.05 and 0 < ω < 1 are considered under purifying selection. The branch-site models were combined with heterogeneous ω across sites and branches to find instances of neutral evolution and negative selection. The clade model was performed to identify subtler differences in site-specific selective constraint among partitions of a phylogeny [77].