The DNA sources were the fresh leaves of B. biternata, B. bipinnata, B. pilosa var. pilosa, B. pilosa var. radiata, B. parviflora and B. tripartita. These species were identified by Prof. Yulin Lin from the Institute of Medicinal Plant Development (IMPLAD), Chinese Academy of Medical Sciences and Peking Union Medical College. Voucher specimens were deposited in the herbarium at IMPLAD. The ID numbers and collecting locations are shown in Supplementary Table S1. The total DNA of the species was extracted by using the DNeasy Plant Mini Kit (Qiagen Co., Germany), and DNA concentration and quality were assessed by using Nanodrop 2000C spectrophotometry and electrophoresis in 1% (w/v) agarose gel, respectively.

The DNA was used to generate libraries with an average insert size of 350 bp and sequenced by using Illumina NovaSeq6000 in accordance with standard protocols, and the sequencing information was shown in Supplementary Tables S2, S3. Paired-end sequencing was performed to obtain 150 bp sequences at both ends of each molecule. Adapters and low-quality regions in the original data were trimmed by applying Trimmomatic software (Bolger et al., 2014). Reference organelle sequences from the family of Asteraceae were downloaded from NCBI genome resources (https://www.ncbi.nlm.nih.gov/genome). The gene sequences of each reference were extracted to build a custom database, and clean reads were mapped by using BWA 0.7.17 (Li and Durbin 2009). NOVOPlasty 4.2 (Dierckxsens et al., 2017) was used to assemble organelle genomes, which needed a sequence as the initial seed. For the chloroplasts, a read from the psbA gene was selected as the seed input, and the output chloroplast sequences were manually adjusted for the start position. For the mitochondria, seed reads from the conserved cox and nad genes were separately tested, and the output contigs were combined and manually linked to obtain the mitochondrial sequences. Then, clean reads were mapped back to the chloroplast and mitochondrial genomes and inspected in IGV (Robinson et al., 2017) to exclude any assembly error. Finally, a custom-made script that took the assembly and the bam files as input was utilised to correct ambiguous bases and generate the complete organelle genome sequences. The sequences were initially annotated by using the CPGAVAS2 software (Shi et al., 2019) and the GeSeq (Tillich et al., 2017) and corrected manually. tRNAs were annotated by using tRNAscan-SE software (Schattner et al., 2005). Genes, introns and coding region boundaries were compared with reference sequences. Then the borders of LSC, SSC and IR regions in the chloroplast genomes were validated by designing primers and polymerase chain reaction (Supplementary Table S4).

Chloroplast and mitochondrial genome maps were generated by using the Organellar Genome DRAW v1.2 (Lohse et al., 2007) and manually corrected. The CodonW software (Sharp and Li 1987) was adopted to analyse codon usage. Simple sequence repeats (SSRs) were detected by using the MISA (Beier et al., 2017) with the definition of ≥ 10 repeat units for mononucleotide SSRs, ≥ 5 repeat units for dinucleotide SSRs, ≥ 4 repeat units for trinucleotide SSRs, and ≥ 3 repeat units for tetranucleotide, pentanucleotide and hexanucleotide SSRs. Long repeated sequences were detected by using REPuter (Kurtz et al., 2001). Sequence homology analysis was carried out by using EMBOSS (Rice et al., 2000). The boundaries of the four regions of the chloroplast genomes were compared by applying IRscope (Amiryousefi et al., 2018).

The chloroplast genomes of Bidens species were compared by using mVISTA software (Frazer et al., 2004). The nucleotide diversity (Pi) values of shared genes and intergenic spacers were calculated with DnaSP software (Librado and Rozas 2009). By combining the mVISTA results and Pi values, seven highly variable regions were screened out. The complete organelle genome sequences of Asteraceae species and common protein-coding genes shared by these genomes were used to construct maximum likelihood (ML) phylogenetic trees by utilising IQ-TREE (Nguyen et al., 2015) with a bootstrap of 1,000 repetitions. ML analysis was conducted based on the TVM + F + R4 (complete chloroplast genomes), TVM + F + I + G4 (complete mitochondrial genomes), TVM + F + R3 (common protein-coding genes shared by chloroplast genomes), and GTR + F + G4 (common protein-coding genes shared by mitochondrial genomes) models. MEGA software (Tamura et al., 2013) was used to construct Neighbor-joining (NJ) phylogenetic trees based on seven highly variable regions. NJ analysis was conducted based on the K2P model.

The 12 chloroplast genomes of these Bidens species showed a typical circular tetramerous structure and included two inverted repeats (IRs), a large single copy (LSC) and a small single copy (SSC) (Figure 1A). The lengths of the chloroplast genomes of the same species were the same or similar. The total lengths of the chloroplast genomes were 150,489 bp (B. tripartita) to 151,635 bp (B. pilosa var. radiata). The sizes of the LSC regions ranged from 83,499 bp (B. tripartita) to 83,899 bp (B. biternata). The sizes of the SSC regions varied between 17,628 bp (B. tripartita) and 18,439 bp (B. pilosa var. radiata). The sizes of the IR regions varied from 24,652 bp (B. bipinnata) to 24,701 bp (B. pilosa var. radiata). The total GC contents of the 12 chloroplast genomes were all 37.5%, indicating that the base composition of the Bidens species was relatively conserved (Supplementary Table S5).

The 12 chloroplast genomes were all annotated with 130 genes, including 85 protein-coding genes, 37 tRNA genes and eight rRNA genes (Supplementary Table S6). Amongst these genes, 17 were located in IR regions. They included six protein-coding genes (ndhB, rpl2, rpl23, rps7, rps12 and ycf2), four rRNA genes (rrn4, rrn4.5, rrn5 and rrn16) and seven tRNA genes (trnA-UGC, trnI-CAU, trnI-GAU, trnL-CAA, trnN-GUU, trnR-ACG and trnV-GAC). In addition, ycf1 and rps19 were annotated as pseudogenes. At the junctions, the gene positions in the boundary regions of the chloroplast genomes of the Bidens species were conserved. The difference was that the ndhF gene of B. tripartita was located at the boundary of the SSC and IRb regions, and the ndhF genes of the other five species were located in the SSC regions (Figure 2). Coding regions (protein-coding regions, tRNA genes and rRNA genes) accounted for 56.0–59.4% of the regions, and the rest were noncoding regions (pseudogenes, introns and gene spacers).

The eight mitochondrial genomes of these Bidens species all had a circular molecular structure (Figure 1B; Supplementary Figure S1) that ranged in length from 183 kb (B. pilosa var. pilosa) to 216 kb (B. tripartita). The GC contents were between 45.4 and 45.8%. The mitochondrial genomes of B. biternata and B. bipinnata were annotated with 83 genes, including 30 protein-coding genes, 18 tRNA genes, three rRNA genes and 32 open reading frames (ORFs). A total of 81, 87, 90 and 82 genes were annotated in the mitochondrial genomes of B. pilosa var. pilosa, B. pilosa var. radiata, B. parviflora and B. tripartita, respectively. The total proportion of all genes in the mitochondrial genomes of the Bidens species was between 24.4 and 28.6% (Supplementary Table S7).

The genes annotated in the mitochondrial genomes of the Bidens species could be divided into 11 categories (Supplementary Table S8). The similar gene types and numbers of complex I, complex III, complex IV, complex V, rRNA genes and maturation enzyme genes indicated that these genes were highly conserved in the mitochondrial genomes. The number of tRNA genes ranged from 18 to 20, and the number of ORFs was quite different.

The relative synonymous codon usage (RSCU) of the chloroplast and mitochondrial genomes of the Bidens species was calculated on the basis of all protein-coding genes (Figure 3). The results showed that the chloroplast and mitochondrial genomes of the Bidens species contained 64 types of codons encoding 20 amino acids. Leucine, serine and arginine all had six types of codons. Amongst all amino acids, leucine had the highest number of codons, whereas cysteine had the lowest.

Given that methionine and tryptophan possess only one codon each, no codon usage bias was found, and the RSCU value was 1. Codon usage bias was found for the rest of the amino acid codons. In the chloroplast genomes, 30 codons were found with RSCU > 1, of which 29 were A/U-ending codons, and 34 codons were found with RSCU ≤ 1, of which 31 were G/C-ending codons. The highest and lowest RSCU values were recorded for UUA and CGC, which encoded leucine and arginine, respectively. In the mitochondrial genomes, 30 codons were found with RSCU > 1, of which 28 were A/U-ending codons, and 34 codons were found with RSCU ≤ 1, of which 30 were G/C-ending codons. The highest and lowest RSCU values were recorded for GCU and CAC, which encoded alanine and histidine, respectively. These results indicated that the chloroplast and mitochondrial genomes exhibited a higher bias towards A/U-ending codons than towards G/C-ending codons.

In this study, a total of 56 (B. parviflora) to 71 (B. tripartita) SSRs were detected in the chloroplast genomes of these Bidens species. The distribution of SSRs in the three samples of B. biternata was the same. It was also the same in the four samples of B. bipinnata. Bidens bipinnata had one more mononucleotide repeat than B. biternata, and the remaining SSR types and numbers were the same. SSR distribution differed between the two B. pilosa var. radiata samples. The MW551955 sample had two more mononucleotide repeats, one more dinucleotide repeat and one more trinucleotide repeat than the MW551952 sample, indicating obvious SSR polymorphism in the chloroplast genome of B. pilosa var. radiata. Mononucleotide to hexanucleotide repeats were present in the Bidens species, most of which were mononucleotide repeats, followed by dinucleotide and tetranucleotide repeats (Supplementary Table S9). The SSRs of the mitochondrial genomes of the Bidens species were also analysed. A total of 50 (B. pilosa var. pilosa) to 62 (B. pilosa var. radiata) SSRs were detected in the mitochondrial genomes of the Bidens species. Six types of SSR were detected in the mitochondrial genomes of B. biternata, B. bipinnata, B. pilosa var. radiata and B. parviflora, whereas only five types were detected in B. pilosa var. pilosa and B. tripartita (Supplementary Table S10).

Some repetitive sequences with length ≥ 30 bp and sequence similarity ≥ 90% were found. These sequences were called long repetitive sequences, namely, forward (F), palindrome (P), reverse (R) and complement (C). In the chloroplast genomes of Bidens species, our analysis revealed 46 (B. biternata and B. bipinnata) to 114 (B. pilosa var. radiata) long repeats, most of which were F and P repeats. The long repeats mainly had lengths of 30–39 bp and included four types. Repeats that were larger than 40 bp were mainly F and P repeats. No repeats with lengths of 50–59 and 60–69 bp were found in B. biternata, B. bipinnata and B. parviflora, and no repeat with lengths of 60–69 bp was found in B. tripartita (Supplementary Table S11). The mitochondrial genomes of Bidens species all contained F and P repeats. The mitochondrial genomes of B. biternata, B. bipinnata, B. pilosa var. pilosa, B. pilosa var. radiata and B. parviflora contained R repeats. Only the mitochondrial genomes of B. biternata, B. bipinnata and B. pilosa var. pilosa contained C repeats. The mitochondrial genome of B. tripartita did not contain R and C repeats. The F and P repeats accounted for more than 90% of the long repeats in the mitochondrial genomes of the Bidens species. The number of long repeats was the highest in B. pilosa var. radiata and B. tripartita, with 106 and 108 F repeats, respectively. In these species, the number of repeats with lengths of 30–39 bp was the largest, followed by that of repeats with lengths of 40–99 bp; repeats with length ≥ 1 kb were rare (Supplementary Table S12).

Consistent with the chloroplast genome analysis above, the global comparison of the chloroplast genomes of the Bidens species showed that the seven chloroplast genomes of B. biternata and B. bipinnata were highly similar. A high degree of similarity was found between the samples of B. pilosa var. pilosa and B. pilosa var. radiata. The difference between B. tripartita and the other five species was the largest, and a mutation locus was found in the intergene region of ndhF-rpl32. In addition, the results revealed that the variation in the noncoding region was considerably greater than that in the coding region. Most of the variation was located in the LSC and SSC regions, and slight variation occurred in the IR regions. The rRNA genes of these species were highly conserved with little variation (Figure 4).

The Pi values of the shared genes and intergenic spacers of the chloroplast genomes of the Bidens species were calculated. Figure 5 shows the intergenic spacers and genes with Pi > 0. Intergenic spacers had more polymorphisms than gene regions, and these results were consistent with the mVISTA analysis results. By combining the mVISTA and Pi results, seven candidate highly variable regions (Pi > 0.018; length > 200 bp) were screened out for species identification and relationship studies.

Consistent with the result based on chloroplast genomes, the results of collinearity and homology analysis showed that the mitochondrial genomes of B. biternata and B. bipinnata had high homology. The mitochondrial genomes of B. pilosa var. pilosa, B. pilosa var. radiata, B. parviflora and B. tripartita were quite different from those of B. biternata and B. bipinnata (Figure 6). The dot plot of the mitochondrial genome sequences between B. biternata and B. bipinnata showed an evident line on the diagonal, indicating that the chloroplast genome sequences of B. biternata and B. bipinnata had high homology (Supplementary Figure S2). In the dot plot of mitochondrial genome sequences between B. pilosa var. pilosa and B. pilosa var. radiata, only several diagonal lines made up of marker points were parallel to the diagonal lines, which represent the same substring of two sequences. This result indicated that the mitochondrial genome sequences of B. pilosa var. pilosa and B. pilosa var. radiata have low homology (Supplementary Figure S3).

In the current study, the complete chloroplast genome sequences of five species and one variety of Bidens and 21 other Asteraceae species were used to construct a phylogenetic tree with Magnolia officinalis and Nicotiana tabacum as the outgroups (Figure 7). The results showed that the Bidens species clustered in a big branch and that different samples of the same species clustered together. The three samples of B. biternata clustered in one branch, which was sister to the cluster comprising the four samples of B. bipinnata. Bidens pilosa var. radiata and B. pilosa var. pilosa clustered together. Bidens tripartita and B. parviflora clustered in a single branch, respectively. In addition, the species in Ligularia were close to those in Bidens.

Then, the common protein-coding genes shared in these chloroplast genomes were applied to construct the ML phylogenetic tree (Supplementary Figure S4). The result was slightly different from the finding based on the complete chloroplast genome sequences. The four samples of B. bipinnata did not cluster together. For the seven candidate highly variable regions, only trnS-GGA-rps4 showed the capability to identify Bidens species (Supplementary Figure S5). The species of B. bipinnata, B. biternata and B. pilosa clustered in different branches, respectively. In contrast to the tree based on the complete chloroplast genome, B. parviflora was sister to the B. pilosa species but with a low bootstrap value.

The ML phylogenetic tree was constructed by using the complete mitochondrial genomes and common protein-coding genes shared by the mitochondrial genomes of Asteraceae species, including the Bidens species (Supplementary Figures S6, S7). The Bidens species did not cluster together in the tree based on the complete mitochondrial genomes, whereas the Bidens species clustered together in the tree based on the common protein-coding genes shared by these mitochondrial genomes. However, species within the genus Bidens were indistinguishable from each other. The results showed that in contrast to the chloroplast genome, the mitochondrial genome was unsuitable for the identification and phylogenetic analysis of Bidens species.