Phylogenomics and adaptive evolution of hydrophytic umbellifers (tribe Oenantheae, Apioideae) revealed from chloroplast genomes

Abstract

Background

Tribe Oenantheae consists mainly of aquatic species within the Apioideae. The unique morphology and habitat distinguish this group from other Apioideae groups. However, the genomic information of these group species has not been widely developed, and the molecular mechanisms of adaptive evolution remain unclear.

Results

We provide comparative analyses on 30 chloroplast genomes of this tribe representing five genera to explore the molecular variation response to plant adaptations. The Oenantheae chloroplast genomes presented typical quadripartite structures, with sizes ranging from 153,024 bp to 155,006 bp. Gene content and order were highly conserved with no significant expansion or contraction observed. Seven regions (rps16 intron–trnK, rpoB–trnC, trnE–trnT–psbD, petA–psbJ, ndhF–rpl32–trnL, ycf1a–rps15, and ycf1a gene) were identified as remarkable candidate DNA markers for future studies on species identification, biogeography, and phylogeny of tribe Oenantheae. Our study elucidated the relationships among the genera of tribe Oenantheae and subdivided the genera of Sium and Oenanthe. However, relationships among the Oenanthe I clade remain to be further clarified. Eight positively selected genes (accD, rbcL, rps8, ycf1a, ycf1b, ycf2, ndhF, and ndhK) were persuasively detected under site models tests, and these genes might have played roles in Oenantheae species adaptation to the aquatic environments.

Conclusions

Our results provide sufficient molecular markers for the subsequent molecular studies of the tribe Oenantheae, and promote the understanding of the adaptation of the Oenantheae species to aquatic environments.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12870-024-05863-2.

Background

Oenantheae Dumort is a special tribe of Apioideae (Apiaceae) that is morphologically and ecologically well-defined. Species of this tribe are generally aquatic with clumped fibrous or tuberous-thickened roots (usually rooted at the base of the stem), and pericarp of the fruits with varying degrees of corky-thickened. Molecular phylogenetic studies in recent years have shown that Oenantheae is currently recognized as a natural and stable major clade of Apioideae, located in the middle of the phylogenetic tree [1–7]. Pollen of Oenantheae species vary in shape, covering almost every type from primitive to evolutionary [8, 9]. The integration of multiple traits (morphological, phylogenetic, ecological, and palynological) has shown that Oenantheae is the key group in studying the evolutionary history of Apioideae. The tribe currently contains about 24 genera and 120 species: including 5 genera (Berula W. D. J. Koch, Cicuta L., Cryptotaenia DC., Oenanthe L., and Sium L.) distributed in both Eurasia and North America; 12 genera (Atrema DC., Ptilimnium Raf., Tiedemannia DC., Harperella Rose, Limnosciadium Mathias & Constance, Cynosciadium DC., Lilaeopsis Greene, Neogoezia Hemsl., Trepocarpus Nutt. ex DC., Daucosma Engelm. & A.Gray, Oxypolis Raf., and Perideridia Rchb.) [10–12] endemic to North America; 3 genera [Caropsis (Rouy & E.G.Camus) Rauschert、Lereschia Boiss., and Trocdaris Raf.] [5, 12] and one hybrid genus (× Beruladium A.C.Leslie) [13] endemic to Europe; one genus (Helosciadium W.D.J.Koch) [14] distributed in both Eurasia and Africa; one monotypic genus (Apodicarpum Makino) [3] endemic to Asia; and one species (Peucedanum sandwicense Hillebr.) [12] endemic to Hawaii. There are 5 genera and around 17 species distributed in China [15].

Previous studies have shown that Oenantheae was recognized as a monophyletic tribe of Apioideae, but the relationships among the genera within this tribe have not been completely resolved. Most of the previous studies on the phylogeny of Oenantheae were carried out based on small molecular markers, including nrDNA ITS [5, 10–12, 16–19], rps16 intron [20], trnQ–trnK [18], psbI–^5’trnK^(UUU) [11], and rps16–^5’trnK^(UUU) [12, 14]. The intergeneric relationships of most genera within Oenantheae have been clarified, except for the genera within the Sium alliance. Recently, a biogeographical study on Oenantheae was implemented under comprehensive sampling using ITS and rps16–trnK markers [12]. The phylogenetic reconstruction results showed that Oenantheae was well divided into four lineages and two individual species, including the North American (NA) endemic clade [10, 19], the Cicuta & Oenanthe group [21], the Sium alliance [14], the Perideridia clade [22], Caropsis verticillatoinundata (Thore) Rauschert, and Trocdaris verticillatum (L.) Raf [12]. The relationships among genera within the North American (NA) endemic clade and the Cicuta & Oenanthe group were highly supported, but that of the Sium alliance (which was formerly known as Old World endemics clade in Spalik et al., 2009) were still controversial. The relationships among genera within the Sium alliance were poorly supported. Mostly, the insufficient informative sites of small molecular markers were suggested to be the main reason for the depressed resolution of phylogenies [2, 6].

Chloroplast genomes have been widely used for the reconstruction of plant phylogenies in recent years [6, 23–28]. Compared with small molecular markers, chloroplast genomes have sufficient information sites to meet the needs of phylogenetic reconstruction. The unique mode of inheritance ultimately reflected the “true” phylogenetic relationships of a given lineage. Nearly 2.2% of the published chloroplast genomes (NCBI, September 2023) were generated from Apiales species [27], most of which were from Apiaceae species. Apiaceae chloroplast genomes generally shared a typical quadripartite structure including a larger single copy region (LSC), a small single copy region (SSC), and two inverted repeat regions (IRA and IRB). However, the characteristics of genome structure varied among different species/genera of Apiaceae [6, 29, 30]. Comparative genomic studies concerning the characteristics of genes or functional groups of genes are of great importance in understanding plastome evolution and phylogenetic inference [27, 29, 31–33]. The chloroplast genome data were successfully used for phylogenetic reconstruction of Apiaceae in many studies [6, 24, 29, 30, 34–37]. However, the chloroplast genome of Oenantheae has not been paid much attention, and most of the chloroplast genome studies on this tribe involved individual species [38–41] or specific genus (for genus Sium) [42]. Regrettably, there was a lack of plastid phylogenomic analysis conducted for this tribe, and the genome differences of intergeneric plastids are still unknown within Oenantheae.

In this study, we filled this gap with a comparative analysis of the chloroplast genomes of 30 taxa (covering five genera) within Oenantheae. We conducted comprehensive analyses (1) to reveal the chloroplast genome characteristics and evolution of Oenantheae; (2) to identify suitable hotspot regions to use as candidate DNA barcodes for genera identification within the tribe; (3) to investigate the efficacy of chloroplast genome in resolving the relationships within the tribe. This is the first comprehensive chloroplast genome analysis of the tribe Oenantheae, which will provide genetic resources and complete the phylogeny analysis of this tribe. This study is crucial for reconstructing the phylogeny and evolution of the global Oenantheae.

Materials and methods

Taxon sampling, DNA extraction, genome sequencing, assembly, and annotation

Seventeen chloroplast genomes were newly sequenced, covering 4 genera and eleven species within Oenantheae (Fig. 1). The collections and voucher information for these newly sequenced species are provided in Table S1. Fresh and well-developed leaves were collected and dried in silica gel. Total genomic DNA was extracted from the leaf tissues using the Plant Genomic DNA Kit following the manufacturer’s protocol (Tiangen Biotech, Beijing, China), and the DNA quality and purity were further determined with the agarose gel electrophoresis. The high-quality DNA was used for library construction and sequencing. Paired-end sequencing libraries were generated on the Illumina Novaseq 6000 platform at Novogene (Beijing, China), with 2 × 150 bp paired-end reads and an insert size of 300 bp. DNA libraries were prepared using Rapid Plus DNA Lib Prep Kit for Illumina (RK20208). We sequenced 3G raw data for each sample.

Fig. 1.

Morphological diversity of Oenantheae species. A, Oenanthe javanica; B, O. hookeri; C, O. thomsonii; D, Sium ventricosum; E, S. serra; F, S. suave; G, Cicuta virosa; H, Cryptotaenia canadensis; I, Cryptotaenia japonica

Qualities of raw reads were checked by FastQC v0.11.9 [43]. The chloroplast genomes were assembled using NOVOPlasty v4.3.3 [44]. The Rubisco-bis-phosphate oxygenase (RUBP) sequences from multiple reference genomes (Table 1) were extracted to serve as seed files for different genera as appropriate. The assembled sequences were checked and annotated under Geneious Prime 2023.2.1 (created by the Biomatters development team, Ltd.), with gaps or degenerate bases corrected by Sanger sequencing. The chloroplast genome maps were drawn using OGDRAW [45]. All newly generated chloroplast genomes have been uploaded to the National Center for Biotechnology Information (NCBI) and are accompanied by detailed accession numbers, as shown in Table 1.

Table 1.

Summary of major characteristics of the Oenantheae chloroplast genomes, including sequence size (bp), number of genes, GC content, GenBank accession number, and Vouchers/References

Taxon	Size (bp)				GC content				Number of Genes					GenBank accession No.	Vouchers/References
	Total	LSC	IRs	SSC	Total	LSC	IR	SSC	Total	CDS	rRNA	tRNA	pseudogene
Cicuta virosa	154,438	84,193	26,355	17,535	37.50%	35.60%	42.70%	30.80%	133	86	8	37	2	PQ283862	LHM1252
Cicuta virosa	154,569	84,177	26,407	17,578	37.50%	35.60%	42.70%	31.00%	133	86	8	37	2	NC_037711	NCBI direct submission
Cicuta virosa	154,341	84,066	26,377	17,521	37.50%	35.60%	42.70%	30.90%	133	86	8	37	2	PQ283863	WJ2368
Oenanthe hookeri	154,597	84,414	26,456	17,271	37.50%	35.60%	42.70%	31.10%	133	86	8	37	2	PQ283867	WJ2334
Oenanthe javanica	154,246	84,096	26,456	17,238	37.60%	35.80%	42.70%	31.30%	133	86	8	37	2	NC_049874	Zhang et al., 2020b [41]
Oenanthe javanica	154,378	84,118	26,491	17,278	37.60%	35.80%	42.70%	31.20%	133	86	8	37	2	PQ283868	WJ2305_02
Oenanthe javanica	154,048	84,063	26,244	17,497	37.60%	35.80%	42.80%	31.30%	133	86	8	37	2	PQ283869	WJ2320
Oenanthe javanica	154,223	84,073	26,456	17,238	37.60%	35.80%	42.70%	31.20%	133	86	8	37	2	PQ283870	WJ2372_AH01
Oenanthe linearis	154,486	84,286	26,458	17,284	37.50%	35.60%	42.70%	31.10%	133	86	8	37	2	PQ283871	ZJW034
Oenanthe linearissubsp.rivulari	154,628	84,348	26,451	17,378	37.50%	35.60%	42.70%	31.00%	133	86	8	37	2	PQ283872	NZD2301
Oenanthe linearissubsp.rivulari	154,660	84,426	26,451	17,332	37.50%	35.60%	42.70%	31.10%	133	86	8	37	2	PQ283873	WJ2319
Oenanthe thomsonii	154,640	84,373	26,479	17,309	37.50%	35.60%	42.70%	31.10%	133	86	8	37	2	PQ283875	WJ2321
Oenanthe virgata	154,464	84,410	26,445	17,163	37.50%	35.60%	42.70%	31.20%	133	86	8	37	2	KX832335	Spooner et al., 2017 [39]
Oenanthe pimpinelloides	154,438	84,415	26,421	17,181	37.60%	35.70%	42.70%	31.30%	133	86	8	37	2	PQ283874	wj20230607_01
Cryptotaenia canadensis	154,101	84,048	26,435	17,183	37.50%	35.50%	42.70%	30.90%	133	86	8	37	2	PQ283864	WJ220908
Cryptotaenia japonica	153,385	83,616	26,419	16,931	37.50%	35.60%	42.70%	30.90%	133	86	8	37	2	PQ283865	WJ2375
Cryptotaenia japonica	153,228	83,451	26,417	16,943	37.50%	35.60%	42.70%	30.90%	133	86	8	37	2	PQ283866	WJ2384
Cryptotaenia japonica	153,259	83,477	26,419	16,944	37.50%	35.60%	42.70%	30.90%	133	86	8	37	2	NC_046737	Luo and Yu, 2019 [40]
Sium crispulifolium	154,099	84,707	26,150	17,092	37.40%	35.40%	42.70%	31.00%	133	86	8	37	2	OP234518	Zhou et al., 2023 [42]
Sium medium	154,598	84,116	26,472	17,538	37.40%	35.50%	42.70%	30.70%	133	86	8	37	2	NC_072108	Zhou et al., 2023 [42]
Sium ninsi	155,006	85,036	26,208	17,554	37.40%	35.50%	42.70%	30.60%	133	86	8	37	2	NC_072107	Zhou et al., 2023 [42]
Sium serra	154,690	84,491	26,223	17,753	37.40%	35.60%	42.70%	30.60%	133	86	8	37	2	PQ283876	WJ2374
Sium suave	154,676	84,155	26,469	17,583	37.40%	35.50%	42.70%	30.70%	133	86	8	37	2	NC_071929	Zhou et al., 2023 [42]
Sium suave	154,642	84,111	26,466	17,599	37.40%	35.50%	42.70%	30.70%	133	86	8	37	2	OP234519	Zhou et al., 2023 [42]
Sium suave	154,646	84,128	26,466	17,586	37.40%	35.50%	42.70%	30.70%	133	86	8	37	2	PQ283877	WBC2301
Sium tenue	154,929	84,952	26,204	17,569	37.40%	35.50%	42.70%	30.70%	133	86	8	37	2	NC_072106	Zhou et al., 2023 [42]
Sium ventricosum	153,024	84,294	24,970	18,790	37.40%	35.50%	43.10%	30.80%	133	86	8	37	2	NC_070095	NCBI direct submission
Sium ventricosum	153,029	84,300	24,970	18,789	37.40%	35.50%	43.10%	30.80%	133	86	8	37	2	OP234514	Zhou et al., 2023 [42]
Sium ventricosum	153,031	84,273	25,027	18,704	37.40%	35.50%	43.10%	30.80%	133	86	8	37	2	PQ283878	WJ2339
Tiedemannia filiformis subsp. greenmannii	154,737	84,581	26,510	17,136	37.30%	35.40%	42.60%	30.40%	133	86	8	37	2	HM596071	Downie and Jansen, 2015 [38]

Chloroplast genomes comparative analyses

A total of 30 chloroplast genomes from five genera of Oenantheae have been adopted to implement comparative analyses. Detailed accession numbers are listed in Table 1. We investigated the chloroplast genome structure of Oenantheae taxa, including gene number, length, and GC content of the whole genomes. Boundaries of the large single copy (LSC), small single copy (SSC), and inverted repeat (IR) regions among Oenantheae chloroplast genomes were compared in Geneious Prime 2023.2.1 to detect the IR expansion/contraction. To identify potential rearrangement and inversion events in chloroplast genomes, we utilized the Mauve Alignment [46] implemented in Geneious Prime.

To identify mutation hotspot regions among species within Oenantheae, the whole chloroplast genomes aligned matrix generated from MAFFT v7 [47] plug-in Geneious Prime 2023.2.1 was used to calculate nucleotide diversity (Pi). The sliding window analysis was conducted using DnaSP v6.10 [48], with a step size of 200 bp and a window length of 600 bp. We performed additional nucleotide diversity calculations for specific regions, including fragments previously utilized in the phylogenetic analysis of Oenantheae, the protein-coding genes (PCGs), their coding sequences (CDSs), and introns.

Positively selected analyses

To understand the process of evolution of chloroplast protein-coding genes, we calculated the non-synonymous (Ka) and synonymous mutations (Ks) of the CDSs. The ratios between non-synonymous and synonymous mutations (ω = Ka/Ks) can reveal the direction and strength of nature selection acting on the protein. Values of ω < 1, = 1, > 1 indicate negative purifying selection, neutral evolution, and positive selection, respectively [49]. A total of 80 CDSs were extracted from 30 chloroplast genomes of Oenantheae taxa, and the protein-coding sequence alignments were generated using Translation Align under the MAFFT program with stop codons removed accordingly. The petN and psbN genes were excluded due to their identical nucleotide sequences among taxa. Thus, only 78 CDSs were included in the later analysis. The DnaSP v6.10 was subsequently utilized to calculate the Ka and Ks.

We tested positive selection on site (codon) based models. Five (petN, psbL, psbN, rps7, and ycf3) were excluded from the testing due to having identical amino acid sequences. A total of 75 CDSs were employed to test positive selection. Firstly, 75 aligned CDSs were manually concatenated into a supermatrix in MEGA v7 [50] and subsequently used to infer an unrooted ML tree under RAxML v8.2.12 [51]. The unrooted ML tree and each CDS alignment served as input files in EasyCodeML analysis [52]. Site models in CodeML [53–55] executed in EasyCodeML were utilized to detect positively selected sites of PCGs in Oenantheae chloroplast genomes. These models treat the ω ratio for any site (codon) in the gene as a random variable from a statistical distribution, thus allowing ω to vary among codons [53, 54, 56]. Two pairs of particularly effective site models were adopted in our study: M1a (Nearly Neutral) vs. M2a (Positive Selection) and M7 (beta) vs. M8 (beta&ω). For these two tests, M1a and M7 are null models that do not allow for any sites with ω > 1, and the M2a and M8 are alternative models that allow some sites with ω > 1. These tests were widely used for positive selection [32, 57–62]. The presence of some sites at which ω > 1 was used to define positive selection [56]. Likelihood ratio tests (LRTs) [54] were performed to compare the null models against the alternative models. The P-values were calculated to measure the significance level of positive selection. If the LRT favors M2a or M8 (P < 0.05), we will identify sites in the genes that are under positive selection [56]. The Bayesian Empirical Bayes (BEB) method [57] was adopted and performed for the identification. A gene with LRT P < 0.05 and positively selected sites was considered a positively selected gene.

Phylogenetic reconstruction

A total of 32 taxa were sampled for phylogenetic analysis, including five genera from Oenantheae and two species from Scandiceae (Apioideae) as outgroups. The outgroups [Anthriscus sylvestris (L.) Hoffm., and Torilis scabra (Thunb.) DC.] were selected according to Wen et al., 2021, the NCBI accession numbers were MT561042 and MT561029, respectively. The 30 Oenantheae taxa contain 11 taxa from each genus Sium and Oenanthe, four taxa from Cryptotaenia species, three taxa from Cicuta, and one taxon from Tiedemannia. The five genera cover three lineages of Oenantheae (except for the Perideridia clade): including the sium alliance (Sium and Cryptotaenia), Cicuta & Oenanthe group, and Nouth American endemics clade (Tiedemannia). Eight datasets were adopted for reconstructing the phylogenetic tree. Five intergenic regions, including three small fragments previously used [1), psbI–trnK, 2), rps16–trnK, and 3), trnQ–trnK] [11, 12, 14, 18], 4), the noncoding region of the total genome, and 5), one dataset concatenated nine introns extracted from the corresponding PCGs (excluded introns of the rpl2 and ndhB genes due to the very small Pi). Two protein-coding sequences (CDSs) matrixes: 6), one dataset concatenated eight CDSs with high nucleotide diversity value, and 7), another concatenated 78 CDS from the whole genome. 8), The entire genome matrix was also used for analyses.

Two methods were employed to conduct phylogenetic analysis: maximum likelihood (ML) and Bayesian inference (BI). The ML analysis was performed by RAxML v8.2.4 [51], under the GTRGAMMA model as suggested in the manual. Rapid bootstrap analysis was implemented using 1000 bootstrap replicates to search for the best-scoring ML tree. MrBayes v3.2.7a [63] was employed to conduct the BI analysis under the best-fit model GTR + I + G recommended by jModelTest v2.1.4 [64]. Two independent Markov chain Monte Carlo (MCMC) runs were performed, each with three heated chains and one cold chain for 10,000,000 generations. The average standard deviation of split frequencies should approach zero, otherwise generations will be added to continue the analysis. Each run started with a random tree, sampling trees every 1000 generations, with the initial 25% discarded as burn-in. The bootstrap support (BS) and posterior probability (PP) were used to measure the supports of the phylogenetic tree implemented under ML and BI methods, respectively. The final tree was viewed in FigTree v1.4 [65].

Results and discussion

Gene characteristic diversity in Oenantheae chloroplast genomes

Among the seventeen newly sequenced plastid genomes, seven species were first reported. The size of Oenantheae chloroplast genomes ranged from 153,024 bp [Sium ventricosum (H.Boissieu) Li S.Wang & M.F.Watson, NC_070095] to 155,006 bp (Sium ninsi L.) (Table 1). Chloroplast genomes of Oenantheae species shared a typical quadripartite structure, which was consistent with that of most other Apioideae species [6, 29, 30, 36, 42], possessing two copies of IR regions (IRA and IRB, 24,970–26,510 bp) separated by the LSC region (83,451–85,036 bp) and SSC region (16,931–18,790 bp). The total GC content ranged from 37.30 to 37.60% roughly comparable to that in some other chloroplast genomes of Apioideae distal clades species [6, 66–70]. The IR region held the highest GC content (42.6–43.1%) when compared to LSC (35.4–35.8%) and SSC (30.4–31.3%) (Table 1, Fig. S1). The North American endemics Tiedemannia filiformis subsp. Greenmannii (Mathias & Constance) Feist & S.R.Downie showed the lowest GC content among the 30 Oenantheae chloroplast genomes in all regions (the total genome, LSC, SSC, IR regions) (Table 1). For the remaining species, GC contents of the total genome, LSC, and SSC were highest in the Oenanthe species, and lowest in the Sium species. The GC contents of IR regions were basically the same among species except S. ventricosum, which possessed a unique extra high GC content of 43.1%. We hypothesized that it might be related to the high-altitude habitat of this species. We reviewed the chloroplast genomes of Apioideae species, the GC content of IR regions of some high-altitude plants (e.g. Pleurospermum s. l.) did have high GC values (42.7–45.4%) [24], while others (e.g. Tongoloa H. Wolff) did not (40.5–42.4%) [37]. Therefore, the higher GC content may be related to the more complex evolutionary history of this species, which needs to be further studied.

Each of the 30 chloroplast genomes contained 133 genes, consisting of 86 PCGs, eight rRNA genes, 37 tRNA genes, and two pseudogenes. A total of 19 genes have been detected duplicated, including seven protein-coding genes (rpl2, rpl23, ycf2, ndhB, rps7, rps12, and ycf1), four rRNA genes (rrn4.5, rrn5, rrn16, and rrn23), seven tRNA genes (trnA-UGC, trnI-GAU, trnI-CAU, trnL-CAA, trnN-GUU, trnR-ACG, and trnV-GAC), and the pseudogene ycf15. Fifteen genes were owning some introns: 11 genes (atpF, ndhA, rpoC1, petB, petD, rpl16, rps16, and two copies of ndhB and rpl2) contained one intron, four genes (clpP, ycf3, and two copies of rps12) contained two introns. No significant expansion or contraction has been detected in Oenantheae chloroplast genomes (Table S2), so did the rearrangement and inversion events (Fig. S2). Each of the LSC/IRA junctions (J_LA) and SSC/IRA junctions (J_SA) in the Oenantheae chloroplast genomes were at the same positions. The J_LA was located between the rpl2 and trnH-GUG genes, and the J_SA was at the ycf1a gene (the copy of the ycf1 gene was largely located in the IRA region). The junction of LSC/IRB (J_LB) was located at the rps19 gene, except for Cicuta virosa L. (no. LHM1252, intergenic region of rps19 and rpl2), which was consistent with that mentioned in Wen et al., 2021. The junction of SSC/IRB (J_SB) was located at the ycf1b gene (the copy of the ycf1 gene was largely located in the IRB region) or the overlap region of ycf1b and ndhF.

Mutation hotspot regions among species of Oenantheae

Gene sequences with high nucleotide diversity (Pi) were suggested to be mutation hotspot regions. The sliding window analyses across the 30 Oenantheae chloroplast genomes showed that the Pi ranged from 0.0000 to 0.0663 (Fig. 2, Table S3). Overall, 188 sequences were owning Pi > 0.015. Among them 89 sequences with Pi > 0.02 (highlighted with red dots in Figs. 2), 22 sequences with Pi > 0.03, and six sequences with Pi > 0.04 (Table S3). The results showed that the sequences with Pi > 0.02 were all located in LSC and SSC, and sequences in IR regions generally had significantly lower Pi (Fig. 2). The sequences with high Pi were identified. Sequences with Pi > 0.02 were located at 30 regions, including matK–psbA, rps16–matK, rps16 intron–trnK, rps16 intron, rps16 intron–psbK, psbK–psbI, psbI–trnS–trnG, trnG–atpA, atpH–atpF intron, atpI–atpH, rpoB–trnC–petN, petN–psbM, trnD–trnY–trnE–trnT–psbD, ycf3–trnS–rps4, rps4–trnT–trnL, trnL–trnF–ndhJ, rbcL–accD, accD, petA–psbJ–psbL, psbE–petL, clpP intron, clpP intron–psbB, rpl16 intron, rps3–rpl16 intron, ndhF–rpl32–trnL–ccsA, ccsA–ndhD, ndhI–ndhG, ndhA intron, ycf1a–rps15–ndhH, ycf1a gene. Among these regions, 14 were formerly used or identified to reconstruct the phylogeny of Apiaceae groups [2, 42, 71–77]. Our study provides more alternative molecular markers for the phylogenetic studies of Apiaceae groups (especially for Apioideae major clades). Sequences with 0.03 < Pi < 0.04 were identified at seven regions, including ycf1a gene, trnE–trnT–psbD, ndhF–rpl32–trnL, rpoB–trnC, ycf1a–rps15, petA–psbJ, and rps16 intron–trnK regions. Among these regions, the ycf1a gene, trnE–trnT–psbD, ndhF–rpl32–trnL regions also contained the sequences with Pi above 0.04 (Table S3, Fig. 2). Sequences with Pi > 0.02 were mostly distributed in intergenic regions or introns of genes, with only three CDSs of the PCGs (rpl32, rps16, and ycf1a) owning Pi > 0.02 (Table S3, S4). In addition to the three, there were a few PCGs (accD, ccsA, matK, ndhF, and rps15) with Pi greater than 0.015 in their CDS regions. The remaining PCGs possessed generally low Pi (Table S4, Fig. 3A).

Fig. 2.

The nucleotide diversity (Pi) of 30 Oenantheae chloroplast genomes. Regions with high Pi (above 0.02) were marked out

Fig. 3.

The nucleotide diversity (Pi) of sequences. A, Pi of 78 protein-coding sequences (CDSs), sequences with Pi above 0.015 were marked out. B, Pi of 11 protein-coding genes, their CDSs, and introns

We calculated the Pi of genes, introns, and coding sequences (CDS) of the 11 genes that had introns (except the rps12 gene). Compared with the genes and coding regions, the introns each possessed a high Pi, except for the case of rps16 and ndhB. The Pi of these two genes are displayed as CDS > Gene > Intron (Fig. 3B). Overall, introns of five genes (including rpl16, ndhA, rps16, clpP, and petB) possessed Pi > 0.015, among them rpl16 (0.0227) and ndhA (0.0213) with Pi above 0.02 (Table S5). The Pi of the 11 CDSs were very low, except for the rps16 CDS (0.0215). The rps16 intron has been used to infer the phylogeny and evolution of Apiaceae, the subfamily, and their major clades for a long time [2, 71–75, 77]. Our study found that both the CDS and intron of the rps16 gene possessed Pi > 0.015. The Pi of other three small fragments previously used (psbI–trnK, rps16–trnK, and trnQ–trnK) in phylogenetic analyses of the Oenantheae group [11, 12, 14, 18] were also calculated with values above 0.02 (Table S6). Based on previous studies, all these fragments (rps16 intron, psbI–trnK, rps16–trnK, and trnQ–trnK) provided a certain resolution in the phylogenetic analysis of Oenantheae, but could not completely resolve all relationships within the tribe. Thus, we proposed that sequence regions with Pi greater than 0.03 (ycf1a gene, trnE–trnT–psbD, ndhF–rpl32–trnL, rpoB–trnC, ycf1a–rps15, petA–psbJ, and rps16 intron–trnK regions) can be considered as candidate barcodes for the phylogenetic analyses. We extracted the seven regions with Pi > 0.03 and aligned. All sequences are free of alignment ambiguities. A comparison of the number of variable sites, parsimony-informative positions, and the percent of parsimony-informative across these regions and the four previously used fragments have revealed that regions with Pi > 0.03 were identified as candidate DNA markers (Table S6).

Phylogenetic relationships among Oenantheae species inferred from multiple datasets

The robust phylogenetic trees were reconstructed from eight concatenated matrices, which exhibited clear relationships among genera within the tribe Oenantheae (Figs. 4 and 5). Our phylogenetic results strongly supported (BS = 100, PP = 1.00) that Oenantheae was divided into three clades: the Cicuta & Oenanthe group, the Sium alliance, and the NA endemics clade. The latter two clades formed the closest sister group (Figs. 4 and 5). This was basically consistent with previous studies [11, 12, 14]. The Sium alliance clade includes two genera, the Sium and the Cryptotaenia. Within the Cryptotaenia, the populations of Cryptotaenia japonica Hassk. gathered a branch, and subsequently formed a sister group to the Cryptotaenia canadensis (L.) DC. The Sium genus here was highly supported to be split into three branches: Sium I, Sium II, and Sium III. Two high-altitude species [Sium crispulifolium (H.Boissieu) Jing Zhou and S. ventricosum] clustered together to form the Sium I clade. Three species [S. ninsi, Sium tenue (Kom.) Kom., and Sium serra (Franch. & Sav.) Kitag.] gathered Sium II. The Sium III consist of Sium suave Walter and Sium medium Fisch. & C.A.Mey. Among these three branches, Sium I and Sium II were supported to be the closest sister (Figs. 4 and 5a–f), gathering the southern Palearctic clade. This topology structure was consistent with the previous studies that combined nrDNA ITS and cpDNA fragments (rps16–trnK) with improved support [12, 14]. Although an alternative result [((Sium I, Sium III), Sium II)] was proved in the phylogenetic analyses inferred from rps16–trnK and trnQ–trnK, this topology was moderately supported at the key nodes (Fig. 5g–h).

Fig. 4.

Phylogenetic relationships inferred from 32 species based on 78 shared CDSs. Support values marked above the branches follow the order Bayesian inference (PP, posterior probability)/maximum likelihood (BS, bootstrap support), * represent the best support (100%). Species names in bold indicate that these species are newly sequenced

Fig. 5.

Comparison of eight phylogenetic trees. Phylogenetic tree each inferred from a, the whole genome matrix; b, the noncoding region concatenated matrix; c, 9 introns concatenated matrix; d, 78 CDSs concatenated matrix; e, eight CDSs concatenated matrix; f, psbI–trnK; g, rps16–trnK; h, trnQ–trnK. Support values marked above the branches follow the order Bayesian inference (PP, posterior probability)/maximum likelihood (BS, bootstrap support), * represent the best support (100%), - represent PP values < 50%. In the black dotted box at the bottom right are the three topologies of the Oenanthe I clade

For the Cicuta & Oenanthe group, the taxa of Cicuta and Oenanthe each formed a nature branch. The genus Oenanthe was strongly supported to be divided into three branches. The relationships among these branches were incontrovertible with Oenanthe I and Oenanthe II forming the closest sister group with high supports inferred from the whole genome, 78 CDSs, 8 CDSs, and psbI–trnK (Figs. 4 and 5). Although the results inferred from the remaining four datasets have only moderate support, the topology is consistent. Two tuberous geophyte species [Oenanthe pimpinelloides L. and Oenanthe virgata Poir.] were gathered together to form Oenanthe III. The Oenanthe II consisted of populations of Oenanthe javanica (Blume) DC. and divided into two branches. The populations of high-altitude and low-altitude clustered separately (Table S1, Fig. 4). The Oenanthe I contained four taxa with narrow leaf blades, two alpine taxa (Oenanthe hookeri C.B.Clarke and Oenanthe thomsonii C.B.Clarke), Oenanthe linearis Wall. ex DC., and Oenanthe linearis subsp. rivularis (Dunn) C.Y.Wu & F.T.Pu. The interspecific relationships within Oenanthe I were ambiguous due to the extremely short differentiated branches of the Oenanthe I clade and the depress supports within the clade. Three topologies have been inferred from the eight datasets. Two populations of O. linearis subsp. rivularis formed a stable branch within Oenanthe I. The first topology (i1) reconstructed based on the whole genome, non-coding region, and 9 introns was very stable with the alpine branch well supported to form the closest sister group with O. linearis subsp. rivularis (Fig. 5a–c). The phylogenetic tree inferred from 78 CDSs has a unique topology (i2) of the Oenanthe I: the alpine branch and O. linearis were the closest sister groups with very lower supports (BS = 39, PP = 0.85). The remaining datasets shared the third topology (i3), the O. linearis and its subspecies clustered together, but the supports were not high (Fig. 5e–h). The interspecific relationships of the genus have been unresolved for a long time, and phylogenetic relationships based on small fragments (nrDNA ITS, cpDNA rps16–trnK) have often received poor support in previous studies [1, 11, 12, 14]. Our study provides a basis for resolving the phylogenetic relationships within the genus, to better solve the problem within Oenanthe I, we recommend obtaining rich nuclear genes for comprehensive analyses. Populations of the S. suave, C. japonica, and C. virosa were divided with very short branch lengths, with moderate node supports in the four large datasets (whole genome, noncoding region, 9 introns, and 78 CDSs) and very low supports in the four small datasets (Fig. 5).

Our results provided stable phylogenetic relationships among species of Oenantheae except the Oenanthe I clade. Molecular data with sufficient non-coding sequences appear to be effective in resolving interspecific relationships within the Oenanthe I clade (Fig. 5a–c). However, some additional data may need to be adopted into the study to clarify the relationships between species within the Oenanthe I clade. At present, many molecular data have been proposed to resolve generic-level phylogenies, such as single-copy nuclear genes (SCG), low-copies nuclear genes, and SNPs extracted from transcriptome or genome [78–82]. Despite the discordance between nuclear-based and plastid-based phylogenetic trees, sufficiently stable topologies can support a more comprehensive interpretation of taxa evolution [6, 83–85]. The previous phylogenetic analyses of the tribe Oenantheae rarely involved taxa distributed in China [11, 12, 14]. Our research contributes in part to the replenishment of Oenantheae’s global resources.

Positively selected gene responses to the adaptive evolution of Oenantheae

We calculated the Ka, Ks, and their ratio Ka/Ks (ω) to evaluate the selection pressure of the 78 PCGs (Table S4). The results showed that the Ka values ranged from 0 to 0.0246 and the Ks values ranged from 0 to 0.0594. Thirteen PCGs (atpH, clpP, ndhC, petG, petL, psaI, psbD, psbE, psbF, psbI, psbL, rps7, and ycf3) have Ka = 0, which indicated these PCGs without any non-synonymous mutations. Only one PCG (psaJ) with Ks = 0, which suggested this gene has non-synonymous mutations but no synonymous mutations. The Ka/Ks (ω) value of psaJ was not calculated because Ks = 0 and the ω value tended to infinity. The ω values for the remaining PCGs ranged from 0 to 1.0161 (Table S4). A total of 12 PCGs were identified with ω values above 0.5 (ccsA, matK, psbK, psaJ, rbcL, rpl20, rpoA, rps14, rps15, ycf1a, ycf1b, and ycf2) (Fig. 6). Among these genes, psaJ and ycf2 possessed ω values greater than 1. The results suggested ten PCGs might be under relaxed selection with 0.5 < ω < 1 and two genes have undergone strong positive selection.

Fig. 6.

The average Ka, Ks, and Ka/Ks values curves of the protein-coding sequences (CDS). CDSs with Ka/Ks > 0.5 were marked out with red circles in the Ka/Ks curve

To detect sites under positive selection in each PCG, we performed tests by comparing null models (M1a and M7) against alternative models (M2a and M8). We have calculated several parameters to estimate the positive selection, including the LRT P-values of the compared models (M1a vs. M2a, M7 vs. M8), the detailed log-likelihood values and free parameters of M7 and M8, and the positively selected sites calculated under BEB method (Table S7). The calculation of LRT P-values showed that eight PCGs (accD, rbcL, rps8, ycf1a, ycf1b, ycf2, ndhF, and ndhK) were significant (P < 0.05) for both M1a-M2a and M7-M8, two PCGs (ndhH and rps11) were not significant (P > 0.05) for the M1a-M2a comparison but significant under the M7-M8. Based on BEB analysis, sites had probability > 50% for the positive-selection class with ω > 1 were detected in 44 PCGs (Table S7). Among these genes, 14 PCGs (rpoC2, rpl20, psbT, psaJ, and the ten PCGs significant under M7-M8 mentioned above) have some positively selected sites with a probability > 95% (Table 2). The LRT of the four PCGs (rpoC2, rpl20, psbT, and psaJ) were not significant (P > 0.05) for both comparisons although their positively selected sites probability greater than 95%. Sites that had probabilities > 99% were presented in seven PCGs (rbcL, rps8, ycf1a, ycf1b, ycf2, ndhF, and rps11).

Table 2.

Positive selected sites identified in the chloroplast genomes of Oenantheae

Gene name	LRT P-value		Positively selected sites (BEB, *: PP > 95%; **: PP > 99%)	sites number
	M7 (beta) vs. M8 (beta&ω)	M1a (nearly neutral) vs. M2a (positive selection)
accD	0.000004515	0.000000015	100 H 0.983*,207 K 0.954*	2
rbcL	0.000000000	0.000000000	28 D 0.974*,91 A 1.000**,95 N 0.997**,97 Y 0.998**,225 I 0.997**,309 I 0.998**,328 S 1.000**,354 I 0.956*,355 A 0.972*,443 E 0.998**	10
rps8	0.000000000	0.000000000	57 D 1.000**	1
ycf1a	0.000000000	0.000000000	469 R 0.960*, 795 V 0.982*,1014 K 0.971*,1085 I 0.969*,1092 K 1.000**,1093 I 0.965*,1094 L 0.986*,1111 Y 0.962*	8
ycf2	0.000000000	0.000000000	103 F 1.000**,503 Q 0.996**,746 L 0.996**,833 R 0.955*,915 P 0.950*,1025 Q 0.951*	6
ycf1b	0.000000888	0.000000131	200 S 0.976*,205 K 0.994**,206 Y 0.980*,215 R 0.975*,217 S 0.975*,244 N 0.971*,270 K 1.000**	7
ndhF	0.000004851	0.000065086	465 Q 0.988*,655 N 0.952*,663 C 1.000**	3
ndhH	0.003524280	0.082768146	198 I 0.960*	1
ndhK	0.011201551	0.022716790	181 F 0.964*,218 K 0.957*	2
rps11	0.012214628	0.054106065	14 G 0.993**	1
rpoC2	0.082830660	0.347241769	239 I 0.969*	1
rpl20	0.096456032	0.117651549	12 R 0.964*	1
psbT	0.225892513	0.225960517	33 K 0.972*	1
psaJ	0.422979316	0.423014425	10 V 0.977*,14 L 0.977*,36 A 0.977*	3

Our Ka and Ks calculation results have suggested ten PCGs (ccsA, matK, psbK, rbcL, rpl20, rpoA, rps14, rps15, ycf1a, and ycf1b) under relaxed selection and two (psaJ and ycf2) undergone strong positive selection. Among these PCGs, four (rbcL, ycf1a, ycf1b, and ycf2) have detected significant positive selection in some sites. The LRTs of these PCGs were significant at a 1% level under both M7-M8 and M1a-M2a tests. Also, the BEB calculation provided strong evidence for sites under positive selection as the posterior probabilities for ω > 1 were high (Table 2). Together, these results suggested strong evidence for sites under positive selection in these four PCGs (rbcL, ycf1a, ycf1b, and ycf2). The BEB calculation also provided some evidence for sites under positive selection in PCGs rpl20 and psaJ as the posterior probabilities for ω > 1 were high (probability > 0.95). However, the LRTs of these two PCGs were not significant (P > 0.05) for both M7-M8 and M1a-M2a comparisons (Table 2). These situations also existed in the rpoC2 and psbT genes. These results suggested some weak evidence for sites under positive selection in the four PCGs (rpl20, psaJ, rpoC2, and psbT). For the remaining six PCGs (ccsA, matK, psbK, rpoA, rps14, and rps15), the LRT P-values were above 0.05 for both two comparisons, and the BEB calculation only provided weak evidence for sites under positive selection as the posterior probability for ω > 1 was low (< 95%) and the posterior distribution for ω was diffuse at every site (Table S7). All these suggested that these six PCGs might be under weak positive selection. Site models have detected extra four PCGs (accD, rps8, ndhF, and ndhK) possessing significant positively selected sites. Both the LRT and BEB calculations have provided strong evidence. The tests for sites under positive selection in the ndhH and rps11 genes were equivocal with the M1a-M2a and M7-M8 comparison giving conflicting results. We tend to accept the test of M1a-M2a because this test is noted to be more stringent when the evidence for positive selection exists but is not very strong [56, 86]. These results together suggested some evidence for sites under positive selection in the ndhH and rps11 genes, although this evidence was not very strong.

Tribe Oenantheae has a high proportion of bog and water species relative to terrestrial members. The occupation of the aquatic environments indicates that Oenantheae species have adapted to conditions of damp, marshy, or truly aquatic habitats. Compared with terrestrial species, emergent aquatic plants occupy stressful ecological habitats characterized by low light, and reduced carbon and oxygen availability [87]. Survival in aquatic habitats has resulted in adaptations specific to low oxygen concentrations. Plants of the tribe Oenantheae exhibit various adaptations to these habitats, such as clusters of fibrous or tuberous roots and corky-thickened fruits [7, 11]. Here possible evidence of positive selection in chloroplast coding genes was detected to reveal the adaptation of Oenantheae species to aquatic environments at the molecular level. Based on our results, a total of eight genes (accD, rbcL, rps8, ycf1a, ycf1b, ycf2, ndhF, and ndhK) are considered positive selection genes. Gene accD encodes the beta carboxyl transferase subunit of ACCase and has been verified to affect several biological processes such as storage compound metabolism, leaf and seed development [88–91]. The two significantly positively selected sites we detected in the accD gene varied mainly between the Cicuta & Oenanthe group and the other species. We identified these genes as possibly associated with the unique traits Cicuta & Oenanthe group owned particularly in leaf and seed development. We have found statistically that the leaf and fruit morphology of Oenanthe and Cicuta are different from other groups. The leaf blades of these two genera species always have one or more pinnates. The fruits of them are overall ovoid with thick and corky fruit ribs, with vittae solitary in each furrow and 2 on commissure [7, 18, 92–95]. The chloroplast ribosomal proteins act as important constituents of protein synthesis machinery and participate in various processes of plant growth, and development as well as reaction to unfavorable conditions [96–99]. Adaptive evolution of the rps8 gene may be helpful for the normal growth and development of Oenantheae species in aquatic environments. The remaining six positive selection genes are functionally associated with photosynthesis. The rbcL gene encodes the Rubisco large subunit, which catalyzes the assimilation of atmospheric CO2 during photosynthesis [100–102]. Positive selection of this gene has been detected in some plants and suggested adaptation to low CO2 concentration, shade-tolerant, and aquatic environments [32, 60, 103, 104], but not prevalent in the Apiaceae. We detected ten positive selection sites in the rbcL gene, the increased amino acid replacement may reflect the continuous fine-tuning of Rubisco under varying species’ ecological conditions. The ndh genes (ndhF and ndhK) encode NADH-dehydrogenase subunits of the Ndh1-complex bound to the thylakoid membrane, which were fundamental to the electron transport chain for the generation of ATP, and photosynthesis of plants [105–108]. Positive selection of these two ndh genes may reflect the adaptation to aquatic environments stress, such as low light and reduced oxygen availability. Positive selections of these two genes in Apiaceae have been detected commonly [24, 36, 109]. Essential genes ycf1 and ycf2 in plant chloroplast genomes encode proteins that are indispensable for cell survival and photosynthesis [110–112]. Overall, positive selection of the eight chloroplasts protein-coding genes (accD, rbcL, rps8, ycf1a, ycf1b, ycf2, ndhF, and ndhK) may be responses to aquatic environment adaptation of Oenantheae species. There may be partial positive selection in six genes that with positive selection sites (posterior probability above 95%), including rpl20, psaJ, rpoC2, psbT, ndhH, and rps11, but it is not statistically significant. Nevertheless, we suggest that these genes still require attention because they may be potential positive selection genes.

Conclusion

In conclusion, we generated the complete chloroplast genomes of 17 Oenantheae species. A deep comparative analysis of 30 Oenantheae chloroplast genomes has been conducted to create an accurate map of the genomic structure of hydrophytic umbellifers. A total of 30 regions (matK–psbA, rps16–matK, rps16 intron–trnK, rps16 intron, rps16 intron–psbK, psbK–psbI, psbI–trnS–trnG, trnG–atpA, atpH–atpF intron, atpI–atpH, rpoB–trnC–petN, petN–psbM, trnD–trnY–trnE–trnT–psbD, ycf3–trnS–rps4, rps4–trnT–trnL, trnL–trnF–ndhJ, rbcL–accD, accD, petA–psbJ–psbL, psbE–petL, clpP intron, clpP intron–psbB, rpl16 intron, rps3–rpl16 intron, ndhF–rpl32–trnL–ccsA, ccsA–ndhD, ndhI–ndhG, ndhA intron, ycf1a–rps15–ndhH, and ycf1a gene) have been identified with high nucleotide diversity (Pi > 0.02), which suggested to be mutation hotspot regions among Oenantheae species. Among these, the ycf1a gene, trnE–trnT–psbD, ndhF–rpl32–trnL, rpoB–trnC, ycf1a–rps15, petA–psbJ, and rps16 intron–trnK regions were considered as candidate barcodes for the phylogenetic analyses of Oenantheae due to well aligned and rich informative sites. Eight genes (accD, rbcL, rps8, ycf1a, ycf1b, ycf2, ndhF, and ndhK) with strong signatures of positive selection were convincingly detected. Two (accD and rps8) are related to plant growth and development, and the remaining are mainly related to the photosynthesis system. Such a comparative genomic analysis of the Oenantheae chloroplast genomes offers us an unprecedented opportunity to reveal footprints of the adaptive evolution of chloroplast genes that make Oenantheae species adapt to aquatic environments. Our phylogenetic analyses also recovered reliable relationships among species of Oenantheae except within the Oenanthe I clade. Phylogeny and evolution of the genus Oenanthe may be further clarified by incorporating more nuclear genes in further studies. In addition, expanding species sampling of North American endemic species will contribute to a deeper understanding of the molecular adaptive evolution of Oenantheae. Our research only serves as an introduction, and we urge other researchers to join us in the future.

Author contributions

Conceptualization, WJ and SCF; validation, WBC, LHM, ZW, YJX; interpretation of data, ZJW and MXD; analysis, WJ and YJX; writing—original draft preparation, WJ; writing—review and editing, SCF and WJ; funding acquisition, WJ. All authors have read and agreed to the published version of the manuscript.

Funding

This work was supported by the National Natural Science Foundation of China (Grant no. 32200191), the Foundation of Jiangsu Key Laboratory for the Research and Utilization of Plant Resources (JSPKLB202214).

Data availability

The datasets generated and analyzed during the current study are available in the NCBI (https://www.ncbi.nlm.nih.gov) repository. The raw reads have been submitted to GenBank at NCBI BioProject PRJNA1154308, with SRA numbers SRR30524793–SRR30524809. The assembly and annotated chloroplast genomes were deposited in NCBI with accession numbers PQ283862–PQ283878. Voucher specimens were identified by Jun Wen and deposited in NAS (Herbarium, Institute of Botany, Chinese Academy of Sciences, Jiangsu Province) with deposition numbers (NAS00640404– NAS00640420, Fig. S3).

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.