Certain species within the genus Panax L. (Araliaceae) contain pharmacological precious ginsenosides, also known as ginseng saponins. Species containing these compounds are of high commercial value and are thus of particular urgency for conservation. However, within this genus, identifying the particular species that contain these compounds by morphological means is challenging. DNA barcoding is one method that is considered promising for species level identification. However, in an evolutionarily complex genus such as Panax, commonly used DNA barcodes such as nrITS, matK, psbA-trnH, rbcL do not provide species-level resolution. A recent in silico study proposed a set of novel chloroplast markers, trnQ-rps16, trnS-trnG, petB, and trnE-trnT for species level identification within Panax. In the current study, the discriminatory efficiency of these molecular markers is assessed and validated using 91 reference barcoding sequences and 38 complete chloroplast genomes for seven species, one unidentified species and one sub-species of Panax, and two outgroup species of Aralia L. along with empirical data of Panax taxa present in Vietnam via both distance-based and tree-based methods. The obtained results show that trnQ-rps16 can classify with species level resolution every clade tested here, including the highly valuable Panax vietnamensis Ha et Grushv. We thus propose that this molecular marker to be used for identification of the species within Panax to support both its conservation and commercial trade.

DNA barcode, Panax genus, Panax vietnamensis, petB, trnE-trnT, trnQ-rps16, trnS-trnG

The genus Panax L. is well-known in culinary and medicinal traditions in many countries including China, Korea, Japan, and Vietnam. Its species produce ginsenosides, also known as ginseng saponins with strong antioxidant, antidiabetic, antitumor, and neuroprotective activities (Jung et al. 2017; Patel and Rauf 2017). Due to their high commercial demand, many species within Panax have been over-harvested and are at risk of extinction (Case et al. 2007; McGraw et al. 2013; Manzanilla et al. 2018). In Vietnam, there are three species of Panax including Panax vietnamensis Ha et Grushv., Panax stipuleanatus H.T.Tsai et K.M.Feng, and Panax bipinnatifidus Seem., all recorded and classified as endangered. Of these species P. vietnamensis is endemic to Vietnam and is considered to have the highest medical potential and is therefore the most commercially valuable (Nguyen 2005; Nguyen et al. 2007). Two varieties of P. vietnamensis, Panax vietnamensis var. fuscidiscus K.Komatsu, S.Zhu et S.Q.Cai and Panax vietnamensis var. langbianensis N.V.Duy. V.T.Tran et L.N.Trieu, are also present in Vietnam (Phan et al. 2013; Nong et al. 2016). Recently, an unidentified sample of Panax discovered on Puxailaileng Mountain of Nghe An Province was also reported by Phan et al. (2014) and referred hereafter as “Panax sp. Puxailaileng”.

Historically, morphological methods have been used to identify ginseng species, though this is challenging due to how similar different ginseng species can appear. Incorrect identification can lead to unintentional or intentional mislabeling and adulteration with low-quality ginsengs, and ultimately affect the consumers’ health and damage the providers’ integrity. Recently, molecular methods have been shown to be efficient for solving problems related to species identification. However, the most commonly used barcoding sequences are challenging to use in the genus Panax, because these often lack sufficient variability to unambiguously identify the species (Komatsu et al. 2001; Janzen et al. 2009; Hollingsworth et al. 2011; Zuo et al. 2011; Li et al. 2015). According to Zuo et al. (2011) rpoC1, rbcL, and rpoB were the low discriminatory with only four to eight variable sites. The region psbK-psbI had the higher discriminatory ability but low chance of successful sequencing (Janzen et al. 2009). Furthermore, psbA-trnH sequence analysis was reported inaccurate because of the complicated microevolution (Li et al. 2015). The attempts using above loci along with matK, trnD and ycf1 for identifying species in the genus Panax also were unable to completely solve the challenge (Komatsu et al. 2001; Shi et al. 2015). Other newly proposed InDel (Nguyen et al. 2017) and dCAPS markers (Nguyen et al. 2020) were also developed but the requirement for multi-locus analyses is time-consuming and labor-intensive. This raises the need to develop better molecular markers for identification of the species within Panax (Shneyer 2009; Li et al. 2015; Manzanilla et al. 2018).

Previously performed an in silico analysis indicated that the chloroplast DNA markers trnQ-rps16, trnE-trnT, petB, and trnS-trnG had high species identification potential within the genus Panax (Manzanilla et al. 2018) and could be used in routine classification processes. In the present study, we assess and experimentally test the discriminatory efficiency of these commonly used and novel chloroplast markers in classifying species of Panax with an emphasis on those distributed in Vietnam.

Phylogenetic studies on Panax using different DNA barcodes, different reference sequences or samples have resulted in conflicting tree topologies and clade placements for several species (Wen and Zimmer 1996; Komatsu et al. 2001; Zuo et al. 2011; Ali et al. 2012; Le et al. 2017; Manzanilla et al. 2018). For example, two sister species P. bipinnatifidus and P. stipuleanatus were separated into two distant clades based on the combined dataset of six markers by Zuo et al. (2011). Meanwhile, these two above taxa could not be separated into two clades using the ITS region alone (Wen and Zimmer 1996). This inability to resolve a clear sister relationship between these two species was also supported by the four novel markers assessed in our study. Similarly, P. vietnamensis was reported to belong to the same clade as Panax japonicus var. major (Burkill) C.Y.Wu et Feng, Panax pseudoginseng subsp. himalaicus H.Hara based on 18S rRNA and matK (Komatsu et al. 2001), and to be closely related to Panax zingiberensis C.Y.Wu et Feng and Panax wangianus S.C.Sun based on ITS2 (Ali et al. 2012), P. notoginseng based on ITS, matK, rbcL, psbA-trnH, and 18S rRNA (Le et al. 2017), as well as P. japonicus based on in silico data of four potential markers (Manzanilla et al. 2018). With regard to an unidentified sample Panax sp. Puxailaileng, our phylogenetic trees based on comprehensive datasets of the marker trnQ-rps16 or combined markers 1+3+4 (Fig. 4) obviously revealed this taxon belonged to the same clade with all samples of P. vietnamensis. Relevant bootstrap values at 90–99 by Maximum Likelihood method indicated that confidence intervals were eligible for genetic correlation of these samples. The obtained results on these novel markers are congruent with or different from previous studies. Panax sp. Puxailaileng was suggested to be P. vietnamensis based on its morphological characteristics and ITS-rDNA sequence, though further studies are still needed to unambiguously resolve its identity (Tran et al. 2016). Similarly, morphology and molecular-based phylogenetic analyses suggested Panax sp. Puxailaileng found in the wild in Ky Son District, Nghe An Province were P. vietnamensis var. fuscidiscus (Pham et al. 2020). In another study, samples of Panax sp. collected from Puxailaileng Mountain were reported to be closely related to P. stipuleanatus based on the commonly use markers ITS-rDNA and matK (Vu et al. 2020). Present results based on empirical data support the results of the in silico study by Manzanilla et al. (2018). P. vietnamensis is closely related to P. japonicus. This apparent contradiction with other studies might result from differences in dataset structures, the number of species and taxa included, and classification methods. Indeed, different approaches can return different results in DNA barcoding analyses (Wen and Zimmer 1996; Komatsu et al. 2001; Zuo et al. 2011; Ali et al. 2012; Le et al. 2017; Manzanilla et al. 2018). The genetic distances calculated in MEGAX suggest that barcoding gaps exist in the markers rbcL, trnQ-rps16, trnE-trnT, and all concatenated markers, whereas TaxonDNA showed overlap between intra and interspecific distances in all analyzed markers (Table 3, Fig. 2). This incongruence could make a tremendous difference in the output of the analyses. Nevertheless, results obtained from the BM/BCM module in TaxonDNA are similar to distances calculated in MEGAX. In general, the discriminatory power assessed in TaxonDNA is higher than in the tree-based method mPTP (Fig. 5). Especially for rbcL, distance-based methods provide the highest percentage of correctly identified species while the tree-based method gave the lowest percentage of correctly identified species. The low number of variable sites within some markers can explain the differences in results between the two methods (Fig. 5). Highly conserved regions with low percentages of variable sites might lead to sound results when calculating genetic distances, but might make the construction of a phylogenetic tree challenging. This limitation for distance-based methods can sometimes result in biologically meaningless results (Meier et al. 2008; Ross et al. 2008). However, the comparison between the two methods generated several common similarities for the marker trnQ-rps16 and a number of different marker combinations with high discriminatory power, e.g., 1+4, 1+2+4, 1+3+4, 1+2+3+4 (Fig. 5).

In the present study, the discriminatory power of four chloroplast markers proposed by Manzanilla et al. (2018), for Panax spp. was evaluated. The highly valuable P. vietnamensis species and collected samples from eighteen distinct populations across the country were analyzed. The obtained results did not show any significant differences between the populations, which suggests that trnQ-rps16 marker is appropriate for identification at the species level and does not lead to misidentifications of P. vietnamensis regardless of their origins or geographical distributions. Only small differences between experimental and in silico analyses were observed. The in silico study included only a limited number of P. vietnamensis accessions, and this may explain why the empirical data contains more variable sites in P. vietnamensis. Our results support the use of trnQ-rps16 as a single molecular marker for species identification in the genus Panax. An advantage of using a single region for species identification is the reduced time and effort necessary for routine analyses, though it can be combined with other markers (petB, trnE-trnT) when necessary if a single marker is not enough to unambiguously provide a species-level identification. Beside its discriminatory power, trnQ-rps16 can also be easily amplified.

This work was supported by the Ministry of Science and Technology of Vietnam under the project: “Transcriptome sequencing and analysis of Panax vietnamensis Ha et Grushv.” (Grant number 16/2017-HĐ-NVQG). We deeply appreciate Le Thi My Hao and Luong Duc Toan who kindly providing the samples of P. vietnamensis and Nguyen Tien Dung for providing the samples of P. vietnamensis var. fuscidiscus, Panax sp.Puxailaileng, P. stipuleanatus, and P. bipinnatifidus. We would like to thank Nguyen Tap and Nguyen Quoc Binh for the morphological identification of the samples. Marcella Orwick Rydmark, Hugo J. de Boer, and Nguyen Tuong Van are acknowledged for proofreading the text.