Introduction

DNA barcoding is a rapid, accurate means of taxonomic identification using one or a few short, standardized DNA region(s)1,2. After one and a half decades of effort, CO1 was selected as the standard barcode for the identification of animal species2, and rbcL + matK as core barcodes for plants3. The internal transcribed spacer (ITS) sequence was also recommended as a complementary marker for plant species identification4. However, the recommended three-barcode system has proven not to work well for the identification of plants used in traditional herbal medicine, since rbcL and matK are very difficult to amplify from the available material, especially powdery processed products, in which DNA degradation is very common. Chen et al.5 therefore compared seven proposed candidate DNA barcodes and suggested that the internal transcribed spacer 2 (ITS2) region of nuclear ribosomal DNA would be the most suitable for barcoding medicinal plant species. Because of its high rate of successful amplification and discrimination power, ITS2 has been successfully used to identify several traditional herbal medicines in recent years6,7,8. However, successful identification should be based on a comprehensive DNA barcode database, which should include samples from all of the geographic sites of a target species’ distribution.

Rhodiola L. (Crassulaceae) consists of about 55 species that are mainly found at high altitudes and in cold regions of the Northern Hemisphere9. There are about 50 Rhodiola species recognized in China (16 of which are endemic), which are concentrated in the western alpine regions (i.e., the Hengduan Mountains and the Qinghai-Tibetan Plateau (QTP))9. Species of this genus are herbaceous perennials that often grow on gravel-covered slopes or in the cracks of exposed rocks at altitudes of approximately 3500–5000 m, making sample collection and study notoriously difficult. Rhodiola species, which have historically been used as adaptogens in Russia and northern Europe, are widely recognized for their ability to enhance human resistance to stress or fatigue and to promote longevity10,11,12. Rhodiola dietary supplements are listed as “good for handling physical and mental stress” in the consolidated list of Article 13 health claims of the European Food Safety Authority (EFSA)13. In China, Rhodiola species known as Hongjingtian have been frequently used as adaptogens, hemostatics, and tonics in traditional Tibetan medicines for hundreds of years12.

The roots and rhizomes of R. crenulata (J. D. Hooker & Thomson) H. Ohba have been included in the Pharmacopoeia of China14 as the authentic Rhodiolae Crenulatae Radix et Rhizoma, where they are listed as enhancing inner spiritual power, concentration, and physical endurance. This species is exclusively found in rock crevices on mountain peaks from 4300 m to 5600 m in elevation9, which makes it one of the highest living vascular plants in the QTP area. Since the 1980s, the accelerated and uncontrolled use of R. crenulata in China has severely reduced its population. Population genetic studies with inter-simple sequence repeat markers have demonstrated the low genetic diversity of this species, and it has been included in the National Class One Endangered Species in China checklist for conservation purposes15. The rapid increase in the commercial value of R. crenulata raw material has led to other Rhodiola species being sold as counterfeit Rhodiolae Crenulatae Radix et Rhizoma.

The type species R. rosea L. is a popular traditional medicinal plant in east Europe and Asia, with a reputation for stimulating the nervous system, decreasing depression, enhancing work performance, eliminating fatigue, and preventing high-altitude sickness16. Rhodiola rosea is recorded as Rhodiola rosea Root and Rhizome in the United States Pharmacopeia, and potentially confounding materials include R. kirilowii, R. yunnanensis, R. crenulata, R. sacra, and R. sachalinensis17. Rhodiola rosea products are sold by 46 companies worldwide, and the rapidly increasing demand for raw materials has caused R. rosea to become a threatened plant in many countries18. Thus, some substitution of other species appears in the market. In the Altay region of Russia alone, there are 24 different species of the genus which can be misclassified as R. rosea19.

A recent phylogenetic study of Rhodiola revealed significant convergent evolution of important morphological characters, such as dioecy and marcescent flowering stems20. Consequently, many of the previously defined infrageneric taxa are not monophyletic. Historical biogeographic studies have suggested that rapid radiations occurred in the evolution of this genus21. These evolutionary processes have made the taxonomy of Rhodiola notoriously difficult to evaluate. The difficulty is greater still when dealing with herbal medicines, since the officinal parts of Rhodiola species are roots and rhizomes, which lack diagnostic morphological characters. Thus, identification of Rhodiola materials based solely on morphology is problematic. DNA barcoding provides an alternative method to identify Rhodiola materials in the market. However, for powdery processed products, in which DNA degradation is very common, as the case in Rhodiola products, a shorter barcode such as ITS2 should be used. Yet we still don’t know the identification efficient of ITS2 barcode in Rhodiola. Besides, the successful implementation of DNA barcoding must be based on a comprehensive barcode database, which is lacking for Rhodiola. Our goal is thus to test the suitability of ITS2 as a barcode for identification of Rhodiola species, and to subsequently establish a comprehensive ITS2 barcode database for Rhodiola based on extensive sampling of R. crenulata and R. rosea, as well as other species in the genus.

Results

Amplification and sequencing

The primers for the ITS barcode were found to successfully amplify DNA from all 47 species of Rhodiola, and we generated 342 new ITS barcodes. We sequenced 253 R. crenulata individuals from 16 populations across its distribution area. We detected 6 haplotypes in these individuals. In R. rosea, we detected 3 haplotypes in 98 individuals. Newly generated sequences were deposited in the GenBank database with the accession numbers shown in Appendix I. In our sampling test, the sequencing success rate was 100% for the ITS sequences and 98% for the ITS2 primers.

Sequences characteristic and barcoding gap

The aligned length of the ITS barcode sequences was 670 bp, and the length of ITS2 barcode was 235 bp after annotation. Length variation exists in both the ITS and ITS2 regions, with a range of 600–642 bp for the former and 209–223 bp for the latter. The number of informative sites and variable sites are shown in Table 1. The original data set was analyzed to calculate the intra- and inter-specific distances. The resulting six parameters (average inter-specific distance, minimum inter-specific distance, theta prime, average intra-specific distance, coalescent depth, and theta) are shown in Table 2. The barcoding gaps between intra- and inter-specific distances are shown in Fig. 1. There was overlap between inter- and intra-specific distance in both ITS and ITS2. The median and Wilcoxon two-sample tests showed that the intra-specific distance was always significantly lower than the inter-specific distance for both ITS and ITS2 (Table 3).

Table 1 Sequence information for ITS and ITS2 data sets used in this study.
Full size table
Table 2 Analysis of inter-specific divergence and intra-specific variation of the ITS and ITS2 sequences in samples of 47 Rhodiola species.
Full size table
Figure 1
figure 1

Relative distribution of inter-specific divergence between congeneric Rhodiola species and intra-specific variation in the ITS (a) and ITS2 (b) regions using K2P genetic distance.

Full size image
Table 3 Results of the median and Wilcoxon two-sample tests based on interspecific versus intraspecific Kimura 2-parameter distances for each barcode
Full size table

The Efficacy of ITS2 for Authentication

The efficacy of the ITS and ITS2 barcodes for authentication was evaluated by both BLAST1 and the nearest distance method, as shown in Table 4. We found that 66.0% and 57.5% of the species were successfully identified as the best hit of a BLAST search using ITS and ITS2 as barcodes, respectively. The results for the nearest distance method are similar to those of the BLAST1 method. Thus, both ITS and ITS2 have a low authentication efficacy. However, in the case of R. crenulata, all six haplotypes could be exactly ascribed to the right species using both BLAST1 and the nearest distance method.

Table 4 Comparison of authentication efficiency for ITS2 using different methods.
Full size table

Monophyly tests of species based on phylogenetic trees

In tests of monophyly based on phylogenetic trees, the unweighted pair group method with arithmetic mean (UPGMA) analysis provided the highest discriminative power, followed by neighbor joining (NJ), and maximum parsimony (MP) analyses. ITS provided a species identification rate of 66%, while ITS2 had a success rate of 57.5%. A tree constructed by the Bayesian method based on ITS2 sequences is shown in Fig. 2. Samples from R. crenulata and R. rosea each formed a well-supported monophyletic clade.

Figure 2
figure 2

NJ tree constructed by MEGA 6 based on ITS2 sequences for Rhodiola species. Numbers above branches indicate bootstrap values of the NJ and maximum parsimony analyses. The width of line indicates the level of support value.

Full size image

Analysis of SNPs

We tried to detect species- or even population-specific SNPs through SNP analysis, which has been used to identify plant species in other studies21,22. We used the entire data set to detect the SNPs at the interspecific level. In the ITS2 data, we found three stable variation sites at positions 141, 152 and 159 (Table 5) which can be used in combination to distinguish R. crenulata from other species. However, we found no stable species-specific sites identifying R. rosea.

Table 5 Three stable SNPs in ITS2 sequences from Rhodiola crenulata and its closely related species.
Full size table

Analysis of the Secondary Structure of the ITS2 region

We used six haplotypes of R. crenulata and its closest relatives (R. gelida and R. coccinea) and one haplotype of R. rosea and its close relatives (R. atuntsuensis and R. tangutica) to predict the secondary structure of ITS2. All ITS2 secondary structures exhibited four similar helices: Helix I, II, III, IV (Figs 3 and 4). The secondary structure of the six R. crenulata haplotypes was similar but distinct from those of its close relatives in the position, size, and number of loops on Helix II and IV, which can be used as molecular morphological characteristics to identify R. crenulata (Fig. 3). The secondary structures of ITS2 in R. rosea and its related species were similar (Fig. 4).

Figure 3
figure 3

Secondary structure of ITS2 in Rhodiola crenulata and its adulterants. (a-f) six haplotypes of R. crenulata. (g) R. gelida; (h) R. coccinea.

Full size image
Figure 4
figure 4

Secondary structure of ITS2 in Rhodiola rosea and its adulterants. (a) R. rosea; (b) R. tangutica; (c) R. atuntsuensis.

Full size image

Discussion

DNA barcoding provides a technique that can improve descriptive taxonomy and help ensure that organisms used for various scientific, commercial, and medicinal applications are correctly identified. After more than a decade of effort, substantial progress has been made in selecting markers for barcoding. Standard markers for animals, plants, and other organisms have been established3,5,23. The next step for DNA barcoding is to construct complete barcode databases to assist in applications, because accurate species identification—assignment of an unknown to a known—requires a comprehensive comparative molecular database against which the unknowns can be compared.

To our knowledge, this is the first time that the ITS2 region has been used to identify Rhodiola species in such a large sample, an endeavor that has expanded the use of the ITS2 region to include medicinal plants. Extensive sampling and analysis of taxonomically well-understood groups is needed to thoroughly validate and standardize the markers and procedures used for DNA barcoding24. A partial barcode database will lead to mistakes in the identification process, especially in groups that have experienced complex evolutionary history. In the present study, we focused on two recorded medicinal plants of the genus Rhodiola: R. crenulata (Rhodiolae Crenulatae Radix et Rhizoma) and R. rosea (Rhodiola rosea Root and Rhizome). We sampled 253 R. crenulata individuals from 16 populations, covering its entire distribution area, from the Hengduan Mountains to the platform of the QTP. We also included 98 R. rosea individuals from 11 populations, representing samples from Asia, North America, and Europe. Only by using such a comprehensive sample to create the barcode database could we ensure that no intraspecific variation would be missed. If an incomplete database was used, future samples of R. crenulata or R. rosea would not be correctly identified, as there would be no corresponding sequences in the database.

A recent study used DNA barcoding to survey the species composition of commercial Rhodiola products25. They collected 100 Rhodiolae Crenulatae Radix et Rhizoma decoction piece samples and found that only 40% of them were authentic R. crenulata. However, their barcode database only included 82 voucher samples from 10 Rhodiola species. There are 55 species in the genus; 10 species therefore represents only 1/6 of the species diversity. Thus, a more comprehensive ITS2 barcode database is needed. In the present study, we included 47 species, covering all the morphological variation and infrageneric taxa described by previous taxonomists9,26. The database also includes some other medicinal Rhodiola species which are used by local people but are not recorded in the pharmacopeia: R. serrate, R. fastigiata, R. himalensis, R. henryi, and others. Using our database, barcoding can not only test whether a sample is authentic but also determine which species the adulterant belongs to.

It has been reported that some species of Rhodiola contain cyanogenic glycoside compounds, such as lotaustralin, epilotaustralin, and rhobupcyanoside A26. After contact with β-glucosidase and α-hydroxynitrile lyase, these compounds degrade and release hydrogen cyanide, which is harmful to humans. Species containing these compounds include R. gelida, R. quadrifida, R. concinna, R. kirilowii, R. litwinowii, R. heterodonta, and R. bupleuroides. Using our database, one can easily determine whether commercial products of Rhodiola include these species.

ITS2 is better than other candidate DNA barcodes, such as the core marker suggested for plants, matK + rbcL, since it is more universal, has less intraspecific variation but higher interspecific divergence, and is shorter. It was therefore recommended by Chen et al.5 as a standard barcode to identify medicinal plant species. In their analysis, 90.3% of their 6600 samples could be assigned to right species using the ITS2 barcode. The efficiency of ITS2 as a barcode for medicinal plants was also evaluated in other case studies in various plant groups7,27,28. Our results confirm some of the previous conclusions: ITS2 is easy to amplify (98% amplification success rate), and the inter-specific genetic distances are significantly larger than intra-specific distances (Tables 2,3). However, an obvious overlap between intra- and inter-specific distance was observed (Fig. 1), and the authentication efficacy of ITS2 as assessed by BLAST1, the nearest distance method, and a tree test (Table 4 and Fig. 2) was much lower than has been reported in other studies. Zhang et al.29 showed that the core barcode combination matK and rbcL could only discriminate 40.4% of their sampled Rhodiola species, while ITS alone could identify 66.0% of the species, according to the phylogenetic tree test. After expanding our sample by adding 253 R. crenulata and 98 R. rosea individuals, the discrimination power of ITS was still 66%. In the ITS2 data set, phylogenetic tree analysis showed that ITS2 could discriminate 57.5% of the sampled species, less than ITS, which is not unexpected. The discrimination power of ITS2 in Rhodiola species is lower than in other groups7,27. This result is partially caused by the complex evolutionary history of the genus. Rhodiola species have undergone rapid radiation triggered by extensive uplift of the Qinghai-Tibetan Plateau21. Rhodiola species are difficult to identify by morphological characters, not only because their taxonomic treatment is somewhat chaotic but also because most species share similar morphology9,26. Using DNA barcodes to identify species has also been problematic because of the group’s complex evolutionary history, which includes interspecific hybridization, introgression, allopolyploidy, mixtures of sexual and asexual reproduction, and recent divergences29,30,31.

Although we found a lower discrimination power using the ITS2 barcode than has been reported in other studies, it is still sufficient to identify the two most widely used medicinal Rhodiola plants, R. crenulata and R. rosea. Since most R. crenulata individuals possessed the same sequences, we analyzed haplotypes of the species. In total, we detected 6 haplotypes, which formed a well-supported clade in the phylogenetic tree test (Fig. 2). The intraspecific distance ranged from 0 to 0.003. The low level of intraspecific divergence is probably due to evolutionary history: during the Pleistocene glacial, the species shrank its distribution area to the Hengduan Mountain as a refuge, then expanded very rapidly to the Qinghai-Tibetan Plateau platform after the glacial period. Due to founder effects, the genetic diversity of the populations on the QTP is very low, and we found only one haplotype in most of the populations. We tried to detect species- or even population-specific SNPs in our SNP analysis, a method which has been used to identify plant species in other studies22,32. Three stable sites specific to R. crenulata were found. These could be used as a nucleotide signature for the identification of Rhodiolae Crenulatae Radix et Rhizoma, though it should be stressed that these three sites should be used in combination. In addition, secondary structure analysis revealed differences between the haplotypes of R. crenulata and its closest relatives, R. gelida and R. coccinea (Fig. 3). Thus, secondary structure could be used as an alternative mini-barcode when a full-length ITS is not present. In R. rosea, we only detected three haplotypes in the ITS data and only one haplotype in the ITS2 data set. However, this is sufficient to correctly identify this species.

Although the ITS region was almost three times as long as the ITS2 region in the present study (670 bp vs 235 bp), the authentication efficacy of ITS2 assessed by BLAST1, the nearest distance method, and a tree test was not significantly lower. In practice, Chinese patent medicine samples which need to be identified are always in processed form, mostly powder or mixed with other ingredients. In these circumstances, the overall ITS region could be difficult to amplify and sequence because of DNA degradation5. Our results demonstrate that both a mini barcode from ITS2 and the structure of ITS2 are effective markers for the identification of R. crenulata and R. rosea based on a survey of representatives from throughout its distribution area, thus greatly broadening the application of DNA barcoding in identifying Chinese patent medicines and other products with degraded DNA.

DNA barcoding could play an important role in identifying traditional Chinese medicines and in drug safety monitoring. In addition, the use of DNA barcoding for market supervision can broaden the application of this technology and provide an excellent inspection method for supervisory institutions. In some studies, DNA barcoding has been successfully applied to supervise medicine markets. Wu et al.33 used barcoding to identify contamination of Aristolochiaceous herbs, offering a method to protect consumers from health risks caused by aristolochic acids. Zhao et al.8 used ITS2 as a barcode to identify Acanthopanacis cortex, and detected five adulterants in nine samples purchased from drug stores and medicine markets. In this study, we constructed a comprehensive DNA barcode database to identify Rhodiola products. Based on this database and a standard DNA barcoding procedure, it is possible to survey Rhodiola dietary supplements sold in markets. As R. crenulata and R. rosea are threatened or endangered in many countries, herbalists may substitute cheaper, more readily available species for these rare ingredients and sell them under the same name. Quality control of raw materials is key to solving this problem. An effective strategy to certify the origin of raw materials and detect adulterants in the herbal medicine markets is needed. DNA barcoding, with the support of a comprehensive barcode database, may help realize this goal by providing a means to monitor and authenticate the species found in raw materials throughout the industrial pipeline.

Methods

Plant Materials

We collected a total of 518 samples of Rhodiola species representing 47 of the 55 species in China from the provinces of Xizang, Qinghai, Gansu, Sichuan, Yunnan, and Xinjiang. We also included samples of R. rosea from North America and Europe (See Appendix I for collection information and NCBI accession numbers). Fresh leaves were dried in silica gel upon collection. For each taxon, three to six individuals were sampled to capture intra-specific diversity and cover most of the species’ respective geographic ranges. For species that are used for traditional medicine (i.e., R. rosea, and R. crenulata), we collected 98 and 253 individuals, respectively. These individuals represent the full range of the two species’ natural distribution area. Voucher specimens of the collected taxa were deposited in the Herbarium of Peking University (PEY) and Herbarium of Shaanxi Normal University (SANU).

DNA extraction, amplification, and sequencing

Genomic DNA was isolated from approximated 15 mg each leaf sample following the CTAB protocol34. Amplification and sequencing primers were ITS-1 and ITS-4 for the ITS35 region (including ITS1, 5.8 S and ITS2). Polymerase chain reaction (PCR) amplification was carried out in a total volume of 20 μL containing 2 μL 10 × buffer, 0.5 μM of each primer, 200 μM of each dNTP, 1 U of Taq polymerase (TianGen Biotech, Beijing, China), and 1 μL template genomic DNA. The amplification followed a program of 5 min at 95 °C; 35 cycles of 1 min at 95 °C, 1 min at 56 °C, and 1 min at 72 °C; and a final extension of 5 min at 72 °C. Direct sequencing was conducted using the same primers as in the amplification. The PCR fragments were purified using polyethylene glycol (PEG) precipitation36. Cycle sequencing was conducted using the BigDye 3.1 reagents with an ABI 3730 automated sequencer (Applied Biosystems, Foster City, California) in Sangon Biotech Corporation (Shanghai, China). To test the utility of ITS2 primers designed by Chen et al.5, we also performed a PCR experiment in a random sample of 100 individuals without sequencing.

Sequence analysis

Contigs were assembled and edited using ContigExpress (a component of Vector NTI Suite 6.0, InforMax). Sequences were aligned using MUSCLE 3.8.3137, followed by manual adjustments in Geneious 9.138. The ITS2 region was delimited based on the GenBank annotation or hidden Markov models (HMMs), which were constructed through a web based tool (http://its2.bioapps.biozentrum.uni-wuerzburg.de). In the following analyses, ITS and ITS2 sequences were analyzed separately. The Kimura 2-parameter distance (K2P distance) for the two DNA regions was calculated in MEGA v. 6.039 to estimate intra- and inter- specific divergence. Three parameters were used to characterize inter-specific divergence24,40: average inter-specific distance between all species; theta prime (i.e. the mean pairwise distance between all samples); and the smallest inter-specific distance. In addition, we used three parameters to determine intra-specific variation: average intra-specific difference between all samples collected within each species; theta, the mean pairwise distance within each species with at least two representatives; and coalescent depth, the maximum intra-specific distance. We used Wilcoxon two-sample tests and median tests (http://www.fon.hum.uva.nl/Service/Statistics.html) to assess the significance of divergence. The barcoding gaps were graphed by comparing the distribution of intra- and inter-specific divergence for each candidate barcode. We used BLAST1 and the nearest distance methods to evaluate the species authentication efficacy5,27,41. In the BLAST1 method, we first constructed an ITS2 database using the BLAST program (http://blast.ncbi.nlm.nih.gov/Blast.cgi), and then used all of the Rhodiola ITS2 sequences as query sequences to search the database. Correct identification occurred when the best BLAST hit of the query sequence was from the expected sequences, while ambiguous identification was when the best BLAST hits for a query sequence were those of several species, including the expected species, and incorrect identification was when the best BLAST hits of the query sequence did not include the expected species. In the nearest distance method, correct identification was when the hit in our database based on the smallest genetic distances was from the query species; ambiguous identification was when several hits from our database were found to have the same smallest genetic distance to the query sequence; and incorrect identification was when the hit based on the smallest genetic distance was not from the expected species.

Several tree-building methods were used to test whether individuals from the same species form a monophyletic group in phylogenetic trees based on different built using the barcodes, tree-based methods were used to display the molecular identification results. Neighbor-joining (NJ) and unweighted pair group method with arithmetic mean (UPGMA) trees were reconstructed using MEGA v. 6.0 with the K2P model39. PAUP v. 4.0b1042 was used to generate a maximum parsimony (MP) tree with a heuristic search strategy followed by random addition starting trees with tree-bisection-reconnection (TBR) branch swapping and MulTrees selected. Gaps were treated as missing data and all characters were weighted equally. Support for individual nodes was assessed using bootstrap values43. Parsimony bootstrap (PB) values were obtained from 1,000 replicates of heuristic searches as described above (TBR branch swapping and MulTrees selected), but with branch swapping limited to 10,000,000 rearrangements per replicate due to memory constraints. Nucleotide substitution model parameters were determined using the Akaike information criterion (AIC) in Modeltest version 3.744,45.

In attempting to identify the origin region of R. crenulata and R. rosea samples, we also used the ITS2 data set to detect single-nucleotide polymorphisms (SNP). If we could find a SNP specific to a geographic region, then it could be used not only to identify the species of a sample but also where it was collected.

Secondary structure analysis

To infer the usefulness of secondary structure of ITS2 for the identification, we chose six haplotypes representing R. crenulata and one haplotype representing R. rosea, as well as their closely related species for secondary structure prediction using the ITS2 Workbench (http://its2.bioapps.biozentrum.uni-wuerzburg.de/)28.