Introduction

Chloroplasts, responsible for photosynthesis in green plants, are cytoplasmic organelles beside the nucleus and the mitochondrion in plant cells. A typical chloroplast genome in angiosperm is double-stranded circular DNA including 120–130 genes and ranging from 120 to 170 kb (Odintsova and Yurina 2006; Ruhlman and Jansen 2014). This genome is generally composed of two inverted repeat copies (IRa and IRb) ranging 25 kb in size separated by one large single-copy region (LSC) and a small single-copy region (SSC) with dissimilar size ~80 and ~20 kb, respectively (Palmer 1991; Raubeson and Jansen 2005; Ruhlman and Jansen 2014). The evolutionary of chloroplast genomes in angiosperm are provided by structure, gene content and organization such as plastid genome rearrangement of Fabaceae, Asteraceae, Poaceae, Oleaceae, and Gesneriaceae (Saski et al. 2005; Kim et al. 2005; Jansen and Palmer 1987; Doyle et al. 1992; Hachtel et al. 1991; Lee et al. 2007; Hoot and Palmer 1994). During the evolution of land plants, many genes were lost and transferred to the nucleus or became pseudogenes. For example, the chloroplast genomes of Orchidaceae species revealed that the ndh genes were deleted or truncated (Wu et al. 2009). Interestingly, the ndh genes were also nonfunctional in diverse plant lineages including Pinus thunbergii, Keteleeria davidiana, Ephedra equisetina, and Welwitschia mirabilis (Wakasugi et al. 1994; McCoy et al. 2008). In addition, ycf15 and ycf68 are nonfunctional in several chloroplast genomes (Guo et al. 2007; Steane 2005; Schmitz-Linneweber et al. 2001). Chloroplast DNA has been a potential candidate for plant evolutionary studies because of its simple structure, highly conserved sequence, and maternal inheritance (Tian and Li 2002).

Colchicum autumnale L. and Gloriosa superba L. are commonly important sources of colchicine which was originally extracted from bulbs and seeds. It is a type of alkaloid used for the treatment of gout and rheumatism, painful muscles, inflammation, and patients with familial Mediterranean fever (FMF) (Kim et al. 2003; Lange et al. 2001; Touitou et al. 2008; Ade and Rai 2010). Since 2009, colchicine has been used in drugs under approval of Food and Drug Administration (FDA, USA). On the other hand, colchicine has been considered a chemical marker for identifying the Colchicaceae family within Liliales (Vinnersten and Larsson 2010). Based on phylogenetic studies, both species belong to the same tribe Colchiceae of subfamily Wurmbeoideae (Vinnersten and Manning 2007; Nguyen et al. 2013). C. autumnale L., a native plant to South and Central Europe and Africa, is herbaceous perennial and has a corm consisting of a brown membrane or scales (Baker 2001). G. superba L. is a perennial flowering plant with tubers (underground stem) and distributed throughout Southern Africa and tropical Asia including foothills of Himalayas, Burma, Indonesia, and so on (Gec et al. 2002; Ade and Rai 2010).

Our understanding of the chloroplast genome organization and evolution has improved due to advances in next-generation sequencing (NGS) techniques. Currently, over 300 chloroplast genomes are available in The Chloroplast Genome Database (http://chloroplast.ocean.washington.edu/tools/cpbase/run). However, few genomics studies have been conducted on medicinal plants while natural plant products are commonly used for drug development. Up to date, there were a few studies and projects focusing on chloroplast genomes of medicinal plants such as Korean Ginseng Panax ginseng C. A. Mey (Kim and Lee 2004), Chinese sage Salvia miltiorrhiza Bunge (Qian et al. 2013), and Mongolia medicine Artemisia frigida (Liu et al. 2013). The traditional Chinese medicine project has established a foundation for the development of natural medicines and the selection of cultivars with good traits based on genomic research (Chen et al. 2011). Additionally, 12 medicinal plant species were examined within the Medicinal Plant Genomics Resource project (http://medicinalplantgenomics.msu.edu) to obtain genomic data and provide novel information to the community regarding genes and markers for medicinal compound synthesis.

Although C. autumnale and G. superba play an importantly commercial role especially in medicinal industries and genetic engineering (Brown 1995), their genomic studies have not been performed. Therefore, here we analyzed the complete chloroplast genomes sequences of both species and it is the first record in colchicine produced plants and Colchicaceae family. The genome organization, gene contents, and order were compared between both species and among previously reported cpDNA within Liliales to understand the genomic feature of Colchicaceae in Liliales. Especially, we discussed about the finding of gene loss only occurred in Colchicum sensu Vinnersten and Manning, and suggested it as a useful molecular marker which can be easily detected in this genus.

Materials and methods

DNA extraction, sequencing and assembly of chloroplast genome sequences

Plant materials used in this study were collected through KNRRC (Medicinal Plant Resources Bank NRF-2010-0005790), supported by Korea Research Foundation, to which resources were provided by the Ministry of Education, Science and Technology in 2013. Fresh young leaves were collected from C. autumnale and G. superba cultivated in a greenhouse, and a voucher specimen was deposited in the herbarium of Gachon University (GCU). Total genomic DNA was isolated using the DNeasy Plant Mini Kit (Qiagen, Seoul, South Korea) from 1 g fresh leaves. The DNA concentration was determined using a UV–visible spectrophotometer (BioSpec-nano; Shimadzu Corp. Japan). High-quality DNA was used as a template for Hiseq 2000. Geneious (version 6.1, Biomatters Ltd. Auckland, New Zealand) was used to assemble pair-end reads from the NGS data based on default setting. We used Alstroemeria aurea (KC968976) (Kim and Kim 2013) as a reference sequence to align the contigs and identify gaps. To fill the gap, Sanger methods were applied and used to identify the borders of the IR, LSC, and SSC regions. PCR products were purified using the PCRquick-spin Kit according to the manufacturer’s protocol (Intron Biotechnology, Korea). We sequenced PCR products using the BigDye Terminator Cycle Sequencing Kit (Perkin Elmer Applied Biosystems). Summary of the sequencing data was described in Supplementary Table 1.

Annotation, codon usage, and comparative analysis

The complete sequences of two species were annotated using Geneious ver. 6.1 program, and manual correction was performed to identify the gene and exon boundaries. All tRNAs were confirmed using the tRNA scan-SE search server (Schattner et al. 2005). Other protein coding genes were verified by BLAST search on the NCBI website (http://blast.ncbi.nlm.nih.gov/), and manual correction for start and stop codons were conducted. GenomeVx software (Conant and Wolfe 2008) was used to construct the visual cpDNA map. The codon usage for all exons of protein coding genes excluding pseudogenes was examined using MEGA5 (Tamura et al. 2011). To calculate the ratio of synonymous (dS) and nonsynonymous (dN) substitutions among reported cpDNA in Liliales, BioEdit (Hall 1999) and DnaSPs software (Librado and Rozas 2009) were used. For the comparative analysis, we used six and four complete chloroplast genome sequences from Liliales and other monocot orders, respectively (Table 1).

Table 1 Comparison of the plastid genome sequence of Colchicum autumnale and Gloriosa superba and the other monocots
Full size table

To examine the gene loss of ycf15, a pair of primers was designed to amplify regions from ycf2 to trnL-CAA using Primer3 software (Untergrasser et al. 2012) based on the sequence assembly results of C. autumnale and G. superba. PCR amplifications were performed in 25-µl reactions containing 10× PCR reaction buffer, 0.5 U e-Taq DNA polymerase (SolGent Co. Ltd, Korea), 25 mM magnesium chloride (MgCl2), 2.5 mM deoxynucleotide triphosphates (dNTPs), 10 pM primer, and 50–100 ng template DNA. The PCR reactions consisted of one initial denaturation step of 3 min at 94 °C followed by 25 cycles of denaturation at 94 °C for 1 min, annealing at 56 °C for 30 s, extension at 72 °C for 20 s, and a final extension of 5 min at 72 °C. The PCR products were sequenced and aligned using MUSCLE (Edgar 2004), and further manual adjustments were made if necessary.

Examination of repeat units

Repeat sequences were analyzed using REPuter (Kurtz and Schleiermacher 1999) in both genomes using a minimum repeat size of 18–20 bp. The GRAMENE Ssrtool (Temnykh et al. 2001) program was used to identify SSRs.

Results

Chloroplast genome assembly and features

For C. autumnale, a total of 9,314,386 reads with an average length of 91 bp were generated from the Illumina HiSeq 2000, and 715,966 reads (7.68 %) were assembled to the A. aurea reference genome (KC 968976). The cpDNA of C. autumnale was 156,462 bp (KP 125337, Fig. 1) in length and composed of a pair of inverted repeats (IRs) of 27,741 bp, which were divided by SSC and LSC of 84,246 and 16,734 bp, respectively (Fig. 1; Table 1). The cpDNA sequence consisted of 132 coding genes, including 87 protein coding genes, 37 transfer RNA (tRNA) genes, and eight ribosomal RNA (rRNA) genes. Among the 132 genes, four rRNA, eight tRNA, and seven coding genes were duplicated in the IR regions (Supplementary Table 2).

Fig. 1
figure 1

Gene maps of the C. autumnale L. and G. superba L. chloroplast genomes. Genes shown outside of the outer circle are transcribed counter-clockwise, whereas those shown inside are transcribed clockwise. The thick lines in small circles indicate the inverted repeats (IR) regions. The asterisks indicate genes containing intron(s). The arrow indicate ycf15 gene which is not present in the C. autumnale chloroplast genome

Full size image

The same NGS platform was applied in the sequencing of the G. superba chloroplast genome, which generated 6,464,451 reads with an average length of 91 bp, of which 19,973 reads (0.3 %) were assembled for genome sequencing. The chloroplast genome size of G. superba was 157,924 bp (KP 125338) and displayed the typical angiosperm cpDNA structure, consisting of the LSC (85,012 bp), SSC (16,786 bp), and two IR copies (28,063 bp) (Fig. 1; Table 1). It also included 134 coding genes, of which 89 were protein coding genes, and 37 and 8 were distinct tRNA and rRNA genes, respectively. The IR regions contained 20 duplicates including four rRNA, eight tRNA, and eight coding genes (Supplementary Table 2).

Comparison of the chloroplast genome sequences between C. autumnale and G. superba revealed that the overall A+T and G+C contents in both whole chloroplast genomes were similar 62.4 and 37.6 %, respectively (Table 1). These percentages are within the range of typical monocots 61–63 % for AT content and 31–38 % for GC content. In both genomes, 19 genes contained introns, 16 of which contained one intron such as atpF, ndhA, ndhB, petB, petD, rpl16, rpl2, rpoC1, rps12, rps16, trnA-UGC, trnG-GCC, trnI-GAU, trnK-UUU, trnL-UAA or trnV-UAC, while two genes (clpP and ycf3) were possessing two introns. For the rps12 gene, the first exon was located in the LSC region, and the second exon was duplicated in the IR regions. Among the genes containing an intron, the trnK-UUU gene had the largest intron (2675–2711 bp) containing matK.

The gene contents were similar between both genomes, excluding ycf15 gene loss found in C. autumnale cpDNA. The infA gene was present as a pseudogene in both genomes because of several stop codons. In addition, an incomplete duplicated copy of ycf1 located at the junction between IRb and SSC was present in both genomes. In the chloroplast genome of G. superba, ycf68 and ycf15 in the IR were thought to be pseudogenes based on the presence of internal stop codons. However, in the C. autumnale chloroplast genome, ycf68 did not contain internal stop codons, and ycf15 was completely deleted (Fig. 2).

Fig. 2
figure 2

Alignment of three pseudogenes in the Colchicine plant chloroplast genome: a infA, b ycf68, and c ycf15. L, L. longiflorum was used as a reference; G, G. superba; C, C. autumnale. The black box with an asterisk represents stop codons

Full size image

The genome size of G. superba was slightly larger than that of C. autumnale (1,154 bp) affected by the presence of the ycf15 gene, and variable lengths of ycf2 and ycf1. Among the available cpDNA sequences of the order Liliales, G. superba had the largest genome size. The A+T and G+C contents in the complete genomes were quite stable and similar among Liliales species (Table 1).

To confirm this hypothetical protein coding gene loss in the tribe Colchiceae of family Colchicaceae, we designed a pair of primers to amplify the region from ycf2 to trnL-CAA (for-TGGATCAAATGACAAAGACA; rev-CTAAAGAGCGTGGAGGTTCG; Fig. 3a). We observed that two different length PCR products, shorter (400–500 bp) and longer (1000–1100 bp) were generated from the same primer set in the Colchicaceae. Sequences of both PCR products indicated that only Colchicum and its alliances had a shorter PCR product and lack ycf15 in their chloroplast genomes against the remaining taxa of tribe Colchiceae (Hexacyrtis, Ornithoglossum, Sandersonia, Gloriosa) and the related tribe Angullarieae (Wurmbea) which had a ycf15 gene on their genomes (Fig. 3b, c). Alignment of this region was shown in more detail, and the location of ycf15 loss was highlighted in Fig. 4.

Fig. 3
figure 3

Analyses of the missing ycf15 gene. a Primer positions were used for amplifying the ycf15 region (psi pseudogenes). b Results of PCR amplification of ycf15. c Modified phylogenetic tree of the Colchicaceae family from Nguyen et al. (2013) result. Bootstrap and posterior probability (PP) values are shown above and below each branch, respectively. The numbers beside taxa name indicate the number of stop codons present in ycf15 sequences. The asterisk beside taxa name indicate the ycf15 gene missing in that species. L, Ladder (100-bp ladder, BIOFACT); W, Wurmbea; G, Gloriosa; S, Sandersonia; O, Ornithoglossum; H, Hexacyrtis; A, Androcymbium; M, Merendera; B, Bulbocodium; C1, C2 and C3, Colchicum

Full size image
Fig. 4
figure 4

Comparison of the sequences of ycf15 and its surrounding region from 11 taxa of tribe Colchiceae. The red letters represent the ycf15 regions. W, Wurmbea; G, Gloriosa; S, Sandersonia; O, Ornithoglossum; H, Hexacyrtis; A, Androcymbium; M, Merendera; B, Bulbocodium; C1, C2 and C3, Colchicum

Full size image

Codon usage

All 78 protein coding genes of C. autumnale and G. superba in the chloroplast genome were encoded by 22,905 and 22,806 codons, respectively (Table 2). The highest codon usage was commonly recorded for isoleucine (especially the ATT codon), and the two genomes shared the same number of stop codons (36—TAA; 24—TAG; 18—UGA). For the start codon, ATG was the most common among coding genes, with the exception of ATC for ndhD, ACG for rpl2, ACT for rps2, GTG for rps19, and ATT for ycf15.

Table 2 Codon usage in Colchicum autumnale and Gloriosa superba chloroplast genomes (excluding pseudogenes)
Full size table

Simple sequence repeats (SSRs) and long repetitive sequences

There was no significant difference in SSRs between C. autumnale and G. superba chloroplast genomes in which 58 and 56 SSRs with lengths of at least 10 bp were found, respectively (Table 3). The majority of mononucleotide repeats were A–T rich, which were common in the two species, while polyG or polyT repeats were rare. Generally, the number of dinucleotide repeat units was slightly higher than those of other repeat units, such as tri-, tetra-, penta- and hexa-nucleotides, observed in the chloroplast genomes of colchicine plants. SSRs were more abundant in noncoding regions than in protein coding genes.

Table 3 Distribution of simple sequence repeats (SSRs) loci in the Colchicum autumnale (Col.) and Gloriosa superba (Glo.) cpDNA
Full size table

Based on the repeat sequences in C. autumnale and G. superba plastid genomes (Table 4), they were divided into three categories including tandem repeat, forward repeat, and palindromic repeat. Three tandem repeats, four forward repeats, and three palindromic sequences were detected in the C. autumnale chloroplast genome. Similar to C. autumnale, one tandem sequence, four forward repeats, six palindromic sequences, and one reverse repeat were found in G. superba cpDNA. By comparing these plastid genomes, three palindromic sequence repeats were identified at the same location namely, ccsA-ndhD, trnH-GUG-psbA, and psbT-psbN regions with similar lengths from 42 to 62 bp. In addition, three palindromic sequences were detected in G. superba at the rrn16-trnI-GAU, accD-psaI, and rpl32-trnL-UAG spacer. Three tandem repeats were found in C. autumnale with a maximum length of 24 bp, and there was only one long tandem repeat (55 bp) in G. superba. We also found the same type of forward repeats (25 bp) at the same location in the IGS trnS-GCU and trnS-UGA in both plastid genomes.

Table 4 Distribution of repetitive sequences in Colchicum autumnale (a) and Gloriosa superba (b) cpDNA
Full size table

Base substitution ratios among Liliales species

Nonsynonymous (dN) and synonymous (dS) substitution rates (dN/dS) are indicators of evolutionary rate and natural selection (Yang and Nielsen 2000). In this report, the substitution ratios among Liliales members were calculated (Fig. 5). We used Dioscorea elephantipes (Dioscoreaceae, Dioscoreales) as a reference instead of Asparagales, because our genomic understanding of Asparagales is restricted to Orchidaceae, which possesses many pseudogenes or lacks the ndh gene (Wu et al. 2009). The majority of genes showed dN/dS ratios of less than 1 (except ycf2). In the SSC region, the highest and lowest ratios were found in ndhD and psaC, respectively. The other ndh genes were stable from 0.1 to 0.3. The highest dN/dS ratios were observed at psbK and rps8 in the LSC and were approximately 0.8–0.9. Additionally, the substitution rates were zero in psbI, psbM, psbJ, psbL, psbE, petG, atpH, all of which are photosynthetic genes, in both C. autumnale and G. superba.

Fig. 5
figure 5

Comparison of nonsynonymous (dN) and synonymous (dS) substitution ratios among eight species of Liliales, using the Dioscorea elephantipes (Dioscoreales) genome as the reference

Full size image

Comparison of IR junction

We compared the IR junction among the Liliales (Fig. 6) and found variable gene composition especially in the IR-LSC junction. However, C. autumnale and G. superba chloroplast genomes showed a similar pattern among them. The IRa-LSC junction was located at a part of rps19 in both genomes, and ndhF and ycf1 gene was overlapped at IRb-SSC junction (Fig. 6).

Fig. 6
figure 6

Comparison of the IR boundaries among the species of Liliales. Psi pseudogenes

Full size image

Discussion

First records of the complete chloroplast genome in colchicine medicinal plants

We firstly analyzed the complete chloroplast genomes of two colchicine plants, C. autumnale and G. superba, and discussed about evolutionary implication and the probability for usefulness of this data for identifying the related taxa from the result. Both chloroplast genomes were very similar in the structure except the gene loss of ycf15 in C. autumnale.

The gene infA encodes translation initiation factor 1, which has been lost completely or is present as a pseudogene in the majority of angiosperm (Millen et al. 2001). In this study, we identified two and three internal stop codons in the infA sequence of C. autumnale and G. superba chloroplast genomes, respectively, suggesting that infA is a pseudogene (Fig. 2a). This has also been observed in the chloroplast genomes of Lilium longiflorum (Kim and Kim 2013) and Veratrum patulum (Do et al. 2013). However, compared with other Liliales members, the infA gene was absent in Smilax china (Liu et al. 2012) and A. aurea (Kim and Kim 2013). In contrast, the complete sequence of the infA gene without any internal stop codons was present in Chionographis japonica (Bodin et al. 2013) and Paris verticillata (Do et al. 2014). Therefore, further studies on the evolution of the infA gene in Liliales are required to improve our understanding of this observation. We also observed that the two hypothetical coding genes of ycf68 and ycf15 were truncated by five and three internal stop codons, respectively, in G. superba (Fig. 2b, c). This situation was not only detected in G. superba, but also in L. longiflorum (Kim and Kim 2013), C. japonica (Bodin et al. 2013), and V. patulum (Do et al. 2013) of Liliales and in Cymbidium of Asparagales (Yang et al. 2013). However, in C. autumnale cpDNA, ycf68 was present as a protein coding gene without internal stop codon (Fig. 2b). This in-frame sequence of this gene was present and could represent a functional protein coding gene in the Nymphaeales (Raubeson et al. 2007) and in the gymnosperms P. thunbergii (Wakasugi et al. 1994) and P. koraiensis (AY228468). ycf15 was absent in the C. autumnale chloroplast genome (Fig. 4). The role of ycf15 as a protein coding gene remains unclear and requires further study.

ycf15 gene loss reflecting the phylogenetic relationship of colchicine plants

Remarkable difference of gene content and genome size between Gloriosa and Colchicum was the loss of ycf15. Using the expanded sampling, we confirmed that ycf15 gene loss was occurred only in the expanded Colchicum genus including Colchicum, Androcymbium, Bulbocodium and Merendera, whereas it is still remained in Ornithoglossum, Sandersonia, Hexacyrtis and Gloriosa, as well as Wurmbea (tribe Anguilarieae) (Fig. 3c). As a result, missing ycf15 may be an effective molecular marker for distinguishing this expanded genus recognized Vinnersten and Manning (2007) from the Colchicaceae family. Four closely related genera of Colchicum, Androcymbium, Merendera and Bulbocodium have been considered to be combined into one genus Colchicum sensu lato based on molecular evidence (Vinnersten and Manning 2007; Nguyen et al. 2013). However, until now this suggestion has not been completely agreed (del-Hoyo and Pedrola-Monfort 2008). Therefore, a finding of specific deletion of ycf15 on chloroplast genome provides important evidence at molecular level to support their phylogenetic relationship and circumscription.

Implications of microsatellites in both chloroplast genomes

SSRs are short repeat motifs (of at least 10 bp) found in DNA sequences (Katti et al. 2001; Shanker et al. 2007). Identification of SSRs is important to develop molecular markers and to map traits of economic, medical, or ecological interest. Early studies on SSRs of chloroplast genome in rice showed that they are abundant in non-coding regions through the genome (Rajendrakumar et al. 2007). In the G. superba chloroplast genome, multiple repeat sequences are found in the trnC-GCA-petN intergenic spacer and ycf1 gene (three repeat sequences for each). Similarly, in the C. autumnale plastid genome, the clpP intron 2 and ycf1 possess three and five repeat units, respectively. In addition, these colchicine plants chloroplast genomes share several SSRs distributed in the same region. For example, mononucleotides were found in trnK-UUU-rps16, rpoB-trnC-GCA, overlapped ycf1-ndhF, ndhF-rpl32, and ycf1; dinucleotides were detected in ndhH and ycf2, and trinucleotides (TCC) were found in the rpoA gene (Table 3). It is interesting that both chloroplast genomes shared several chloroplast SSRs co-located in noncoding regions and it may be variable in other Colchicaceae species. This information of SSRs can be used as a reference when the molecular identification marker development and application in the population genetics are tried in future studies.

IR expansion and its evolutionary implications in Liliales

The boundaries of IR regions with LSC and SSC play a critical role in expansion and contraction of the genome; thus, the gene contents of IRs vary significantly (Goulding et al. 1996; Plunkett and Downie 2000). Eight complete plastid genome sequences including the data generated in the present study are available currently and are represented from five families in the order Liliales, namely, C. autumnale and G. superba (Colchicaceae), A. aurea (Alstroemeriacae), L. longiflorum (Liliaceae), S. china (Smilacaceae), C. japonica, V. patulum and P. verticillata (Melanthiaceae). The IR boundaries between the C. autumnale and G. superba chloroplast genomes showed a similar pattern (Fig. 6). The IRB-LSC junction is located within the sequence of rps19 in both genomes, and part of the rps19 gene was copied in IRA and its 3′ end adjacent psbA in LSC, which is common in the typical monocot chloroplast genome structure. This junction was similar to A. aurea, which shares a close relationship with the Colchicaceae family and L. longiflorum. However, it was expanded to include the whole rps19 gene and part of rpl22 in S. china. In case of family Melanthiaceae, which is a sister clade of Liliaceae and Smilacaceae, it was varied within the family. In V. patulum, it was located in the intergenic spacer between rps19 and trnH-GUG (Fig. 6). In contrast, it was expanded to part of rps3, and the entire rps19 and rpl22 genes in the chloroplast genomes of C. japonica and P. verticillata. Comparative studies among cpDNAs from eight species in Liliales indicated that they share the same structures and gene orders in the SSC region and IR-SSC junction (JLA, JLB) at the position of ycf1 region. However, the overlapped length between the ycf1 pseudogene and ndhF (JSB) varied among the species; e.g., C. autumnale (39 bp), G. superba (53 bp), A. aurea (78 bp), L. longiflorum (17 bp), and C. japonica (5 bp). The overlap of these genes was also observed in the other chloroplast genomes of angiosperm species (Yang et al. 2010).

In addition, the trnH-GUG-psbA spacer region located in the junction between IRA and LSC is common in monocots. The trnH-GUG-psbA chloroplast intergenetic spacer region has been used in many DNA barcoding studies (Pang et al. 2012) and has become a frequently used marker for molecular phylogenetic studies at the lower taxonomic level (Degtjareva et al. 2012). However, this study showed various gene compositions in these regions among the species of Liliales. Therefore, the trnH-GUG-psbA intergenetic spacer may not be a good candidate for resolving phylogenetic relationships (at least in this order), because this spacer showed significantly different gene contents and lengths, which cause the complicate alignment and analysis.

Author contribution statement

Nguyen PAT carried out the analyses and drafted the manuscript. Kim JS participated in the design and coordination of the study, advised to analyze the data, and drafted the manuscript. Kim JH participated in the coordination of the study and helped to draft the manuscript. All authors read and approved the final manuscript.