To develop an effective SSR-HRM assay, in this study we surveyed, documented, collected, and selected 161 specimens of original licorice plants as well as common adulterants (Supplementary Table S1). Leaf material was collected from these plants and was desiccated in silica gel before DNA extraction. To verify the effectiveness of the SSR-HRM method in detecting adulteration of commercially available licorice products, 47 products labeled as G. uralensis and/or “gan cao” in Chinese were purchased from different raw herbal material retail markets or online stores. Total DNA was extracted according to a modified CTAB protocol as reported by Sun et al. (2016). High-quality genomic DNA is an important prerequisite for accurate HRM-based plant identification. Generally, the concentrations of polysaccharide and polyphenol compounds in root and rhizome-based herbs are relatively high. These interferents must be removed using polyvinylpyrrolidone during the early stages of DNA extraction. In addition, licorice roots are also rich in fiber, and so the volume of material and of the corresponding extraction solution should be increased. DNA concentration was quantified using a spectrophotometer (Qubit 3.0, Invitrogen Co., United States) and was then adjusted to a 100 ng/μL working concentration and stored at -20°C until needed.

Genome data for G. uralensis was retrieved from the ngs-data-archive². To identify potential SSR markers, the G. uralensis genome sequences were screened using MISA³. We searched for loci with di- and trinucleotide motifs of at least eight or seven repeats, respectively, or with five repeat units for tetra-, penta-, and hexanucleotide SSRs. After screening, more than 100,000 putative microsatellite loci were identified. Because so many microsatellite loci were found in the genome, we selected at random 350 putative microsatellite loci distributed in different scaffolds and in different locations to design SSR primers.

Microsatellite loci were initially amplified using five DNA samples from the five licorice species. PCR was carried out in a final volume of 20 μL containing: 10 μL of 2× Taq PCR mix (Sangon Biotech Co., Ltd., China), 1 μL of DNA template (100 ng/μL), 1 μL of each primer (10 μM), and 7 μL of ultrapure water. The following PCR program was used: 94°C for 3 min; 35 cycles at 94°C for 30 s, at primer annealing temperature for 30 s, and at 72°C for 30 s; and a final extension cycle at 72°C for 7 min. PCR products were separated using 8% non-denaturing polyacrylamide gels. SSR markers showing polymorphism among the five species examined were chosen to amplify in 161 samples. Only those SSR markers that were monomorphic within each species were used for further SSR-HRM analysis (Table 1).

SSR-HRM analysis was performed using a Rotor-Gene Q MDx (QIAGEN) real-time PCR machine. The HRM reaction mixture contained 50 ng of genomic DNA, 5 μL of 2× HRM master mix (QIAGEN, Germany), 0.5 μL of 10 μmol/L of both forward and reverse primers, and distilled water was added to reach the final volume. HRM-PCR was conducted in a 72-well carousel with the following reaction conditions: an initial denaturing step of 94°C for 3 min, followed by 40 cycles of 94°C for 30 s, primer annealing temperature for 30 s, and 72°C for 30 s. Fluorescence data was collected at the end of each extension step of each PCR cycle. HRM was performed immediately after PCR amplification. For HRM, the temperature rose from 70 to 90°C at 0.15°C increments with a 2 s hold time for each acquisition step. As the temperature increased, fluorescence was continuously monitored. A sharp drop in fluorescence occurred as the sample temperature reached the denaturation temperature. Data was analyzed using the Rotor-Gene Q software. Melting curve profiles were processed in three steps: (1) normalization, (2) temperature shift, and (3) difference plot visualization. Melt data were first normalized by removing background fluorescence and adjusting the fluorescence signals for the variation among wells. The negative derivative of the fluorescence (F) over temperature (T) (dF/dT) curve identifies the melt temperature (T_m), and the normalized raw curve shows the decrease in fluorescence against increasing temperature. Finally, to create more easily distinguishable plots, a base melting curve was selected. To genotype commercial licorice products, a two-step procedure was used to assess the similarity between unknown HRM curves (i.e., commercial samples) and known ones (i.e., original plants). First, the normalized HRM curve of the commercial product was compared to known normalized HRM profiles of the five Glycyrrhiza species. The HRM profile of the unknown genotype either resembled one of the known samples or generated a unique melting curve. Second, the HRM genotype confidence percentage (GCP) of the licorice product was calculated by setting each known species as a “genotype” (i.e., reference species), and the confidence threshold was set to 90%.

To validate the results of HRM analysis, we labeled all forward SSR primers with the fluorescent dye 6-FAM and performed PCR before conducting capillary electrophoresis. The PCR was performed in a 10 μL reaction mixture consisting of 0.4 μL of genomic DNA, 1 μL of 10× buffer, 0.3 μL of 25 mmol/L MgCl₂, 0.2 μL of 10 mmol/L dNTPs, 0.1 μL each of 10 μm/L forward and reverse primers, 1 μL of 10 mg/mL BSA-V, 0.1 μL of 5 U/L Taq polymerase, and distilled water up to the final volume. PCR amplification reaction parameters were as previously mentioned. Capillary electrophoresis test samples were prepared by diluting amplified SSR fragments and dispensing in each well 2 μL of the diluted products and 9 μL Hi-Dye formamide solution premixed with 500 Liz^TM dye standard following the manufacturer’s instructions (Life Technologies). Samples were then denatured at 95°C for 3 min followed by quick cooling on ice. Fragment analyzes were conducted on an ABI 3730XL genetic analyzer (Applied Biosystems, Foster City, CA, United States). During capillary electrophoresis, the separation processes were automatically recorded in a single GeneScan file; this file was then analyzed with GeneMapper V4.0 (Applied Biosystems, Foster City, CA, United States).

The genome sequence of G. uralensis was extracted from a genome data bank for SSR marker mining. In total, more than 10,000 microsatellite loci were identified, of which 350 microsatellite loci with dimer, trimer, tetramer, pentamer, and hexamer motifs with lengths longer than 10 bp were randomly selected for further analysis. Mononucleotide repeats were excluded from this study due to the fact that they are associated with unstable polymorphism. The 350 SSR primer pairs were used to amplify DNA-fragments from five Glycyrrhiza species, and the resulting PCR products were analyzed using polyacrylamide gel electrophoresis to define their repeat numbers and to detect polymorphism. The electrophoresis data showed that the discriminatory power of the SSR primers ranged from 0 to 100%, and the SSR markers could be separated into five classes as follows: (1) none of the five Glycyrrhiza species could be distinguished from each other, (2) two of five species could be distinguished from one another, (3) three of five species could be distinguished, (4) four of five species could be distinguished, and (5) all five species could be distinguished from one another. After one round of screening, 83 primer pairs that had the ability to distinguish five Glycyrrhiza species were qualified for the next step of testing. In this round, the 83 candidates were characterized using 161 samples of the five species. Finally, nine primer pairs which displayed monomorphism within samples of same species were used for SSR-HRM analysis in this study (Table 1).

The potential resolving power of HRM is much greater than conventional melting curve analysis and PAGE gel electrophoresis, since HRM permits distinguishing between the shapes of melting curves from different amplicons even when they have the same T_m. Here we used both T_m values and the shapes of melting curves to assess differences among five Glycyrrhiza species (Figure 2). Figure 2a presents conventional derivative melting curves, which shows peaks indicating the T_m values for the GM-S9 SSR fragment amplified from the five Glycyrrhiza species. The melting curve is produced by slowly melting the DNA of tested samples in the presence of a dsDNA binding dye within a range of temperatures. The melting temperature peaks of the tested plant species are calculated as T_m values (Table 2). The T_m values of G. inflata, and G. glabra were similar, but these species could be distinguished by plotting normalized melting curves (Figure 2b) and difference plot melting curves (Figure 2c). Analysis of the normalized melting curves of the SSR marker GM-S9 (Figure 2b) proved that all test species could be easily distinguished. G. uralensis and G. pallidiflora were easily distinguished by their distinct melting curves, while in contrast the curve profiles of G. inflata and G. eurycarpa were similar and were not easy to visually distinguish from each other. To better visualize small differences between these two species, close examination of difference plot melting curves was performed with the melting curve of G. eurycarpa used as the baseline; this revealed that a part of the curve was outside the 90% CI curve, suggesting that the melt curves of G. inflata and G. eurycarpa are indeed different (Figure 2c). Similar results were obtained for all other eight SSR markers except GM-S4; this marker could not be amplified in G. inflata (Supplementary Figure S1). Furthermore, the confidence values for the similarity between each pair of the Glycyrrhiza species was calculated by setting each species as a “genotype” (reference species) with a confidence threshold that was set in an appropriate control. As shown in Figure 2d, the highest GCP (30.36) was found between G. inflata and G. eurycarpa, while the lowest (0) was between G. uralensis and other species. According to calculated GCPs, the most informative microsatellite marker was GM-S1, while GM-S7 was the least informative. However, while GM-S1 was the best marker for this study, other SSR primers—either alone or in combination—might be better able to identify other Glycyrrhiza species.

To validate the HRM results, we performed capillary electrophoresis on heat denatured PCR products labeled with FAM. These results revealed that each Glycyrrhiza species had distinct patterns and that the five species could be differentiated from each other. This demonstrates that the newly designed SSR primers can be used for licorice species identification (Figure 3 and Supplementary Figure S2) and is consistent with our HRM results. Interestingly, analysis of sequenced DNA found that some amplicons not only exhibited different microsatellite alleles, but also contained single nucleotide polymorphisms (SNPs) in regions flanking the SSR (data not shown here due to the poor quality of SSR sequencing). The presence of SNPs in these flanking sequences has been previously reported in other species (Distefano et al., 2012). HRM is a powerful method that can be used to detect SNPs, which means that relative to capillary electrophoresis, its higher sensitivity means that it is better able to discriminate PCR products of the same size that have different base compositions.

After the confirmation that each species could be identified using the SSR-HRM protocol, we used the same method to identify the species used in commercial licorice products. A total of 47 test samples were collected from products sourced from both pharmacies and online vendors. All samples were labeled as “gan cao” and were available to consumers and patients. Genomic DNA was extracted from all 47 samples using a modified CTAB method (Sun et al., 2016). DNA samples from the five known Glycyrrhiza species were used as controls. The normalized and difference plot curves of the amplicons from the five Glycyrrhiza species and 47 commercial licorice products, based on HRM analysis of SSR marker GM-S1 are shown in Figure 4. By using each known species as a “genotype” (i.e., reference species), and with the confidence threshold set to 90%, we were able to calculate the degree of confidence in the similarity of each commercial product to each of the five known species (shown in Supplementary Table S2). For example, HRM analysis allowed us to identify G. uralensis in sample 13 with a confidence level of 99.6%, which agrees with the labeled information indicating that this product is 100% G. uralensis (Supplementary Table S2). However, in sample 4, HRM analysis detected G. pallidiflora with a confidence level of 98.9%, which suggested that this product contained an adulterant. The same results were also observed for nine other samples in which G. pallidiflora and G. eurycarpa were identified with a high level of confidence. It was also noteworthy that the melting curve of tested sample 40 was intermediate between the curves of G. uralensis and G. eurycarpa; this may indicate that this product is an admixture of G. uralensis and G. eurycarpa (Figure 4). To back up this speculation, TA-cloning was conducted on PCR product from this sample using the universal ITS2 primer. Sequencing of the cloned products confirmed that DNA from two different species were present. We also found that 15 samples (31.9%) were authenticated as G. glabra, 11 (23.4%) were authenticated as G. uralensis, and 10 (21.3%) were authenticated as G. inflata. This confirms that the most common species found in available commercial licorice products are species included in the Chinese Pharmacopoeia. This finding also provides evidence that some adulterants are present in herbal products sold in Chinese markets, and that this adulteration may be a serious issue for the herbal drug industry.

The HRM method has been widely used for many purposes (Druml and Cichna-Markl, 2014), including the detection of infectious diseases, testing of bacterial pathogens in foods, and the identification of genetically modified organisms. HRM has notable advantages due to its simplicity, rapidity, sensitivity, and accuracy. Based on the protocol used in this study, we propose a brief operating procedure for traditional herbal medicine authentication using SSR-HRM (Figure 5). First, genomic DNA must be extracted from both the authentic medicinal plant and the adulterant(s). The extracted DNA can then be used to acquire a microsatellite library using next-generation sequencing. Second, the fragments of the selected microsatellite loci should be amplified by PCR and the amplicons run on a PAGE gel to assess polymorphism. Third, SSR-HRM amplification is performed using primers for polymorphic loci with DNA from both authentic, known plants as well as their adulterants. The best primer pair—i.e., those that are both stable and efficient—are then rechecked. Fourth, a preliminary test is conducted to detect adulterants in commercial products by performing SSR-HRM PCR using the selected primers. Fifth, if the SSR primers do not amplify in the adulterant samples, DNA barcoding can be used to assist with species identification.

The supply and safety of herbal medicines are major global public health issues, and are particularly important in heavily populated countries such as China and India. Rapid modernization and globalization continue to have a profound impact on both herbal medicine supply and safety. Many traditional herbal medicine products are available not only in pharmacies but are now also sold on the Internet. According to a WHO survey, in 99 out of 142 countries a majority of herbal products can be purchased without prescription, and many natural remedies and medicines are used for self-medication, and/or are bought or prepared by friends, acquaintances, or the patient (World Health Organization [WHO], 2004). These trends bring highlight the importance of the quality and safety of the products used. Moreover, with increased use of traditional herbal medicines, reports of adverse reactions have also increased.

In China, a country where traditional therapies are widely used to complement conventional medicine, there were 9854 known reported cases of adverse drug reactions in 2002 alone, which was higher than the 4000 that occurred between 1990 and 1999 (Li et al., 2016b). Numerous and irrefutable cases of adverse herbal drug reactions have been reported in the literature (Zhou et al., 2004; Chen W.L. et al., 2010; Liu et al., 2011; Dunnick and Nyska, 2013; Kim et al., 2013). Misidentification of medicinal plants and mislabeling herbal medicinal products can sometimes cause adverse consequences, so it is of great importance to examine herbal medicines for the presence of adulterants. The Former Director General of the WHO Dr. Lee Jong-wook said that “WHO supports traditional and alternative medicines when these have demonstrated benefits for the patient and minimal risks.” Thus, it is easy to conclude that WHO highlighted the contribution of traditional medicine in improving health care. “But as more people use these medicines, governments should have the tools to ensure all stakeholders have the best information about their benefits and their risks,” Lee added. Whether intentional or not, substitution and adulterants are highly undesirable.

To address this issue, DNA barcoding technology provides a novel method that can be used to confirm the identification of raw herbal materials and to develop a standard quality within the marketplace. This technology uses a short fragment (<1000 bp) of the genome (also called a “DNA barcode”) to identify species (Lahaye et al., 2008). In plants, however, this method cannot be used to distinguish between in closely related species or sub-species, due to the slow evolution and limited divergence in these barcodes. Further studies are therefore necessary for this protocol improvement, especially exploiting a complementary tool to facilitate DNA barcoding for a broader coverage of plant species. Here we argue that an SSR-based method may serve as a useful alternative. SSR markers result from differences in the lengths of mono- to hexa-nucleotide repeat sequences. These markers are popular because of their high stability and co-dominant transmission. Moreover, conserved microsatellites can be readily used for taxonomic studies (Küpper et al., 2008; Tuler et al., 2015). Because of these advantages, the SSR method can be effective and for the identification of closely related species, and can therefore assist in identifying ingredients present in herbal medicine.

This study is the first to use SSR-HRM to detect the species composition in currently available licorice herbal supplements. Our results indicated that genomic DNA can be obtained from medicinal parts (i.e., dried roots) as well as from their processed or modified forms (i.e., decoction pieces or powder), and its quality was suitable for HRM analysis. We used 161 voucher samples to verify the reliability of the newly developed SSR markers. SSR patterns obtained from polyacrylamide gel and HRM analyzes were similar, which indicated that these SSR markers were able to accurately distinguish between the five species. Therefore, we speculate that the SSR markers developed in this study have the power to identify the original plant species used in commercial Glycyrrhiza herbal supplements as well as their substitutes.

Among the 47 herbal samples sold in pharmacies and Internet in China as “gan cao,” we confirmed that 75% of the samples were authentic, while the rest of the samples included adulterants. Seven test samples were authenticated as G. pallidiflora, and three of the remaining samples matched G. eurycarpa. One sample was authenticated as an admixture of G. uralensis and G. eurycarpa. This result demonstrates that licorice products available in the market may have complex sources, and that samples purchased on the Internet may contain adulterants. Using such substitute species as dietary supplements may be detrimental to patients and consumers and may ultimately also be harmful to public health. Although 36 out of 47 (75%) products tested were identified as the labeled species, we speculate that a similar analysis of a larger number of samples from additional markets could lead to figures lower than 75%. Detecting adulteration in herbal medicines has been reported in previous studies, and therefore it is not surprising that licorice produce are also mislabeled. The China Food and Drug Administration should monitor commercial licorice dietary supplements. The SSR-HRM method developed here is an effective way to accurately distinguish between closely related species in herbal products.

It has been shown that HRM is a rapid and cost-effective approach for detecting sequence variants in clinical research and diagnostics (Akiyoshi et al., 2013). HRM has also been applied in plants to detect SNPs used for genotype identification (Wu et al., 2008). Potentially, HRM can also be used for the identification of homozygous and (known) heterozygous variants (Montgomery et al., 2010). The time cost and analysis is similar to conventional PCR but HRM does not require post-PCR procedures (e.g., gel electrophoresis). Although some studies have described HRM applications where microsatellite loci are used to identify food adulterants (Ganopoulos et al., 2011; Bosmali et al., 2012), the merits of employing this method have not been fully reported. In this study, we analyzed nine novel SSR markers using HRM and compared the results with those produced from capillary electrophoresis analysis. DNA patterns from five Glycyrrhiza species were distinguished successfully using both applications. Although the power of both methods was the same, in general HRM is a more accurate approach and can also detect SNPs in the sequences flanking SSR motifs. Therefore, the use of HRM may be helpful to overcome the limits of capillary electrophoresis-based SSR analyzes. This is especially true in cases where SSR are monomorphic but SNPs are present elsewhere in the amplicon. Nevertheless, SSR-HRM is still not widely used to identify closely related species in the herbal medicine supply chain. In the future, the use of this application for examining the contents of herbal products will likely increase.

The most important properties for routine techniques used to examine the contents of herbal medicines are: the robustness of the process, the availability of testing devices, and the existence of extensive and accessible primer resources. Despite it proven effectiveness, SSR-HRM is still in the initial stage of being adopted as a routine method used to identify the contents of herbal products, and its use for herbal product traceability remains limited due to the lack of SSR primers. In fact, many techniques that are now well-established suffered from a similar limitation early on but have since developed into robust, universal tools (e.g., DNA barcoding). Therefore, there is a need to develop and establish a primer database for SSR-HRM as a universal tool for herbal medicine identification. In last decade, the emergence of next-generation sequencing technologies has increased the availability of molecular markers, including genomic SSRs and SSRs from expressed sequence tags (Wu et al., 2017). Hence, these resources can be used to establish reference SSR databases for herbal medicine identification, and continual updating can guarantee the integrity of the SSR primer database and improve its application. In conclusion, the use of SSR-HRM for the identification of closely related species in herbal products appears to be promising but remains to be fully exploited. In the future, SSR-HRM will not replace classical DNA barcoding but may provide a supplementary tool to facilitate DNA barcoding to identify closely related species in herbal medicine products.