Identification of common horsetail (Equisetum arvense L.; Equisetaceae) using Thin Layer Chromatography versus DNA barcoding

The global herbal products market has grown in recent years, making regulation of these products paramount for public healthcare. For instance, the common horsetail (Equisetum arvense L.) is used in numerous herbal products, but it can be adulterated with closely related species, especially E. palustre L. that can produce toxic alkaloids. As morphology-based identification is often difficult or impossible, the identification of processed material can be aided by molecular techniques. In this study, we explore two molecular identification techniques as methods of testing the purity of these products: a Thin Layer Chromatography approach (TLC-test) included in the European Pharmacopoeia and a DNA barcoding approach, used in recent years to identify material in herbal products. We test the potential of these methods for distinguishing and identifying these species using material from herbarium collections and commercial herbal products. We find that both methods can discriminate between the two species and positively identify E. arvense. The TLC-test is more cost- and time-efficient, but DNA barcoding is more powerful in determining the identity of adulterant species. Our study shows that, although DNA barcoding presents certain advantages, other established laboratory methods can perform as well or even better in confirming species’ identity in herbal products.

Tens of thousands of plant species are used medicinally 1 and a substantial portion of the world's population depends on traditional medicine 2 . In recent decades, public interest in herbal products has grown [3][4][5] , but these products are not always regulated. The safety of herbal products can be compromised through accidental adulteration, misidentification and deliberate contamination 6,7 , which can lead to severe side effects due to the presence of toxic compounds 8 . This creates a need for authentication of species included in these products. The qualitative and quantitative composition of herbal products is regulated by international and national monographs such as the European Pharmacopoeia 9 , which presents a series of monographs for herbal products, including recommended tests for identification and quality. These tests are often based on morphology. However, macroscopic or microscopic identification of plant species requires considerable expertise to differentiate between closely related or similar looking species. Furthermore, morphological characters may be indistinguishable in bulk, pulverised or otherwise processed material 10,11 .
To circumvent these problems, most monographs define a maximum allowance of foreign matter often based on a Thin Layer Chromatographic (TLC) test using chemical markers allowing distinction between the correct species and other, potentially toxic species 12,13 . However, such chemical markers or fingerprinting analyses have certain drawbacks. First, it is often difficult to find chemical markers that are unique to the target species. Different species can produce the same marker, hindering species' identification. Second, chemical composition can demonstrate considerable intraspecific variability depending on season, growth, storage conditions and harvesting process 14 . Third, herbal products are sometimes spiked with synthesised compounds 15 . In these cases, the TLC-test may lead to false species' identification.
An alternative method that has been used to identify components of herbal products is DNA barcoding 10,11,[16][17][18][19] . DNA barcoding relies on sequencing of short fragments of the genome, which are unique to the target species 20 . The DNA sequences from the product are compared to a reference database, based on which the identity of the species can be confirmed 21,22 . DNA-based identification methods have often revealed adulteration in traditional medicinal preparations and herbal products. For example, potentially toxic Ephedra L. and Asarum L. material was found in Traditional Chinese Medicinal products administered in Australia 23 , and several adulterant plant species were found in herbal products from North America 17 . Nevertheless, DNA barcoding also has limitations. First, depending on the condition of the plant material, amplification of the target DNA marker may not be possible. Second, DNA barcodes might show low interspecific variability, particularly among closely related species. Finally, because DNA barcoding relies on the presence of a reference database, the absence of a species from the database will impede its identification success 19 . Despite its limitations, DNA barcoding has often been discussed as the primary method of molecular identification of plants in the last decade 11,16,22 .
In this study, we explore molecular identification of the genus Equisetum L. (Equisetaceae), also known as horsetails. The genus comprises 15 species and has a more or less cosmopolitan distribution 24,25 . Equisetum arvense L. is used traditionally against numerous conditions 26 and many E. arvense herbal products are sold on the market mainly against urinary and renal conditions 27 , as well as skin, hair and nail remedies, potentially due to the species' high silica content 28 . The separation of E. arvense from other Equisetum species -especially E. palustre L. that contains toxic levels of the pyridine alkaloid palustrine -is challenging 29,30 , particularly based on microscopic examination of commercial herbal products. Therefore, the European Pharmacopoeia monograph for the common or field horsetail, E. arvense, includes a TLC-test (Identification C) that tests for its positive identification, including a test for foreign matter from E. arvense. However, it is not clear whether this test can positively identify either E. arvense or E. palustre among other morphologically similar Equisetum species, several of which overlap geographically with E. arvense 31 . This is a potential problem because palustrine is not specific to E. palustre, but it is found in other horsetail species. An early study detected palustrine in E. arvense and E. hyemale L. 32 . A later study did not detect it in E. arvense, E. telmateia Ehrh., and E. sylvaticum L. 33 , but a more recent compendium of poisonous plants cites palustrine and palustridine alkaloid content for E. fluviatile L., E. hyemale, E. palustre, E. sylvaticum, and E. telmateia 34 .
The main objective of this study was to investigate the resolution power of the European Pharmacopoeia's TLC-test and of the DNA barcoding approach for i) distinguishing between E. arvense and E. palustre and ii) positively identifying these two species and discriminating them from other Equisetum species. In order to perform these investigations, we needed to have a reliable species' delimitation. Therefore, we also reconstructed a molecular phylogeny of Equisetum to test currently accepted species boundaries. Our study is based on herbarium collections of wild origin, as well as exemplar herbal products from the market.

Results
Phylogeny of Equisetum. We reconstructed a phylogenetic tree of Equisetum ( Fig. 1) in order to test the monophyly of the species. Previous studies have provided phylogenetic hypotheses for the genus using plastid DNA markers [35][36][37] , but these studies only included one specimen per species. The topology obtained here from nuclear and plastid markers, and including several accessions per species, largely corresponds to the topology found previously [35][36][37] . Equisetum bogotense Kunth is recovered as sister to the rest of the genus and not as a member of subg. Hippochaete (Milde) Baker. The remainder of the genus is resolved into two major clades, each comprising seven species and corresponding to the two subgenera Equisetum and Hippochaete (Fig. 1). With the exception of E. diffusum D. Don and E. sylvaticum, all species were recovered as monophyletic, including E. arvense and E. palustre. These two species are resolved in the same clade (subg. Equisetum), but not as sister species (Fig. 1).

Distinction between E. arvense and E. palustre.
Chemistry. The distinction between the two species based on the TLC-test of the European Pharmacopoeia is based on the presence of a combination of marker bands in each species, shown in Fig. 2. The results of the TLC-test recommended by the European Pharmacopoeia are shown in Fig. 3 for the E. arvense -E. palustre comparison. The two bands at the bottom of the plate that are used for the identification of E. palustre are present in all accessions of this species, but not in any of the E. arvense accessions. Although some of the marker bands used to identify E. arvense can be found in E. palustre accessions, the combination of the four marker bands (Fig. 2) is not seen in any E. palustre accessions Scientific RepoRts | 5:11942 | DOi: 10.1038/srep11942 (Fig. 3). Therefore, the marker zones used as the distinguishing characters between the two species in the monograph (Fig. 2) could consistently distinguish between E. arvense and E. palustre. We observed a typical E. arvense TLC chromatogram for five out of eight commercial products included in the analysis (Fig. 4). In one product (B -Bulgaria), we observed the marker bands that are used to identify both E. arvense and E. palustre in the TLC-test of the European Pharmacopoeia, suggesting this product includes a mixture of the two species (Fig. 4). One product (I -UK) seemed to not contain any Equisetum material at all and another one (HB -UK) returned no chromatogram (Fig. 4).
DNA barcoding. The two plastid markers we used for the DNA barcoding of E. arvense and E. palustre resolve the samples into two well supported, monophyletic clades, shown in Fig. 5. We were able to amplify DNA from two of the herbal products, only; one was resolved within the E. arvense clade (BP 100). The other, which was shown to be a mixture from the TLC-test (B -Bulgaria), is recovered with the E. palustre (BP 77) clade (Fig. 5). For both barcoding regions, we found 36 substitutions (25 for matK and 11 for trnH-psbA) that can distinguish E. arvense and E. palustre. Some of them are unique to each species and others are shared with other species, but not between E. arvense and E. palustre (Table 1).

Positive identification of E. arvense and E. palustre.
Chemistry. We analysed one exemplar specimen of all Equisetum species using the TLC-test recommended by the European Pharmacopoeia (Fig. 6). For E. diffusum and E. sylvaticum, which were not monophyletic in the DNA analysis, we could test only one sample, as the other sample did not come from our study. The TLC-test (Identification C) of the European Pharmacopoeia can positively identify E. arvense. Although some of the marker bands outlined in the TLC identification test for E. arvense ( Fig. 2) are seen in the chromatograms of other Equisetum species, E. arvense is the only species with the combination of all these markers bands (Fig. 6). As shown in Fig. 6, one or two of the greenish-blue fluorescent zones used in the TLC-test to detect E. palustre were not detected in any other species within subg. Equisetum, but were present in all species in subg. Hippochaete. Therefore, the TLC-test of the European Pharmacopoeia cannot be used to identify E. palustre, because the trait of this species (Fig. 2) is shared with other Equisetum species as well (Fig. 6).
DNA barcoding. We investigated whether the two DNA markers we used as barcodes can not only differentiate E. arvense from E. palustre, but also include a combination of unique traits for these species, which can be used to successfully identify them from all other Equisetum species. Table 1 shows that there are unique substitutions in these two markers, the combination of which can positively identify both E. arvense and E. palustre from other Equisetum species. For matK, we found no substitutions to be unique for E. arvense, but five substitutions were unique for E. palustre (Table 1). For trnH-psbA, one substitution was unique for E. arvense and two for E. palustre (Table 1). Regarding the identification of material in herbal products, DNA sequences from one product (F -Germany) show the combination of characters that can identify E. arvense. For the other product (B -Bulgaria), we only managed to amplify matK. This sequence is actually a chimeric sequence (several doubles peaks are observed in the DNA chromatograms), showing some characters that are characteristic of E. arvense and some of E. palustre.

Discussion
Some Equisetum species are morphologically quite variable and can be difficult to identify based on morphology alone 29,30 . To the untrained eye, E. arvense may superficially resemble other species within subgenus Equisetum, including E. palustre, as well as E. pratense, E. fluviatile, E. telmateia and E. diffusum. Positive identification of material lacking strobili, or where information about dimorphism is lacking, may be challenging even for trained botanists, as micro-morphological or anatomical characteristics may be required to separate some species, e.g. E. arvense and E. palustre 38 . Within their respective ranges, taxa sharing similar morphological characters, such as E. arvense and E. palustre, may be found co-occurring in the same habitat 31 . A further complication to field-identification is that E. arvense is known to form hybrids with E. palustre (E. × rothmaleri C.N. Page) and E. fluviatile (E × littorale Rupr.) 39,40 , with morphological and chemical traits that are intermediate between the parent taxa 30,40 .
Due to the risk of misidentification or adulteration of E. arvense with E. palustre, laboratory techniques are needed for the quality control of herbal products of E. arvense. The European Pharmacopoeia has devised a simple method using TLC (Identification C) to distinguish the two species 9 , and we found this test to be straightforward and consistent. It can confirm that the material is from E. arvense, through a combination of marker bands unique to this species (Figs 2,3 and 6). Further, as shown in Fig. 3, the two greenish-blue bands at the bottom are present in all E. palustre accessions, but none of the E. arvense accessions. The presence of these bands can be used as an indication of adulteration with E. palustre, but the identity of the adulterant is not confident, because these bands are also found in other Equisetum species besides E. palustre (Fig. 6). Also, even in the case of absence of these bands, a partial adulteration with another Equisetum species that does not demonstrate them in the chromatogram (Fig. 6) cannot be ruled out.
The current TLC-test is testing for the presence of kaempferol glucosides (flavonoids), instead of directly testing for the presence of alkaloids. We tested for alkaloids using the material which had already been extracted for the flavonoid analysis. This method could only detect alkaloid bands (two bands) present in the reference E. palustre HRS and the E. palustre accession used on the TLC-test across the genus, whereas possible alkaloids present in other Equisetum species were below the detection limit of this method (results not shown). We suggest that a method testing directly for alkaloids be developed and included in the monograph.
DNA barcoding may be an alternative or supplementary method to identify material in herbal products with higher certainty [17][18][19]41 . We found that two plastid markers can successfully distinguish between E. arvense and E. palustre. In total, there are 36 characters (25 for matK and 11 for trnH-psbA) differentiating the two species (Table 1), and the phylogenetic analysis of these two DNA barcoding markers assigns material from these two species to two well-supported clades (Fig. 5). Further, this approach can positively identify the two species, as we found six substitutions (five for matK and one for trnH-psbA) that are unique to E. palustre (Table 1), allowing high confidence in the identification of this species. For E. arvense, we only found one unique substitution in trnH-psbA and none in matK (Table 1), making assignment of material to this species less robust. However, a number of other substitutions are only shared by E. arvense and its two closest relatives, E. fluvatile and E. diffusum (Fig. 1), which co-occur in Asia. Including more DNA barcoding regions that have been proposed by the Consortium for the Barcode of Life Plant Working Group 22 could provide further discriminatory power for E. arvense. However, we found rbcL to show too little interspecific variation, and ITS2, which has been proposed for the DNA barcoding of medicinal plants 16 , did not amplify consistently in Equisetum. Other DNA barcoding markers that have been shown recently to perform better than the ones we used here [e.g., ycf1 42 ] could provide more species-specific substitutions in future investigations.
We included eight commercial products claiming to be E. arvense, seven of which produced TLC chromatograms allowing assignment of the herbal product to either E. arvense or E. palustre following the European Pharmacopoeia's TLC-test for foreign matter. Of these seven products, five were assigned to E. arvense (Fig. 4). We were only able to gather DNA sequence data for one of these samples (herbal product F -Germany), and it was confirmed to be E. arvense (Fig. 5). For one product (herbal product B -Bulgaria), the TLC-test showed the presence of E. arvense and potentially E. palustre material (Fig. 4). The DNA sequence data confirmed that this product is most likely a mixture, as the resulting sequence was chimeric. Although this product is resolved within the E. palustre clade (Fig. 5), the sequence we amplified shows a combination of substitutions characteristic of E. arvense and E. palustre. It could be adulterated, misidentified or even be of hybrid origin. The product is a tea from south-eastern Europe, an area that is a major source of commercial E. arvense products 27 , and where the two species co-occur, raising concerns about the risk of contamination with E. palustre in commercially available material presumed to be E. arvense. Finally, one sample (herbal product I -UK) produced a chromatogram that was different from those characteristic of any Equisetum species (Figs 4 and 6), suggesting the botanical material in that sample might not be Equisetum. Unfortunately, no DNA sequence data could be gathered from that sample.
Our objective was to explore and compare the power of the European Pharmacopoeia's TLC-test and of the DNA barcoding approach for distinguishing between E. arvense and E. palustre, as well as for positively identifying the two species. We found both methods to be useful, however with different advantages and shortcomings. In terms of success rate of data collection, the TLC-test approach is more efficient. First and foremost, the laboratory work is less laborious and cheaper than DNA barcoding. Second, the TLC-test had a greater success rate with commercial herbal products: we obtained chromatograms for seven out of eight of these products, while the amplification success of the barcoding regions from these products was limited (only two samples). On the other hand, in terms of resolution and confidence in identification, the DNA barcoding approach is better. Although both methods can successfully discriminate between E. arvense and E. palustre and positively identify E. arvense, only DNA barcoding provides a combination of traits that is unique to E. palustre among horsetail species. However, the amplification of these barcoding markers might prove difficult in processed commercial products. Additionally, an advantage of the TLC-test is that contamination can be quantified based on the level of visibility of the greenish-blue bands on chromatograms, an aspect in which the DNA barcoding approach lacks.
Which method would we recommend as being the best? Given the pros and cons of each method, we believe that it depends on the application. Our results show that, when it comes to confirming whether an herbal product contains E. arvense, the TLC-test is the most cost-and time-efficient option. However, the presence of the marker bands described in the TLC-test as characteristic of E. palustre can be seen in cases of adulteration with other Equisetum species, as these bands are common within the genus (Fig. 6). Similarly, the absence of these marker bands does not guarantee that the product has not been adulterated with other Equisetum species, which do not show those bands (Fig. 6). Given that there is uncertainty about which Equisetum species produce toxic alkaloids, this could be an important shortcoming of the TLC-test. In these cases, DNA barcoding can be used as a complementary test for quality control, when possible.
Our study also highlights the immense potential of herbarium collections for a wide range of modern approaches to biodiversity research 43,44 , and DNA barcoding in particular 45,46 . The majority of the material used in this study was obtained from the collections in herbarium of the Natural History Museum of Denmark (C). The age of the material ranged from 1900-2013. We did not detect any apparent age-related difference in the intensity of the TLC chromatograms or the amplification success of the DNA markers, showing that the chemical profiles and the DNA are not substantially degraded in carefully stored collections 47 . Our findings demonstrate how available collections can be used to set up a modern framework of chemical and molecular identification of economically important species. Without conducting substantial fieldwork, we managed to sample across all Equisetum species, as well as within E. arvense and E. palustre, covering their geographic ranges, hence ensuring that both inter-and intra-specific variation is covered. An incidental advantage of using herbarium material is that the link between the chemical and molecular data and the voucher is established by default. Missing vouchers is a serious problem in many studies 41 which makes replication by future researchers almost impossible 48 .

Conclusions
Given the recent growth of the herbal products market 3-5 , efficient methods for regulating these products against accidental adulteration, deliberate contamination and misidentification are more relevant than ever for public healthcare 41 . We tested the European Pharmacopoeia's E. arvense TLC-test for foreign matter, particularly from the closely related E. palustre. We also tested a DNA barcoding approach to distinguish and identify these species. We found that each method has advantages and disadvantages, but the TLC-test is the most efficient way of confirming that material in herbal products is indeed E. arvense.    On the other hand, the DNA barcoding can be used as a complementary test to determine the identity of adulterant species, particularly E. palustre. Future work can focus on systematically studying which Equisetum species produce toxic alkaloids, which will assist the quality control of E. arvense herbal products. Further, a chemical method that directly tests for the presence of alkaloids in herbal products can circumvent problems in species identification, directly testing for the quality and appropriateness for human consumption of herbal products. Additionally, the steadily dropping price of next generation sequencing techniques -which massively amplify short DNA fragments -may considerably enhance the success rates of DNA barcoding in degraded or processed material. Finally, given the presence of several putative hybrids between E. arvense and other Equisetum species, further techniques can be applied to investigate the presence of hybrid material in herbal products.

Methods
Plant material. For the phylogenetic reconstruction, we sampled at least one accession of each Equisetum species, mostly from material deposited in the herbarium of the Natural History Museum of Denmark (C), in order to produce a well-sampled phylogenetic hypothesis for the genus. From these specimens, we chose one per species for the TLC-test across Equisetum species. For the DNA barcoding and TLC-test of E. arvense and E. palustre, we sampled several accessions of each of the two species covering their distribution ranges to the extent possible. Additionally, we sampled eight herbal products sold on the market as E. arvense. Details of plant materials are listed in Supplementary Tables 1 and 2. DNA sequencing. Complete genomic DNA was extracted using the DNeasy Mini Plant Kit (Qiagen Ltd, Crawley, UK), following the manufacturers protocol. For the DNA barcoding of E. arvense and E. palustre, we sequenced the trnH-psbA spacer and the barcoding fragment of matK, which have been used in previous DNA barcoding studies 21,22,49,50 . For the genus wide analysis, we sequenced the plastid regions rps4, rbcL, the barcoding fragment of matK, the trnH-psbA spacer, and the nuclear ribosomal ITS2 region. The rps4 marker was amplified using primers rps5 ( Reactions of 25 uL were carried out using standard procedures with 1 or 2 uL DNA template. Moreover, for matK and ITS DMSO was added to reduce the effects of secondary structure on primer biding. BSA was added to all reactions to enhance polymerase activity. The PCR products were purified using the Qiagen PCR purification kit (Qiagen Inc.) according to the manufacturer's instructions. Direct sequencing of purified PCR products was either performed using BIGDYE v1.1 (Applied Biosystems, Wellesley, Massachusetts, U.S.A.) and purified sequencing products were run on an AB3130 × 1 automated sequencer (Applied Biosystems) or sent to GATC-biotech in Germany (http://www.gatc-biotech.com). Forward and reverse sequences were edited and assembled in Geneious v. 7.1.7 (http://www.biomatters. com). Alignments were conducted using the MAFFT v.7 plugin 57 in Geneious with default options and inspected manually afterwards. Regions that were ambiguously aligned were excluded from the analyses. Genbank accession numbers for all sequences used in the study are shown in Supplementary Table 3 Phylogenetic methods. Two matrices were assembled: (1) a genus wide matrix combining our datasets with DNA sequences from previous phylogenetic studies of Equisetum 35-37 to achieve a sampling scheme of multiple accessions per taxon allowing test of species monophyly, and (2) an Equisetum arvense-E. palustre dataset for the development of the DNA barcoding methodology. All sequences were aligned with MAFFT 57 and sequence data were analysed under the Maximum Likelihood (ML) criterion, with RAxML 58 using the partitioned model option (five partitions -one per DNA marker) with the GTR+ I+ G model and running 100 bootstrap replicates 59 . Angiopteris angustifolia and Ophioglossum reticulatum were used as outgroup for the phylogenetic analysis of the genus, and E. variegatum was used as outgroup for the phylogenetic analysis of DNA barcodes.  9 . Due to the limited availability of material from the herbarium specimens in general, only about 20-50 mg of powdered stem was extracted and the amount of methanol used adjusted accordingly. For the test solutions, powdered Equisetum stems were extracted with methanol R (VWR BDH Prolabo Chemicals) in the ratio 100 mg/mL. The mixture was heated in a water-bath at 60 °C for 10 min with occasional shaking, allowed to cool and then filtered. The reference solution (a) of Equisetum palustre HRS (European Directorate for the Quality of Medicines) was prepared in the same way as the test solutions. Another reference solution (b) was made by dissolving 1.0 mg of caffeic acid R (Sigma), 2.5 mg of hyperoside R (Roth) and 2.5 mg of rutin R (Sigma) in 20 mL of methanol R. For commercial products, 1 g material was extracted with 10 mL methanol R in the same way as the test extracts.
2 μ l bands of 8 mm of each solution were applied with a GAMAC nanomat 4 to HPTLC silica gel plates R (5-6 μ m; Merck). HPTLC plates were developed over a path of 6 cm using a mobile phase consisting of anhydrous formic acid R (Emsure), glacial acetic acid R (Merck), water R, and ethyl acetate R (Sigma Aldrich) (7.5:7.5:18:67 V/V/V/V). After development, plates were air-dried for 5 min. Detection was achieved by heating at 100 °C for 3 min followed by treatment of the still warm plate with a 10 g/L solution of diphenylboric acid aminoethyl ester R (Roth) in methanol R, and then treatment with a 50 g/L solution of macrogol 400 R in methanol R. Finally plates were air-dried and examined after 10 min in ultraviolet light at 365 nm. System suitability was observed by the appearance of two greenish-blue fluorescent zones from kaempferol glucosides (flavonoids) characteristic of E. palustre L. in the reference solution (a) just above the line of application. In the chromatogram of the test solution any greenish fluorescent zones just above the line of application may not be more intense than the corresponding zones (characteristic for E. palustre) in chromatogram of the reference solution.
For alkaloid detection, the dried, powdered stem material, which had already been extracted for flavonoid-analysis, was moistened with 10% (1 μ L/μ g dry plant material). 1 ml dichloromethane (VWR BDH Prolabo) was added, and the mixture was extracted for 24 h at room temperature. 900 μ L of the liquid was taken to dryness. The extract was redissolved in 20 μ L dichloromethane and applied to a Merck Silica gel 60 F254 TLC plate and eluted in toluene:ethyl acetate:diethylamine (VWR BDH Prolabo; Sigma; Merck) 7:2:1 over 7 cm. 1 mg/mL brucin was used as positive control. The plate was sprayed with 0.15% chloroplatinic acid hydrate (Sigma Aldrich) in a 3% KI solution.