Background

Seseli L. is one of the largest genera of Apiaceae Lindl. [1] and consists of 125–140 species. Seseli species are distributed in the Old World from western Europe and northwestern Africa to China and Japan [1,2,3,4]. Nineteen Seseli species are distributed in China with nine of them endemic [5].

As one of the largest genera of Apiaceae, the taxonomy of Seseli has been controversial so far. One of the prominent taxonomic problems is the inclusion of Libanotis Haller ex Zinn and Eriocycla Lindl. within Seseli. For Libanotis, Drude regarded Libanotis as one of the four subgenera of Seseli [6]. Then, Pimenov and Sdobnina classified Libanotis distributed in Russia into different groups within Seseli [7]. Pimenov summarized previous studies and type specimens, and concluded that most Chinese Libanotis species were synonyms of Seseli species [e.g., Libanotis buchtormensis (Fisch.) DC. \(\equiv\) Seseli buchtormense (Spreng.) W. D. J. Koch, Libanotis montana Crantz \(\equiv\) Seseli libanotis (L.) W. D. J. Koch] [8]. In the latest research, Duran et al. [9] found that species of Libanotis form a clade, but this clade was falling into polytomy with other Seseli species. However, Schischkin advocated the generic status of Libanotis according to its different morphological characteristics (e.g., conspicuous calyx teeth, bracts numerous, the separation of bracteoles, and almost always pubescent fruits) [10]. For Eriocycla, Kljuykov proposed that the type species of Eriocycla (Eriocycla nuda Lindl.) with its related species and several Seseli species should be divided into a new section of Seseli as there were no substantial carpological differences between Eriocycla and Seseli and were similar in non-carpological characteristics [11]. Moreover, Pimenov treated all Eriocycla species distributed in China as synonyms of Seseli species [e.g., Eriocycla nuda \(\equiv\) Seseli nudum (Lindl.) Pimenov et Kljuykov, Eriocycla pelliotii (H. Boissieu) H. Wolff \(\equiv\) Seseli pelliotii (H. Boissieu) Pimenov et Kljuykov] [8]. However, Degtjareva et al. proposed that several Seseli species [e.g., S. delavayi Franch., S. afghanicum Pimenov] should be transferred into Eriocycla [12]. In addition, taxonomic boundaries were uncertain between Seseli and Ligusticum mucronatum (Schrenk ex Fisch. & C.A. Mey.) Leute. For instance, Pimenov treated Ligusticum mucronatum and Ligusticum thomsonii C. B. Clarke as synonyms of Seseli mucronatum (Schrenk) Pimenov et Sdobnina [8]. Most of the above-mentioned taxonomic treatments were based on limited morphological data, but this was insufficient to define the boundaries between Seseli and its related genera. For example, fruit structures of the type species of Seseli (S. tortuosum) and Libanotis (Libanotis montana Crantz) were almost identical [13], and different types of fruit structures existed among Seseli species [14]. Therefore, combining more abundant morphological characteristics is critical to resolving the taxonomy of Seseli.

An ideal genus should be monophyletic and clearly defined based on morphology [15]. Seseli is not a monophyletic group, which is the same as other large genera of Apiaceae (e.g., Angelica L., Ligusticum L., Peucedanum L.) and is one of the most taxonomically complicated genera within Apiaceae [16,17,18]. Previous studies have used several molecular fragments (e.g., ITS, rps16 intron, rpl16 intron) to show that members belonging to Seseli are distributed into three tribes: Selineae (including the majority of Seseli species), Pimpinelleae [S. diffusum (Roxb. ex Sm.) Santapu & Wagh] and Apieae (S. webbii Coss.) [15, 17, 19,20,21]. However, these molecular fragments contained too few informative sites. Consequently, the results of these molecular studies showed low support and resolution and were insufficient to resolve the phylogeny of Seseli. Therefore, it is urgent to use extensive sampling and abundant molecular data to reconstruct a more robust phylogeny of Seseli.

Seseli is an important genus with a high number of aromatic species used as traditional medicine due to their richness in coumarins, terpenoids, and essential oils. They have many important pharmacological activities such as reduction of inflammation, swelling, rheumatism, pain, and minimization of the common cold’s symptoms [22,23,24]. Five Seseli taxa (S. mairei H. Wolff, S. mairei var. simplicifolia C. Y. Wu ex R. H. Shan & M. L. Sheh, S. yunnanense Franch., S. delavayi Franch., and S. squarrulosum Shan & M.L. Sheh) are used as traditional Chinese medicine “Fang feng” [5]. The identification of medicinal materials is almost entirely based on morphological characteristics and traditional recognition. However, medicinal materials are used indiscriminately given the considerable intraspecific morphological variations of Seseli species. Hence, it is necessary to define species boundaries and develop more molecular markers to ensure the correct identification and usage of medicinal Seseli species.

The length of a plastome is usually 115 kb-165 kb. The plastome’s typical quadripartite structure is composed of a pair of inverted repeat (IR) regions of 22-25 kb separating the large single-copy (LSC) region of 82-90 kb and the small single-copy (SSC) region of 15-20 kb [25]. Plastome of angiosperms has the advantages of low nucleotide substitution rates and without gene recombination [26]. Plastomes have been widely used in studying the phylogeny of Apiaceae, Ranunculaceae Juss., Saxifragaceae Juss., Allium L., Liliaceae Juss., etc. [27,28,29,30,31,32,33,34,35,36,37]. However, there has been no study on the phylogeny of Seseli based on plastomes and therefore there is ample opportunity to investigate phylogenetic and taxonomic issues of Seseli using plastomes.

Plastome is valuable for phylogenetic studies but should be combined with morphological characteristics (especially carpological characteristics) given the significance of morphology in the taxonomy and evolution of Apiaceae. For example, Wen et al. and Li et al. have combined molecular phylogenetic analyses with carpological characteristics to obtain relatively reliable results [29, 38]. Thus, we use plastomes and morphological data to explore the phylogeny of Seseli. Here, our aims were to: (1) investigate the plastome features and evolution of Seseli; (2) develop appropriate mutation hotspot regions as candidate DNA barcodes for species identification of Seseli; (3) test the ability of plastomes to study the phylogeny of Seseli; and (4) explore the effectiveness of comparative plastome analyses and morphological data for studying the phylogeny and taxonomy of Seseli. Overall, our study can provide new insights into the phylogenetic relationships and taxonomic framework of Seseli.

Results

Features of the eleven Seseli plastomes

The size of the eleven Seseli plastomes ranged from 144,957 bp in S. mairei var. simplicifolia to 155,617 bp in S. eriocephalum (Pall. ex Spreng.) Schischk. All eleven plastomes shared the typical quadripartite structure consisting of a pair of IRs (17,473-26,992 bp) separating the large single copy (LSC) region (84,243-92,935) and the small single copy (SSC) region (16,501-17,698 bp) (Fig. 1, Table 1). The overall GC content ranged from 37.4% to 37.6%, while the IR regions were 42.5-44.9% and much higher than the LSC regions (35.6-36.0%) and SSC regions (30.8-31.2%). In addition, the eleven plastomes contained 128-134 genes, including 84-89 protein-coding genes, 36-37 transfer RNA (tRNA) genes, and eight ribosomal RNA (rRNA) genes (Table 1). No gene rearrangement or loss was found in the eleven Seseli plastomes (Fig. 2, Table S3).

Fig. 1
figure 1

Gene map of the eleven Seseli plastomes. Genes shown outside and inside the outer layer circle are transcribed in the clockwise and counterclockwise, respectively. Different functional groups of genes are marked with different colors. The dark gray area of the inner circle represents the GC content of the plastome. IR, inverted repeat, LSC, large single copy, SSC, small single copy

Full size image
Table 1 Features of the eleven Seseli plastomes
Full size table
Fig. 2
figure 2

Mauve alignment of the eleven Seseli plastomes. Local collinear blocks within each alignment are represented by blocks of the same color connected with lines

Full size image

Repeat sequence and nucleotide diversity analyses

In the eleven Seseli plastomes, we found 501 repeats of four types, with the number of repeats varying from 36 to 49 between species (Fig. 3A, Table S4). The most abundant repeats were forward repeats (256), followed by palindromic repeats (229), reverse repeats (13), and the least were complementary repeats (3), which only appeared in three taxa (S. glabratum Willd. ex Schult., S. mairei var. simplicifolia, and S. valentinae Popov.) (Fig. 3A). Most of the repeats were distributed in intergenic or intron regions (e.g., ycf2-trnL-CAA, trnL-CAA-trnH-GUG, ycf3 intron, ndhA intron). However, protein-coding genes also contained a few repeats, among which the ycf2 gene contained the most, with 29, 24, 21 repeats in S. eriocephalum, S. delavayi, and S. incisodentatum K. T. Fu, respectively (Table S4).

Fig. 3
figure 3

Analyses of repeats in the eleven Seseli plastomes. (A) Total number of four repeat types, (B) Total number of SSRs

Full size image

We discovered 868 SSRs and the number of SSRs differed between the eleven Seseli plastomes, with S. delavayi having the fewest (66 SSRs) and S. glabratum having the most (90 SSRs). The number of Mono-, Di, Tri-, Tetra-, Penta- and Hexa- SSRs were 520, 189, 25, 103, 18, and 13, respectively (Fig. 3B, Table S5). Most of the SSRs were situated in LSC regions and intergenic spacers. However, the protein-coding gene ycf1 contained the most abundant SSRs, with 48, and ccsA contained 20 SSRs (Table S5).

We calculated the nucleotide diversity (Pi) of 63 protein-coding genes, 81 non-coding regions and introns in LSC, SSC, and IR (Fig. 4, Table S6). The Pi values of the gene regions ranged from 0 (psbE gene) to 0.01409 (matK gene) with a mean value of 0.00428, and the Pi values of the non-coding regions and introns ranged from 0.00077 (trnI-GAU intron) to 0.08368 (trnH-GUG-psbA) with a mean value of 0.01325 (Table S6). Two protein-coding genes (matK and ccsA) with relatively high nucleotide diversity (Pi > 0.01) were detected, while ten non-coding regions and introns with high nucleotide diversity (Pi > 0.02) were detected, namely trnH-GUG-psbA, ycf2-trnL-CAA, trnG-UCC-trnR-UCU, psbA-trnK-UUU, psbK-psbI, petA-psbJ, rps2-rpoC2, ndhC-trnV-UAC, rpl32-trnL-UAG and cemA-petA (Fig. 4, Table S6). These mutation hotspot regions were selected as candidate DNA barcodes.

Fig. 4
figure 4

Comparative analysis of the nucleotide diversity (Pi) values among the eleven Seseli plastomes: (A) protein-coding genes, (B) non-coding and intron regions

Full size image

Phylogenetic analyses

Thirty-seven plastome CDS and 59 nrDNA sequences, the two supermatrices, were used to reconstruct the phylogeny of Seseli, respectively. We found several incongruences in topologies between CDS-based and nrDNA-based phylogenetic trees. Nevertheless, the topologies indicated that the Seseli taxa fell into two tribes (Selineae and Echinophoreae) and were not clustered as a monophyletic group (Fig. 5, Fig. S1). Two types of support values: Bayesian inference (BI) posterior probabilities (PP) and ML bootstrap values (BS) were shown on the phylogenetic trees.

Fig. 5
figure 5

The plastome CDS-based phylogenetic tree constructed by Bayesian inference (BI) and maximum likelihood (ML) with the posterior probabilities of BI and the bootstrap values of ML above the branches, respectively, (*) represents maximum support in both two analyses

Full size image

For CDS-based phylogenetic trees, the ML and BI trees were highly consistent in topology (Fig. 5). S. delavayi clustered with Eriocycla nuda, belonging to Echinophoreae, with strong support (PP = 1.00, BS = 100) and it was far from the main branches of Seseli. Whereas other Seseli taxa were members of Selineae (Fig. 5). In Selineae, the Seseli taxa failed to cluster in a clade, but were divided into five subclades (Fig. 5). S. glabratum clustered with Libanotis buchtormensis, Libanotis spodotrichoma K. T. Fu, and Saposhnikovia divaricata (Turcz. ex Ledeb.) Schischk. with relatively high support in BI (PP = 1.00) but the support in ML was rather weak (BS = 62), belonging to Subclade A. S. intramongolicum Ma formed a single Subclade B and diverged from Subclade A + Peucedanum subclade [Peucedanum harry-smithii var. grande (K.T.Fu) Shan et Sheh + Peucedanum praeruptorum Dunn + Peucedanum ampliatum K.T.Fu] with high support (PP = 1.00, BS = 98). S. yunnanense, S. mairei, S. mairei var. simplicifolia, S. squarrulosum, and S. incisodentatum formed a robust Subclade C (PP = 1.00, BS = 99), in which S. yunnanense + S. mairei, clustered with S. mairei var. simplicifolia with strong support (PP = 1.00, BS = 100), and S. squarrulosum and S. incisodentatum clustered together (PP = 1.00, BS = 100). S. valentinae and Seseli montanum L. formed a robust Subclade D (PP = 1.00, BS = 100). Seseli coronatum Ledeb. and S. eriocephalum were in Subclade E, in which S. coronatum was more closely related to the Ligusticum subclade (Ligusticum thomsonii + Ligusticum mucronatum) but with moderate support (PP = 0.81, BS = 63), and S. eriocephalum was located at the base of Subclade E with strong support (PP = 1.00, BS = 100).

The concatenated nrDNA dataset included 984 aligned characters. The ML and BI trees were nearly identical in topology, and several branches with quite low support (< 50% bootstrap support) were treated as parallel branches (Fig. S1). S. delavayi was also far from the other Seseli taxa and clustered with Eriocycla nuda within the tribe Echinophoreae with strong support (PP = 1.00, BS = 100). The other Seseli taxa belonged to Selineae. Nine Seseli taxa, being S. arenarium M.Bieb., S. hartvigii Parolly & Nordt, S. andronakii Woronow ex Schischk., S. grandivittatum (Sommier & Levier) Schischk., S. serpentinum B.L.Burtt ex H.Duman & E.Doğan, S. alexeenkoi Lipsky, S. globiferum Vis., S. leptocladum Woronow, and S. tortuosum formed the robust “S. tortuosum” clade (PP = 1.00, BS = 95). Eight Seseli taxa, being S. corymbosum Boiss. & Heldr., S. paphlagonicum Pimenov & Kljuykov, S. lehmannii Degen, S. ponticum Lipsky, S. rupicola Woronow, S. resinosum Freyn & Sint., S. dichotomum Pall. ex M.Bieb., and S. gummiferum Pall. ex Sm. formed the robust “S. gummiferum” clade with high support in BI (PP = 1.00) but weak support in ML (BS = 67). Five Peucedanum taxa and Saposhnikovia divaricata formed a clade (PP = 1.00, BS = 64). Then, these three clades clustered together with weak support (PP = 0.72, BS = 31). S. yunnanense, S. mairei, S. mairei var. simplicifolia, S. squarrulosum, and S. incisodentatum formed a robust clade (PP = 1.00, BS = 98). S. coronatum clustered with S. glabratum (PP = 0.83, BS = 62). S. eriocephalum formed a single branch. S. transcaucasicum (Schischk.) Pimenov & Sdobnina clustered with Libanotis sibirica (L.) C. A. Mey. (PP = 1.00, BS = 100), while S. intramongolicum clustered with S. marashicum E.Doğan & H.Duman (PP = 1.00, BS = 100), then, the two subclades clustered together. S. valentinae and Ligusticum mucronatum formed a clade (PP = 0.86, BS = 58).

Comparative plastome analyses

We investigated the plastome structural differences, mainly focusing on the borders of LSC/IRb (JLB), IRb/SSC (JSB), SSC/IRa (JSA), and IRa/LSC (JLA) (Fig. 6). Taxa within Subclade A had relatively similar structures, in which S. glabratum had the shortest SSC region (16,501 bp) and the longest ycf1 gene extended into the IRa region (3,041 bp) among the eighteen plastomes. Subclade B: S. intramongolicum had the longest part of ycf2 gene extended into the IRb region (3,712 bp) and an extreme distance of trnL-CAA gene from the JSA (4,905 bp) among the eighteen plastomes. Subclade C: S. yunnanense and S. mairei had highly similar structures, while S. mairei var. simplicifolia had 63 bp between the ycf2 gene and the JLB. S. squarrulosum and S. incisodentatum had similar structures: the JLB was within the rpl23 gene and the JLA was located between the trnI-CAU and trnH-GUG gene. Subclade D: S. valentinae and S. montanum had relatively similar structures, in which S. valentinae had the farthest distance between the trnH-GUG gene and the JLA (1,032 bp) among the eighteen plastomes. Subclade E: Ligusticum thomsonii and Ligusticum mucronatum had similar structures, the ycf2 gene of S. coronatum was distant from the JLB (446 bp), while the JLB of S. eriocephalum was within the rps3 gene and the JLA located between the rpl22 and trnH-GUG gene. Echinophoreae: S. delavayi and Eriocycla nuda had the JLB within the rpl2 gene and the JLA located between the rpl23 and trnH-GUG gene (Fig. 6).

Fig. 6
figure 6

Comparison of the borders of LSC, SSC, and IR regions among the eighteen plastomes. Different boxes for genes represent the gene position

Full size image

The eighteen plastome sequence divergence analysis suggested that the coding regions were more conserved than the non-coding regions, and the IR regions were more conserved than the single copy regions (Fig. 7). Taxa within different subclades varied widely in some regions (e.g., rps16-trnQ-UUG, rpoB-trnC-GCA, petA-psbJ, rpl32-trnL-UAG, ycf1, ycf2), while the sequence divergence of taxa within the same subclade was relatively low.

Fig. 7
figure 7

mVISTA-based sequence identity plots for the eighteen plastomes with S. yunnanense as the reference

Full size image

The length of the 80 common CDS of the eighteen plastomes ranged from 68,148 bp in Libanotis buchtormensis to 68,313 bp in S. eriocephalum (Table S7). Codons encoding leucine (Leu) had the most (2,399–2,426), while codons encoding cysteine (Cys) had the fewest (235–243). The RSCU values of all codons ranged from 0.34 to 2.00 (Fig. 8, Table S7). The heatmap showed that 30 types of codons were used more frequently (i.e., RSCU value > 1) and ended with a purine (A/U) except for UUG. Considering the three types of terminator codons (UGA, UAG, and UAA), taxa that belonged to Subclade C (S. yunnanense, S. mairei, S. mairei var. simplicifolia, S. squarrulosum, and S. incisodentatum) had lower RSCU values of UAG (RSCU = 0.71) and higher values of UAA (RSCU = 1.65) than others. Overall, usage of codons showed bias in different subclades, and similarity in different taxa within the same subclade (Fig. 8, Table S7).

Fig. 8
figure 8

The RSCU values of 80 protein-coding genes for the eighteen plastomes. (*) to denote the terminator codons

Full size image

Mericarp morphology

We mapped the mericarps of the eleven Seseli taxa to the two phylogenetic trees (Fig. 9). S. glabratum in Subclade A, had mericarp elliptic or narrowly ovoid, slightly dorsally compressed, finely papillose or slightly scabrous, sometimes subglabrous, endosperm flat on commissural side, calyx teeth obsolete, ribs equal, prominent, filiform or shortly keeled, and vittae 1 in each furrow, 2 on commissure. S. intramongolicum in Subclade B, had mericarp oblong, slightly compressed dorsally to laterally, densely papillose-pubescent, endosperm flat on commissural side, calyx teeth small and triangular, all ribs equal, keeled and filiform, and vittae 1 in each furrow, 2 on commissure. Five taxa in Subclade C, had mericarps ovoid to elliptic, compressed dorsally, glabrous, endosperm flat on commissural side, and numerous vittae in commissure (2–10) and each furrow (1–4, not solitary). However, these five taxa were different in their ribs and calyx teeth: S. yunnanense, S. mairei, and S. mairei var. simplicifolia had ribs narrowly keeled or rounded, while S. squarrulosum and S. incisodentatum had median and lateral ribs keeled and filiform, marginal ribs winged. The first three taxa had calyx teeth obsolete, while S. squarrulosum had small calyx teeth and S. incisodentatum had broadly triangular calyx teeth. S. valentinae in Subclade D, had mericarp ovoid or oblong-ovoid, dorsally compressed, densely puberulent, endosperm flat on commissural side, calyx teeth small and triangular, all ribs equal, obtuse-keeled, and vittae 1 in each furrow, 2 on commissure. For Subclade E, S. coronatum had mericarp oblong, compressed dorsally, sparsely puberulent, endosperm flat on commissural side, calyx teeth obsolete, median and lateral ribs keeled and filiform, marginal ribs winged, and numerous vittae in commissure (8–12) and each furrow (3–5). S. eriocephalum had mericarp oblong, slightly dorsally compressed, densely tomentose, calyx teeth obsolete, ribs thick, marginal ribs slightly winged, and vittae 1 in each furrow, 2 on commissure. S. delavayi in Echinophoreae, had mericarps ovoid, slightly dorsally compressed, densely white hispid, endosperm slightly concave on commissural side, calyx teeth present and triangular, all ribs equal, keeled and filiform, hidden by indumentum, and vittae 1 in each furrow, 2 on commissure.

Fig. 9
figure 9

Mapping carpological characteristics to the two phylogenetic trees, different colors indicating correspondence. Scale bars: dorsal side views = 0.5 mm, transverse sections = 0.5 mm. (A) CDS-based phylogenetic tree, (B) nrDNA-based phylogenetic tree

Full size image

Discussion

In this study, we newly generated eleven plastomes to investigate the plastome features and evolution of Seseli. The genome length, gene numbers, IR/SC borders, and repeat composition were variable, which further implied the complexity of plastomes evolution and the non-monophyly of Seseli. Then, we reconstructed the phylogeny of Seseli based on plastomes and nrDNA sequences. Both CDS-based and nrDNA-based phylogenetic trees indicated that the Seseli taxa did not form a monophyletic group, which was consistent with previous studies using molecular fragments [15,16,17,18,19,20,21]. The eleven newly sequenced Seseli taxa did not cluster with S. tortuosum, in which S. delavayi clustered with Eriocycla belonging to Echinophoreae and the others belonging to Selineae. The comparative plastome analyses and morphological characteristics confirmed the reliability of the phylogenetic analyses and implied the complex evolution of Seseli. We suggest that “a narrow sense” of Seseli is meaningful for further study and the current taxonomic system of Seseli needs to be revised.

Plastome features and evolution

We observed that the genome structure (the typically quadripartite structure) and GC content of these eleven plastomes were conserved, and there was no gene rearrangement or loss among these eleven plastomes. These conservations are commonly in other genera of Apiaceae [28, 29], which might be related to stable plastome function. However, the genome length (144,957 bp-155,617 bp), gene numbers (128–134), IR/SC borders, and repeat composition were variable, which might imply the complexity of plastomes evolution and the non-monophyly of Seseli. In addition, twelve mutation hotspot regions (matK gene, ccsA gene, trnH-GUG-psbA, ycf2-trnL-CAA, trnG-UCC-trnR-UCU, psbA-trnK-UUU, psbK-psbI, petA-psbJ, rps2-rpoC2, ndhC-trnV-UAC, rpl32-trnL-UAG and cemA-petA) longer than 200 bp with high Pi values were selected as candidate DNA barcodes for phylogenetic analysis and species identification of Seseli. Among them, matK gene, ccsA gene, and trnH-psbA region have been used as universal DNA barcodes in studying plant phylogeny [39,40,41]. We will further explore the reliability and effectiveness of these regions in future studies.

A suggestion of “a narrow sense” for Seseli based on phylogeny and morphology

In the nrDNA-based phylogenetic tree, the Seseli taxa did not form a monophyletic group and were divided into several branches. The “S. tortuosum” clade and the “S. gummiferum” clade were separated with high support, which was consistent with a previous study [42]. However, these two clades were closer to the clade consisting of five Peucedanum taxa and a monotype genus, Saposhnikovia Schischk., while they did not cluster with S. transcaucasicum, S. marashicum, and the eleven newly sequenced Seseli taxa. After determining the morphological characteristics of the taxa belonging to the “S. tortuosum” and “S. gummiferum” clade (Table S8), we found that the nine Seseli taxa belonging to the “S. tortuosum” clade shared many common characteristics: the segments of basal leaves linear to lanceolate, bracts nearly absent, bracteoles linear to lanceolate, the number of rays no more than 20 (except S. globiferum) and unequal, calyx teeth very minute, mericarps ovoid or oblong, and ribs prominent. According to the literature and the results of Lyskov et al. [42], taxa belonging to the “S. gummiferum” clade were morphologically different from the “S. tortuosum” clade (e.g., S. gummiferum and S. lehmannii had several bracts, S. rupicola had conspicuous calyx teeth, S. corymbosum had 30–70 rays) (Table S8). In addition, taxa of the two clades differ geographically: taxa belonging to the “S. tortuosum” clade are mainly distributed on the northern coast of the Mediterranean, while taxa belonging to the “S. gummiferum” clade are mainly distributed near the Black Sea. The nrDNA-based phylogenetic tree had several parallel branches, which resulted in lack of resolution for interspecific relationships, but it was still clear that Seseli was non-monophyly.

Most Seseli species are regional endemics and the composition of the type section of Seseli (i.e., the species closely related to the type species S. tortuosum L.) has not been clearly defined, which leads to imperfect and incomplete classification among Seseli species [42]. Moreover, given that the genera (Peucedanum, Ligusticum, and Libanotis) related to Seseli were also non-monophyly [16, 17, 28, 30], which made studying the phylogeny and taxonomy of Seseli harder. It is worth noting that many phylogenetic and taxonomic problems of Peucedanum and Ligusticum have been effectively resolved after considering a narrower sense [15, 43]. Overall, to further investigate the phylogeny and taxonomy of Seseli, we suggest that it is important to treat Seseli in a narrow sense. Thus, according to our results, the “S. tortuosum” clade could be referred to as the narrow sense of Seseli (within Selineae), containing common features such as leaf segments linear to lanceolate, bracts nearly absent, bracteoles linear to lanceolate, rays unequal, calyx teeth very minute, mericarps ovoid or oblong, ribs prominent, and consisting of the following species: S. tortuosum, S. arenarium, S. hartvigii, S. andronakii, S. grandivittatum, S. serpentinum, S. alexeenkoi, S. globiferum, and S. leptocladum.

The phylogenetic position of S. delavayi and the taxonomic relationship between Seseli and Eriocycla

Previous studies have revealed that members belonging to Seseli are distributed into three tribes: Selineae, Pimpinelleae, and Apieae [17]. However, S. diffusum belonging to Pimpinelleae has been treated as Psammogeton difusum (Roxb ex Sm.) Rech.f. ex Pimenov [20], while S. webbii belonging to Apieae has been treated as Canaria tortuosa (Webb & Berthelot) Jim.-Mejías & P.Vargas [21]. Consequently, the remaining Seseli taxa still belong to Selineae. However, in our study, we found that S. delavayi was furthest from the main branches of Seseli (within Selineae), but clustered with Eriocycla nuda and belonged to Echinophoreae [4]. S. delavayi had a similar plastome structure and SC/IR borders to Eriocycla nuda. It also possessed the most different morphological characteristics from other Seseli taxa but is similar to Eriocycla species: pubescent throughout, peduncles elongate, bracts 5-7, bracteoles longer than 2 × pedicels, prominent calyx teeth, and fruit densely white hispid (Table S9).

Taxonomic controversies between Seseli and Eriocycla have always existed: Degtjareva [12] proposed that several Seseli species (e.g., S. delavayi, S. afghanicum) should be transferred into Eriocycla, while Pimenov [8] treated Eriocycla taxa as synonyms of Seseli taxa based on type specimens. In our study, it was clear that the type species of Seseli, S. tortuosum, was located in Selineae, while the type species of Eriocycla, Eriocycla nuda, was nested in Echinophoreae. Additionally, S. tortuosum and Eriocycla nuda were morphologically dissimilar (base with clothed in fibrous remnant sheaths vs. base without clothed in fibrous remnant sheaths, peduncles not elongate vs. peduncles elongate, bracts absent or 0-1 vs. bracts 3-5, petals white-violet vs. petals bright yellow) [5, 44]. Therefore, Eriocycla should be identified as a separate genus based on our molecular and morphological evidence. Additionally, introducing more Eriocycla taxa would be useful to establish a robust phylogenetic framework for Eriocycla and resolve the taxonomic problem of S. delavayi.

The taxonomic complexity of Seseli based on phylogeny, comparative plastome analyses and morphological characteristics

Seseli taxa within Selineae failed to form a monophyletic group but were divided into five subclades. The comparative plastome analyses and morphological characteristics confirmed the reliability of our phylogenetic analyses and implied the complex evolution of Seseli. For the comparative plastome analyses, Seseli taxa belonging to different subclades showed distinguishing SC/IR borders, sequence divergence, and codon usage, which implied the complexity of plastome evolution and the non-monophyly of Seseli. Additionally, Seseli taxa had significantly different morphological features. For example, S. yunnanense, S. mairei, S. mairei var. simplicifolia, S. squarrulosum, and S. incisodentatum had glabrous mericarps, while other Seseli taxa had more or less hairy mericarps: S. glabratum had sparsely papillose, S. intramongolicum had densely papillose-pubescent, S. valentinae had densely puberulent, S. coronatum had sparsely puberulent, and S. eriocephalum had densely tomentose. In addition, mericarp in transverse section was slightly dorsally to laterally compressed in S. intramongolicum but strongly dorsally compressed in S. coronatum. Even though these taxa had different morphological characteristics, they also possessed several common characteristics: bracts absent or 1-2, bracteoles numerous, calyx teeth almost obsolete, and mericarp ovoid or oblong, which were very important taxonomic characteristics of Seseli [45]. These morphological similarities and differences of taxa indicated that Seseli was indeed a taxonomically complex genus.

The incongruences between nrDNA-based and plastome-based phylogenetic trees often appear in plant phylogenetic analyses [28, 46,47,48], and there is no exception in our results. For example, S. glabratum clustered with Libanotis buchtormensis, Libanotis spodotrichoma, and Saposhnikovia divaricata in the CDS-based phylogenetic tree, while it clustered with S. coronatum in the nrDNA-based phylogenetic tree, S. valentinae clustered with S. montanum in the CDS-based phylogenetic tree, while it was clustered with Ligusticum mucronatum in the nrDNA-based phylogenetic tree. These incongruences might be the result of hybridization, introgression, and incomplete lineage sorting (ILS) [49]. Moreover, Wen [50] proposed that chloroplast capture events in Apiaceae induced by early hybridization explained the incongruence of positions between tribes in the two phylogenetic trees. Further study is needed to identify the cause of the nuclear-plastome conflict in Seseli.

It is impossible for us to conduct taxonomic treatments of several Seseli taxa due to the lack of adequate morphological and molecular data of S. tortuosum. We recommend that comprehensive studies of morphological characteristics and molecular phylogeny should reduce the uncertainties in the taxonomy of Seseli. Overall, in our study, we verify the non-monophyly of Seseli based on both plastomes and nrDNA sequences, and this provides a foundation for studying the evolution, phylogeny, and taxonomy of Seseli.

Conclusion

In this study, we newly sequenced, assembled and annotated complete plastomes of eleven Seseli taxa. We observed that the genome length, gene numbers, IR/SC borders, and repeat composition of the Seseli plastomes were variable. Several appropriate mutation hotspot regions might be developed as candidate DNA barcodes for evolution, phylogeny, and species identification of Seseli. Thirty-seven plastome CDS and 59 nrDNA sequences were used to perform the phylogenetic analysis of Seseli. The phylogenetic results identified that Seseli was not a monophyletic group. Moreover, the eleven newly sequenced Seseli taxa did not cluster with S. tortuosum (the type species of Seseli, belonging to the tribe Selineae), where S. delavayi clustered with Eriocycla belonging to Echinophoreae and the other ten belonged to Selineae. The comparative plastome and morphological characteristics analyses confirmed the reliability of the phylogenetic analyses and implied the complex evolution of Seseli. We suggest that “a narrow sense” of Seseli will be meaningful for further study and the current taxonomic system of Seseli needs to be revised. Overall, our study can provide new insights into the phylogenetic relationships and taxonomic framework of Seseli.

Methods

Sample collection

Fresh and mature green leaves from adult plants of eleven taxa, namely S. mairei, S. mairei var. simplicifolia, S. yunnanense, S. squarrulosum, S. incisodentatum, S. glabratum, S. intramongolicum, S. valentinae, S. coronatum, S. eriocephalum, and S. delavayi, were collected from the wild (including Xinjiang, Gansu, Ningxia, and Yunnan provinces) and immediately dried with silica gel for subsequent treatment. These taxa belong to four of the five sections (Sect. Seseli, Sect. Hippomarathroidea DC., Sect. Macrostylopodium Schischk., and Sect. Pseudosilaus Schischk.) according to the treatment of Flora Republicae Popularis Sinicae (FRPS) [51]. The formal identification of these eleven samples was undertaken by Associate Professor Songdong Zhou (Sichuan University). Voucher specimens of the above taxa were deposited in the herbarium of Sichuan University (SZ) (Table S1).

DNA extraction, sequencing, assembly and annotation

Total genomic DNA was extracted from silica gel-dried materials using the modified CTAB method [52]. Then, we amplified ITS and ETS (internal and external transcribed spacer) sequences of these eleven Seseli taxa with primers ITS-4 (5’-TCCTCCGCTTATTGATATGC-3’), ITS-5 (5’-GGAAGTAAAAGTCGTAACAAGG-3’) [53], 18S-ETS (5’-ACTTACACATGCATGGCTTAATCT-3’) [54], and Umb-ETS (5’-GCGCATGAGTGGTGAWTKGTA-3’) [55]. Polymerase chain reactions (PCRs) were performed in a 30 µL volume with 2 µL plant total DNA, 1.5 µL forward primer,1.5 µL reverse primer, 15 µL volume 2 × Taq MasterMix (cwbio, Beijing, China), and 10µL ddH2O. We used the software DNAstar-SeqMan to edit and obtain the newly assembled ITS and ETS sequences [56].

For plastomes, raw data of the eleven newly sequenced Seseli taxa were generated by Illumina platform, generating 150 bp paired-end reads at Novogene (Beijing, China). The raw data was filtered through fastP v0.15.0 (-n 10 and -q 15) to ensure high quality [57]. After quality control, we acquired at least 5 GB of clean reads for each taxon. The clean reads were assembled using the program NOVOPlasty v2.6.2 [58], with default parameters and the rbcL gene sequence of S. tortuosum (MW662022) as the seed sequence. Genome annotation was performed using Plastid Genome Annotator (PGA) [59] with S. montanum (KM035851) as the reference. Manual adjustment compared with related species’ plastomes was conducted in Geneious v9.0.2 [60]. Then, we used the online program Organellar Genome DRAW (OGDRAW) [61] to draw circular plastome maps. Additionally, the gene rearrangements among the eleven Seseli plastomes were detected using Mauve Alignment [62] in Geneious v9.0.2 [60].

The newly generated plastomes, ITS and ETS sequences of the eleven Seseli taxa have been submitted to the GenBank under accession numbers ON975056-ON975066, ON980800- ON980810, and ON980787-ON980797 (Table S2).

Repeat sequence and nucleotide diversity analyses

The online REPuter program [63] was used to identify repeat sequences, and four types were included: forward, palindromic, reverse, and complementary repeats. The parameter settings were as follows: (1) a minimum repeat size of 30 bp; (2) more than 90% sequence identity between two repeats; and (3) Hamming distance = 3. In addition, we used the Perl script MISA (http://pgrc.ipk-gatersleben.de/misa/sleben.de/misa/) to detect simple sequence repeats (SSRs) of the eleven Seseli plastomes with thresholds (the minimum number of SSRs) of 10, 5, 4, 3, 3, and 3, for mono-, di-, tri-, tetra-, penta-, and hexanucleotide SSRs, respectively.

DnaSP version 6.12.03 [64] was used to calculate the nucleotide diversity (Pi) of protein-coding genes, non-coding regions, and introns to identify mutation hotspot regions. To develop potential and useful molecular markers for future analyses, the length of regions we selected was longer than 200 bp [65].

Phylogenetic analyses

Previous studies have shown that Seseli species are divided into three tribes: Selineae, Pimpinelleae, and Apieae [17]. Thus, in our study, we chose Aegopodium podagraria L. and Carum carvi L. belonging to the tribe Careae as the outgroup to root the phylogenetic tree, according to the results of Wen et al. [50]. The names of these tribes were mainly based on the work of Lyskov et al. [4] and Wen et al. [50]. All taxa and their accession numbers in GenBank included in the phylogenetic analysis are listed in Table S2.

Eighty common CDS of 37 Apiaceae taxa were extracted, respectively aligned, and concatenated as the plastome CDS dataset, using PhyloSuite v1.2.2 [66]. Previous studies of Apiaceae indicated that the nrDNA ITS, in conjunction with the ETS region, can provide more informative variation for phylogenetic reconstruction and allow for better resolution of relationships [55, 67, 68]. Thus, ITS and ETS sequences of 59 Apiaceae taxa were first respectively aligned using MAFFT v7.308 [69] and concatenated as the nrDNA dataset in PhyloSuite v1.2.2 [66]. Then, the two datasets (plastome CDS and nrDNA) were used to reconstruct the phylogeny of Seseli.

Maximum likelihood (ML) analysis was conducted using RAxML v8.2.8 [70] based on the best-fit GTRGAMMA model and 1000 bootstrap replicates. Bayesian inference was performed using MrBayes v3.2.7 [71] after the program Modeltest v3.7 [72] calculated the best-fitting models of nucleotide substitutions under the Akaike information criterion (AIC), and the best-fitting models were GTR+I+G for both nrDNA and plastome CDS dataset. Four independent Markov chains were run for 10,000,000 generations, with one tree sampled every 1,000 generations. The first 25% of the trees were discarded as burn-in. FigTree v1.4.2 [73] was used to edit the phylogenetic trees, with nodes under 50% bootstrap support being treated as parallel branches.

Comparative plastome analyses

Comparative plastome analyses were based on phylogenetic results, and a total of eighteen plastomes belonging to Selineae and Echinophoreae were selected. Of the eighteen, eleven were the newly generated Seseli plastomes from this study. The remaining seven plastomes were from published data (Libanotis buchtormensis, Libanotis spodotrichoma, Saposhnikovia divaricata, S. montanum, Ligusticum thomsonii, Ligusticum mucronatum, and Eriocycla nuda).

The online program IRscope [74] was used to display the borders between the inverted repeat (IR) and single copy (SC) regions, aiming to illustrate the structural differences in the eighteen plastomes. Then, sequence divergence of the eighteen plastomes was performed with the online program mVISTA in Shuffle-LAGAN mode [75], with S. yunnanense as the reference.

Eighty common single-copy coding sequences of the eighteen plastomes were extracted, and then codon usage analysis and relative synonymous codon usage (RSCU) values [76] were calculated using the CodonW v1.4.2 program [77]. TBtools [78] was used to make a heatmap to visualize the RSCU values.

Morphological data

Mericarps of the eleven Seseli taxa were collected from the field and were then photographed using a stereomicroscope (Nikon SMZ25). The mericarp terminology followed Kljuykov et al. [79]. Morphological data of these Seseli taxa and other related taxa involved in this study was obtained during our field observation, consulting type specimens, and previous literature review.