Comparative analysis of the plastid and mitochondrial genomes of Artemisia giraldii Pamp.

Artemisia giraldii Pamp. is an herbaceous plant distributed only in some areas in China. To understand the evolutionary relationship between plastid and mitochondria in A. giraldii, we sequenced and analysed the plastome and mitogenome of A. giraldii on the basis of Illumina and Nanopore DNA sequencing data. The mitogenome was 194,298 bp long, and the plastome was 151,072 bp long. The mitogenome encoded 56 genes, and the overall GC content was 45.66%. Phylogenetic analysis of the two organelle genomes revealed that A. giraldii is located in the same branching position. We found 13 pairs of homologous sequences between the plastome and mitogenome, and only one of them might have transferred from the plastid to the mitochondria. Gene selection pressure analysis in the mitogenome showed that ccmFc, nad1, nad6, atp9, atp1 and rps12 may undergo positive selection. According to the 18 available plastome sequences, we found 17 variant sites in two hypervariable regions that can be used in completely distinguishing 18 Artemisia species. The most interesting discovery was that the mitogenome of A. giraldii was only 43,226 bp larger than the plastome. To the best of our knowledge, this study represented one of the smallest differences between all sequenced mitogenomes and plastomes from vascular plants. The above results can provide a reference for future taxonomic and molecular evolution studies of Asteraceae species.

www.nature.com/scientificreports/ whole-genome sequence (WGS) data. We then assembled the extracted reads into a unitig graph. All the 'edges' of the unitig graph had the same coverage depth, suggesting the absence of plastid and nuclear sequences, which tend to show significantly higher or lower coverage depths. The unitig graph contained multiple double-bifurcation structures ('> = <' , DBSs) resulting from the presence of repeat sequences in the genome. To resolve the sequence path around these DBS, we constructed all possible sequences around the DBSs and mapped them to the Nanopore reads with minimap2 tool 48 . For each DBS, we selected the sequence path with the largest number of Nanopore reads mapped as the dominant sequence path. Finally, we identified a cyclic path on the unitig graph covering all the 'edges' . This path corresponded to a circular DNA sequence, which was considered the mitogenome.
Homology sequence analysis between plastid and mitochondrion. Sequence similarity comparison between the plastome (OK128342) and mitogenome (NC_064134.1) was carried out for the identification of homologous sequences between two organelles. BLASTN was used, and the e-value cutoff was 1e-5 56 . The final results were visualised using the Circos package implemented in TBtools 57,58 .
Selective pressure analysis of A. giraldii mitogenome. We used EasyCodeML (v1.4) software 69 to conduct the selective pressure analysis of 28 protein-coding genes in the mitogenome. The running model was 'Preset (Nested Models)' . The site model in EasyCodeML can be used in identifying positively selected sites in a multiple-sequence alignment 70 . The required inputs for analysing selection are aligned sequences in PAML format and a tree file in Newick format. Firstly, we aligned each gene from 10 species with MAFFT (v7) 63 and converted the alignment into PAML format by using the 'Seqformat Convertor' tool in EasyCodeML (v1.4). Then, we used IQ-TREE (v2) 65 to generate a tree file in Newick format. Finally, we ran the CodeML with the following parameters: nt = 0 and icode = 0' . On the basis of the lnL and np values of the null model (M0, M1a, M7 and M8a) and alternative model (M3, M2a and M8), the likelihood ratio test (LRT) p-value of each PCG was calculated. Then, the p-values were adjusted using the Benjamini-Hochberg correction method 71 . Genes with adjusted p-values of < 0.05 were considered positively selected.

Molecular marker development.
To discover universal primers that can be used in distinguishing the Artemisia species, we downloaded the 17 plastome sequences of Artemisia species from GenBank. They were analysed using ecoPrimers 72 with the following parameters: '-l 300 -L 600 -e 0 -3 2 -t species -U -f -O 25' . Here, '-l 300' specified the minimum barcode length as 300, excluding primers. '-L 600' specified the maximum barcode length as 600, excluding primers. '-e 0' specified the maximum number of mismatches allowed per primer as www.nature.com/scientificreports/ 0. '-3 2' specified the number of nucleotides on the 3′ end of the primers as 2, and these primers should have a strict match with their target sequences. '-t species' specified the taxonomic level used for evaluating barcodes and primers as 'species' . '-U' meant that no multi match of a primer on the same sequence record is allowed. '-f ' indicated the removal of data mining step during strict primer identification. '-O 25' specified the primer length to be 25. A custom script was used to extract the regions adjacent to the identified DNA barcode region for designing PCR primers.

Hypervariable region analysis.
To identify the hypervariable regions among the 18 Artemisia species, we wrote a custom script to extract the intergenic spacer regions (IGS) from the GenBank files of the 18 plastomes. Firstly, we extracted the IGS sequences using extractseq. Then, we aligned the extracted sequences using clustalw2 73 with options '-type = DNA -gapopen = 10 -gapext = 2' . Finally, we calculated the genetic distance of the intergenic regions using the K2p evolution model implemented in the distmat program from the EMBOSS package 74 with the parameter '-nucmethod 2' . Fourteen hypervariable IGS were identified (Fig. 6). To verify whether these molecular markers can distinguish the 18 Artemisia species, we extracted the top three most variable IGS regions from 18 Artemisia species for the alignment.
Ethics approval and consent to participate. We collected fresh leaf materials from A. giraldii for this study. No specific permits were required from the local government for the collection. In addition, we conducted the study in compliance with relevant institutional, national and international guidelines and legislation. We prepared the voucher specimens and deposited them in the Institute of Medicinal Plant Development (Beijing, China) with the accession number implad201910017.

Results
DNA sequencing, genome assembly and validation. In the Illumina sequencing data, a total of 21,579,647 sequences was generated, and the total number of bases was 3,236,947,050. The average read length was 150 bp. In the Nanopore sequencing data, a total of 10.225 Gb of 1,800,259 reads were obtained, and 8.227 Gb of 1,389,001 reads had Q of > 7, which were used in subsequent analysis. The average length of the remaining reads was 5.923 kb, N50 was 14.074 kb and the longest read was 114.470 kb. We used two strategies to assemble the plastome. In the first strategy, we used Illumina data alone. In the second strategy, we used the Illumina and Nanopore reads. The assembled results were identical except that the small single-copy (SSC) region was inverted between the two assemblies ( Supplementary Fig. S1A). In the mitogenome assembly, we used Illumina and Nanopore reads. We mapped the Illumina reads to the assembly results to obtain the coverage depth of the plastome and examine the quality of the assembly ( Supplementary Fig. S2). To determine the coverage depth of the mitogenome, we mapped the Illumina reads to the hybrid assembly results ( Supplementary Fig. S3). The average coverage depth was 121× for the mitogenome and 430× for the plastome. For locations with low coverage depths in the mitogenome and plastome, we used Tablet software 75 to visualise read cover in the genome. All low-coverage locations had spanned reads (Supplementary Figs. S2 and S3). We found more than 30 reads that covered the plastome locations with low coverage depths. By contrast, we found more than 10 reads that covered the mitogenome locations with low coverage depths. We used Bandage 76 to visualise the structure of the A. giraldii plastome ( Supplementary Fig. S4A) and mitogenome ( Supplementary Fig. S4B). The plastome was a typical circular sequence containing a large single-copy (LSC) region, a pair of identical inverted repeats (IRs) and an SSC region ( Supplementary Fig. S4A).
The unitig graph of the mitogenome showed a branched polymeric structure ( Supplementary Fig. S4B). Different contigs ( Supplementary Fig. S4B, left side) were linked to form a master chromosome (Supplementary Fig. S4B, right side). The principle chromosome can undergo rearrangement through repeat-mediated recombination, generating chromosomes with different rearrangements, called isomers 40 . We manually removed non-mitochondrial nodes from the graph according to the stratified coverage depth, and the repeat paths were resolved by aligning with the Nanopore long reads. Finally, a circular mitochondrial molecule was obtained ( Supplementary Fig. S4). The master chromosome encoded 54 genes: 32 PCGs, 3 rRNAs and 21 tRNAs. The quantities were consistent with those found in other Asteraceae species.
The total length of PCGs in A. giraldii plastome was 78,009 bp, representing 51.64% of the whole length of the plastome sequence. By contrast, the size of the rRNA was 9046 bp, and the size of the tRNA was 2693 bp, representing 5.99% and 1.78% of the total length of the A. giraldii plastome sequence, respectively. The GC content analysis showed that the overall GC content was 37.47%. In particular, the GC content for the protein-coding regions, rRNA genes and tRNA genes was 37.78%, 55.10% and 52.73%, respectively. The GC content in the LSC, SSC and IR regions was 35.56%, 30.78% and 43.09%, respectively.  www.nature.com/scientificreports/ 32 PCGs in the mitogenome (Fig. 1B). According to these functions, these 32 genes can be divided into 10 classes: ATP synthase (atp1, atp4, atp6, atp8 and atp9), cytochrome (ccmB, ccmC, ccmFc and ccmFn), ubichinol cytochrome c reductase (Cob), cytochrome c oxidase (cox1, cox2 and cox3), maturases (matR), transport membrane protein (mttB), NADH dehydrogenase (nad1, nad2, nad3, nad4, nad4L, nad5, nad6, nad7 and nad9), large subunit of ribosome (rpl5, rpl10), small submit of ribosome (rps1, rps3, rps4, rps12 and rps13) and succinate dehydrogenase (sdh4; Table 1). Repeat sequence analysis. In addition to difference in intergenic region, diversity in mitogenome size can be attributed to a large number of repeat sequences and foreign fragments 43,78 . Therefore, we analysed three   www.nature.com/scientificreports/ common types of repeated sequences. Microsatellites (simple repeat sequences, SSRs), also called tandem repeats of 1-6 bp, are abundant in the genomes of higher organisms and usually show high levels of polymorphism 79 . Therefore, they are generally used as molecular markers for identifying similar species 80 . SSRs can be classified into different types according to repeat unit. For instance, SSRs are classified into mono-, di-, tri-, tetra-, pentaand hexanucleotide repeats according to the length of their major repeat units 81 . We identified 36 SSRs in the plastid sequence and 51 SSRs in the mitochondrial sequence (Fig. 3, Supplementary Tables S3, S4). The most abundant SSRs in the plastome were single-nucleotide SSRs, including 19(A) and 12(T), accounting for 79.49% of the total SSRs. However, the SSRs in the A. giraldii mitogenome were dominated by tetranucleotide polymers, which accounted for 43.14% of all repeats. The types of SSRs in the mitogenomes were more evenly represented than in the plastomes. Tandem repeat sequences exist in the DNA of all organisms whose genomes have been sequenced. These sequences consist of multiple contiguous repeat units and exhibit extremely high mutation rates in eukaryotes and prokaryotes because they tend to gain or lose repeat units 82 . We identified 23 tandem repeats in the plastome and 15 in the mitogenome (Supplementary Tables S5, S6). The repeats can be further tested for their suitability as DNA fingerprinting markers.   Table S7). All the dispersed repeats in the plastome were less than 100 bp, the longest was 60 bp and the shortest was 30 bp. However, the number of dispersed repeats in the mitogenome was larger than those in the plastome. In the mitogenome, we found 135 dispersed repeats comprising 85 forward repeats, 49 palindromic repeats and 1 reverse repeat. They accounted for 62.96%, 36.30% and 0.74% of all dispersed repeats, respectively (Supplementary Table S8). The length of the dispersed repeat sequences ranged from 30 to 248 bp, but only 17 were longer than 100 bp.

Analysis of homologous sequences between two organelles. The transfer of mitochondrial and
plastid DNAs to the nucleus has been considered a part of the ongoing genome evolution and influences eukaryote evolution 83,84 . This process not only occurs from the organelle to the nucleus but also from the plastid DNA to the mitochondrial DNA 85,86 . For example, the plastid gene rbcL is transferred to the mitogenome numerous times during angiosperm evolution, and all evaluated sequences are pseudogenes 87 . To investigate whether plastid DNA is transferred to mitochondrial DNA, we used BLASTN 56 to identify potential homologous sequences between the plastome and mitogenome in A. giraldii, and the cutoff e-value was 1e-05. Nine DNA fragments were found between two organelle genomes (Fig. 4, Supplementary Table S9). The total length of the nine fragments was 4806 bp and accounted for 2.47% of the whole mitogenome. The longest fragment was 888 bp in the mitogenome, and the shortest was 79 bp. The location of the nine homologous fragments in the mitochondrial and plastid genomes is shown in Supplementary Table S9.

Phylogenetic inference analysis.
We constructed phylogenetic trees with the concatenated PCG sequences, using the maximum likelihood (ML) and BI methods (Fig. 5). The phylogenetic trees constructed with plastome and mitogenome sequences had minor differences in topological structures. In both trees, the 12 species were first divided into two main clades: a large clade composed of 10 Asteraceae species and a small clade composed of two outgroup species. A. giraldii was closely related to C. indicum and C. boreale in the two trees. In the mitochondrial genome tree, H. grosseserratus and H. annuus were clustered on one branch, and H. strumosus and H. tuberosus was clustered on another branch. However, in the plastome tree, H. annuus and H. tuberosus were separated into different branches, whereas H. grosseserratus and H. strumosus were clustered in a clade. The second difference was that L. sativa was located in different positions in the two trees. In the plastid tree, L. sativa was located in the outermost clade formed by the Asteraceae family. In the mitochondrial tree, L. sativa was located within the clade formed by the Asteraceae species (Fig. 5).  Table S10). The detailed analysis results can be found in Supplementary Table S11. The adjusted p-value of ccmFc, nad1, nad6, atp9, atp1 and rps12 is below 0.05, suggesting these six genes are subject to positive selection.

Molecular marker development.
Based on the 18 plastome sequences of Artemisia species, we found one molecular marker for distinguishing among 18 Artemisia species (Supplementary Table S12). It was a pair of highly conserved regions that can be used for primer design. The regions amplified by the primer pairs contained one or more SNP and INDEL sites that can be used in distinguishing among the 18 Artemisia species. However, the lengths of the regions were about 30 kb, which is extremely long for practical uses.

Analysis of hypervariable regions.
A total of 14 IGS were hypervariable regions (Fig. 6). The top three regions: ndhG-ndhI, ccsA-ndhD and rpl32-trnL-UAG had K2p values of 1.50, 1.22 and 1.06, respectively. We first extracted the top three hypervariable regions and aligned them ( Supplementary Fig. S7). However, the only two variant sites in ccsA-ndhD regions also existed in rpl32-trnL-UAG regions. Hence, we selected two regions: ndhG-ndhI and rpl32-trnL-UAG for molecular marker development. The variant sites in the two hypervariable regions can be used in distinguishing among the 18 species completely, including 11 SNPs and six indel sites ( Supplementary Fig. S7). As indicated in Supplementary Fig. S7A

Discussion
Artemisia giraldii is a medicinal plant primarily used as a source of traditional medicines. Obtaining its genomic information is the critical step for understanding the biosynthesis of its active components. As the first step, we sequenced and assembled the mitogenome and plastome of A. giraldii in the current study. Then, we analysed the mitogenome and plastome's general features and compared them in detail.
In the plastome, two copies of IRs separate SSC and LSC regions 88 . When an IR region is present, homologous recombination occurs between the two copies and results in the frequent 'flip' inversion of the SSC region between the two copies, thus allowing two heterogeneous genomic orientations to coexist in a single plant with approximately the same frequency 89,90 .
In this study, we used two strategies to assemble the plastome of A. giraldii. The two strategies generated two assemblies that were identical, except that the SSC region was inverted ( Supplementary Fig. S1A). The reverse and complement of the SSC region in the plastome assembly from Illumina and Nanopore data generated an assembly identical to that assembled by Illumina data (Supplementary Fig. S1B). Coverage depth is an indicator used in evaluating the correctness of an assembly in the mitochondria and the plastid genome assembling process. The drop of coverage depth is often considered a sign of misassembly. We observed several regions with low depths (Supplementary Figs. S2A and S3A,B). To determine whether assembling problems occurred, we visually examined the regions. The mapped results (Supplementary Figs. S2B and S3C) showed the reads sufficiently covered cover the regions, suggesting that the regions were correctly assembled. Further examination showed that the regions were AT rich. The AT-rich regions tend to be highly polymorphic and are error prone for long-read sequencing and result in a low coverage depth 91 .
The mitogenome of plants is much larger than the plastome 92 because of frequent exchange with nuclear and chloroplast DNA 93 , repeat sequences, AT-rich non-coding regions, large introns and non-coding sequences 94 . The mitogenome size commonly ranges from 200 to 2400 kb in angiosperms 95 . By contrast, the plastome size commonly ranges from 100 to 200 kb. We compared the sizes of the mitogenomes and plastomes of plants released in GenBank to determine if the small difference between the two organelles is unusual. Our results showed a small difference in size between the mitogenome and plastome in A. giraldii among the 318 species having both mitogenomes and plastomes released in GenBank by August 1st, 2022 (Supplementary Table S13).
The size difference between the mitochondria and plastid genomes in A. giraldii was extremely small, only 43,226 bp, compared with the size difference in other species. Among the 318 species, 95 showed the smaller difference between mitogenome and plastome sizes than A. giraldii. 94 of the 95 species were algae and mosses. The only angiosperm plant having a smaller size difference was Bidens pilosa from Asteraceae, with 1236 bp. www.nature.com/scientificreports/ Actually, its size difference was the smallest among all pairs of mitogenomes and plastomes in this study. These observations suggested that mitogenome expansion develops along with plant evolution. Among the Asteraceae species, A. giraldii had the second smallest difference. The other seven Asteraceae species, Bidens parviflora, Bidens biternate, Bidens bipinnata, Chrysanthemum indicum, Chrysanthemum boreale, Bidens tripartite, and Ageratum conyzoides, also had small size differences between their two organelle genomes, which were 44,511, 46,989, 46,990, 57,125, 59,990, 66,297 and 67,873 bp, respectively. This result indicated that small size difference is a common phenomenon in Asteraceae. The cause of this phenomenon has not yet been reported, and thus the specific mechanisms need to be further explored.
We drew a figure to show the sizes of the seven most representative mitogenomes. The largest known mitogenome was obtained from Cucumis melo. The smallest known angiosperm mitogenome was obtained from Bidens pilosa. The sizes of four Asteraceae mitogenomes were in between (Fig. 7). The mitogenomes of different plants differ greatly in size.
We analysed the homologous sequence between mitogenome and plastome. Sequence migration is common in plants 96 . The plastid or nuclear DNA fragments can be inserted into mitochondrial DNA, resulting in an expanding mitogenome. These cp-derived mtDNAs can contain complete or partial PCG sequences 87,97 and some tRNA sequences 86 . Frequently, these transfer sequences have no functions. We found nine homologous fragments between the plastid DNA and mitochondrial DNA. The total length of the nine fragments was 4806 bp and accounted for 2.47% of the whole mitogenome. To determine whether these homologous sequences originated from their common ancestor (vertical transfer) or were transferred from plastid to mitochondria (horizontal transfer), we determined whether these homologous sequences were present in the plastome and mitogenome of C. boreale with BLASTN. We found homologous sequences for eight fragments: F1, F2, F3, F5, F6, F7, F8 and F9 (Supplementary Table S9) in the plastome and mitogenome of C. boreale. We only found a homologous sequence for fragment F4 in the plastome of C. boreale. Therefore, we speculated that eight homologous fragments (F1, F2, F3, F5, F6, F7, F8 and F9) may have originated from their common ancestor and have been preserved throughout evolution. Another homologous fragment (F4) may have been transferred from the plastome to the mitogenome in A. giraldii. Thus, we suspected that a low degree of DNA exchange between the mitochondria and plastid DNAs is responsible for the low level of mitogenome expansion in A. giraldii.
Compared with plastomes and nuclear genes, the mitogenome has been rarely used in reconstruct phylogenies partly because of the slower nucleotide substitution rate and the difficulty of complete assembly and direct alignment 98,99 . We used the sequences of common genes to construct mitochondrial and plastid trees with ML and BI methods. A. giraldii was placed in the same locations in both trees. However, the plastid and mitochondrial trees differed in topology, particularly in the branch containing L. sativa and four Helianthus species. In the Figure 6. The hypervariable regions between the Artemisia genus. The horizontal direction represents the intergenic spacer regions that are highly variable among the 18 Artemisia species. The vertical direction is the arbitrary K2P distance of these regions. The square in the middle of each line represents the main distance of each intergenic spacer region. www.nature.com/scientificreports/ plastid tree, the L. sativa was located in the outermost clade formed by the Asteraceae family. In the mitochondrial tree, L. sativa was located within the clade formed by the Asteraceae species. According to the taxonomy (https:// www. ncbi. nlm. nih. gov/ Taxon omy/ Brows er/ wwwtax. cgi), L. sativa belongs to the Cichorioideae, whereas the other nine Asteraceae species belong to Asteroideae. Hence, the plastid tree was more in line with the taxonomic classification compared with the mitochondrial tree. L. sativa and Asteroideae species are located in different branches of the phylogenetic tree 100,101 . To further understand the relationship of mitochondrial genomes among 10 Asteraceae species, we aligned the mitogenome of A. giraldii (NC_064134.1) by using the BLASTN suite in NCBI (https:// blast. ncbi. nlm. nih. gov/ Blast. cgi). The results showed that the sequence similarity between 10 Asteraceae species was consistent with those shown in the mitochondrial tree (Supplementary Table S14). Compared with the four Helianthus species and A. conyzoides, the sequence similarity between A. giraldii and L. sativa was higher.
Previous report and sequence alignment results confirm the incongruence between the plastome tree and mitogenome tree for L. sativa. We hypothesised that the difference in topology between the two trees results from the inconsistent evolutionary rates of the plastome and mitogenome. Further analysis of the mitogenome of L. sativa is required to elucidate the incongruence. However, the support value between H. grosseserratus and H. strumosus in the plastome and between H. grosseserratus and H. annuus were less than 50 because of the high sequence similarity among Helianthus species, making the branches inseparable. The A. giraldii reported in this study had the same branch structure in the two trees and had a high support value, suggesting high credibility for its evolutionary relationship. The closest relatives to A. giraldii were C. indicum and C. boreale. This result is consistent with their taxonomic relationship, as they both belong to Artemisiinae. The collinearity results confirmed this conclusion. C. indicum and C. boreale had a larger synteny fragment than the A. giraldii mitogenome. Overall, the results revealed that the gene orders on the mitogenomes of the 10 Asteraceae species differed significantly.
Most mitochondrial genes are highly conserved and have undergone neutral and negative selection. The selective pressure analysis is commonly used in identifying positively or negatively selected genes to adapt to a particular lifestyle. In this analysis, the adjusted p-values of ccmFc, nad1, nad6, atp9, atp1 and rps12 were below 0.05, suggesting that these genes underwent positive selection in the evolution process. The other 22 genes were more conserved and not subject to positive selection. The adjusted p-values of ccmFc and nad1 were 0, suggesting that they are subject to strong positive selection. ccmFc was a protein similar to the C-terminal part of the bacterial ccmF. It is involved in cytochrome c maturation and is present in a large-sized complex in wheat mitochondria 102 . nad1 is one of the NADH dehydrogenases and plays an important role in mitochondrial electron transport 103 . Given the limited availability of mitogenomes in Artemisia, we used the plastome sequences of 18 Artemisia species to predict one pair of primers that potentially amplify a variable DNA region to distinguish among 18 Artemisia species. However, the length of the predicted amplified fragment was extremely long to validate. We concluded that this molecular marker may not be applicable to distinguish them. Instead, we analysed the hypervariable regions of the 18 species to obtain available molecular markers. Owing to the large number of species, the variant site in one hypervariable region cannot be used in distinguishing 18 species from one another. The variant site in ccsA-ndhD is present in rpl32-trnL-UAG, and thus the 17 variant sites in the two hypervariable regions (ndhG-ndhI and rpl32-trnL-UAG) were combined (11 SNPs and six indels). We were able www.nature.com/scientificreports/ to completely distinguish among 18 Artemisia species (Supplementary Fig. S7). Further experimental verification of these molecular markers is needed.

Conclusions
In this study, we assembled the mitogenome and plastome of A. giraldii for the first time. Phylogenetic analysis showed that the branch locations of A. giraldii in the phylogenetic trees constructed with the mitochondrial and plastid protein sequences were identical, suggesting the possible co-evolution of the genomes from the two organelles. Homologous sequence analysis identified nine homologous fragments between two organelles, and one fragment might have transferred from the plastome into the mitogenome. This study may provide a reference for studying the evolutionary relationship between mitochondria and plastids in Asteraceae species.

Data availability
The plastome and mitogenome sequences of A. giraldii reported in this article are available in GenBank (https:// www. ncbi. nlm. nih. gov/) with accession numbers OK128342 and NC_064134.1, respectively. The raw data have been submitted to the SRA database (BioSample: SAMN25050459; BioProject: PRJNA798221; SRA: SRR17652243). The sample has been deposited in the Institute of Medicinal Plant Development (Beijing, China) with accession number implad201910017.