Authentication of Ginkgo biloba Herbal Products by a Novel Quantitative Real-Time PCR Approach

Ginkgo biloba is a widely used medicinal plant. Due to its potential therapeutic effects, it is an ingredient in several herbal products, such as plant infusions and plant food supplements (PFS). Currently, ginkgo is one of the most popular botanicals used in PFS. Due to their popularity and high cost, ginkgo-containing products are prone to be fraudulently substituted by other plant species. Therefore, this work aimed at developing a method for G. biloba detection and quantification. A new internal transcribe spacer (ITS) marker was identified, allowing the development of a ginkgo-specific real-time polymerase chain reaction (PCR) assay targeting the ITS region, with high specificity and sensitivity, down to 0.02 pg of DNA. Additionally, a normalized real-time PCR approach using the delta cycle quantification (ΔCq) method was proposed for the effective quantification of ginkgo in plant mixtures. The method exhibited high performance parameters, namely PCR efficiency, coefficient of correlation and covered dynamic range (50–0.01%), achieving limits of detection and quantification of 0.01% (w/w) of ginkgo in tea plant (Camellia sinensis). The quantitative approach was successfully validated with blind mixtures and further applied to commercial ginkgo-containing herbal infusions. The estimated ginkgo contents of plant mixture samples suggest adulterations due to reduction or almost elimination of ginkgo. In this work, useful and robust tools were proposed to detect/quantify ginkgo in herbal products, which suggests the need for a more effective and stricter control of such products.


Introduction
Ginkgo (Ginkgo biloba L.) is a millenary Chinese tree that belongs to the Ginkgoaceae family whose leaves are widely used for medicinal purposes [1]. Owing to its composition in pharmacologically active compounds, such as flavonol glycosides and terpene trilactones (bilobalides and ginkgolides) [2,3], ginkgo is used for its capacity to improve cognitive impairment in the elderly and quality of life in mild dementia. It is also known for its therapeutic action in peripheral circulatory illnesses, improving blood circulation and preventing clot formation [1,2,[4][5][6]. Currently, different herbal products that have ginkgo as an ingredient are readily available in the global market, including in plant food supplements (PFS) and herbal infusions. According to recent surveys, ginkgo was the most popular botanical in PFS and is used in six European Union countries [7], while in the United States it ranked among the top 10 dietary supplements in the category of herbal/botanicals [8]. Moreover, the global market of G. biloba extracts, mainly intended for pharmaceutical and food supplement industries, was estimated to be US $1590.5 million in 2018 and projected to reach US $2379.2 million by 2028 [9].

Plant Species and Commercial Samples
Leaves from G. biloba were kindly provided by the Botanical Garden of University of Porto, Botanical Garden of Bern, Serralves Garden and Botanical Garden of Madeira (Table S1, Supplementary Material). Leaves or seeds of 73 plant species corresponding to medicinal plants, fruits and spices were used for cross-reactivity testing (Table S1, Supplementary Material). A total of 20 herbal infusions were bought at local stores, including specialized herbalists, and from the internet (Table 1).
For method development, model mixtures with known amounts of dried leaves of G. biloba in Camellia sinensis were prepared to contain 50%, 10%, 5%, 1%, 0.5%, 0.1%, 0.05% and 0.01% (w/w). Firstly, a reference mixture with 50% of G. biloba was prepared by adding 10 g of ground G. biloba leaves to 10 g of ground plant material of C. sinensis. All the subsequent mixtures were prepared by sequential additions of C. sinensis plant material up to the level of 0.01% (w/w). For method validation, blind mixtures were independently prepared as described for the reference mixtures, with the proportion of 20%, 8%, 2%, and 0.2% (w/w) of G. biloba in C. sinensis plant material and were further analyzed as unknown samples.
Seeds were ground with a mortar, while the leaves and herbal infusions were ground in a laboratory mill Grindomix GM200 (Retsch, Haan, Germany).

DNA Extraction
The NucleoSpin Plant II kit (Macherey-Nagel, Düren, Germany) was chosen to perform the DNA extraction from 50 mg of each sample, according to the manufacturer's instructions with slight modifications, as described by Costa et al. [22]

DNA Quality and Purity
The yield and purity of DNA extracts were assessed by UV spectrophotometry using a Take3 micro-volume plate accessory, on a Synergy HT multi-mode microplate reader (BioTek Instruments, Inc., Winooski, VT, USA). The nucleic acid protocol was set for double-strand DNA in the Gen5 data analysis software version 2.01 (BioTek Instruments, Inc., Winooski, VT, USA), which was applied to absorbance data measured at 260 and 280 nm.
The quality of DNA extracts was further assessed by electrophoresis with 1% of agarose gel as previously described [22].

Target Gene Selection, Oligonucleotide Primers and Probes
A set of primers (Gkb2-F/Gkb2-R) and a specific probe (Gkb2-P) labelled with fluorescein (FAM) as a fluorescent reporter and black hole quencher 1 BHQ-1 as quencher, were designed to target the Internal Transcribed Space (ITS) region of G. biloba (GenBank: Y16892.1) ( Table 2). In silico analysis of sequences and primers was performed using the BLAST and Primer-BLAST tools to verify fragment and primer specificity, respectively. OligoCalc software was used to check primer properties and ensure the absence of primer hairpins and self-hybridization. Table 1. Results of the application of qualitative PCR and normalized quantitative real-time PCR systems targeting the ITS1 region (primers Gkb2-F/Gkb2-R) for the detection and quantification of G. biloba in samples of herbal infusions and a universal gene (primers EG-F/EG-R).

Samples Label
Qualitative PCR a Quantitative Real-Time PCR Ginkgo  To ensure the presence of amplifiable DNA, a universal eukaryotic primer pair (EG-F/EG-R), targeting a conserved 18S rRNA nuclear region, was used [23]. The same primer pair together with a probe (EG-P) was used as an endogenous control gene for developing the normalized real-time PCR system [24]. The primers and probes were synthesized by Eurofins MWG Operon (Ebersberg, Germany).  Table 2). The reactions were carried out in a MJ Mini™ Gradient Thermal Cycler (Bio-Rad Laboratories, Hercules, CA, USA), using the following optimized programs: initial denaturation at 95 • C for 5 min; 35 or 40 cycles (for EG-F/EG-R or Gkb2-F/Gkb2-R primers, respectively) of amplification at 95 • C for 30 s, 63 • C or 62 • C (for EG-F/EG-R or Gkb2-F/Gkb2-R primers, respectively) for 30 s and extension at 72 • C for 30 s; and a final extension at 72 • C for 5 min.

Qualitative PCR
PCR products were further analyzed by electrophoresis in a 1.5% agarose gel stained with 1× Gel Red (Biotium, Hayward, CA, USA) and running in 1× SGTB buffer (GRISP, Porto, Portugal) for 20-25 min at 200 V. Each extract was amplified in at least two independent assays.

Real-Time PCR
The reactions were performed using 20 µL of total reaction volume, containing 2 µL of DNA (20 ng), 1× SsoFast Probes Supermix (Bio-Rad Laboratories, Hercules, CA, USA), 300 nM or 400 nM of each primer set (EG-F/EG-R or Gkb2-F/Gkb2-R, respectively) and 200 nM of each probe (EG-P or Gkb2-P, for eukaryotic and G. biloba genes, respectively). A fluorometric thermal cycler CFX96 Real-time PCR Detection System (Bio-Rad Laboratories, Hercules, CA, USA) was used to amplify, simultaneously and in parallel reactions, each target sequence, under the following conditions: 95 • C for 5 min, 45 cycles at 95 • C for 15 s and 65 • C for 45 s, and the fluorescence signal was collected at the end of each cycle. The data evaluation, from each real-time PCR assay, was made using the software Bio-Rad CFX Manager 3.1 (Bio-Rad Laboratories, Hercules, CA, USA). Real-time PCR assays were performed, at least, in two independent runs using n = 3 or n = 4 replicates in each one.
For the construction of a calibration curve and for the determination of the absolute limits of detection (LOD) and quantification (LOQ), 10-fold serially diluted ginkgo DNA extracts (20 ng-0.002 pg) were amplified by real-time PCR. Additionally, a normalized calibration model was constructed based on the parallel amplification of the ITS1 region of G. biloba (target sequence) and the 18S rRNA gene (reference for eukaryotes) using the model mixtures (0.01-50%) of G. biloba in C. sinensis. The acceptance criteria established for real-time PCR assays were the PCR efficiency between 90-110%, the slope within −3.6 and −3.1 and the correlation coefficient (R 2 ) above 0.98 [25,26]. The lowest amplified level for 95% of the replicates was considered as the LOD and the LOQ was set as the lowest amplified level within the linear dynamic range of the calibration curve, which should cover a minimum of 4 orders of magnitude and should extend to ideally 5 or 6 log 10 concentrations [25,26].

DNA Quality and Selection of Target Region
In general, DNA extracts from the leaves, seeds and commercial samples showed adequate yields and purities, being in the range of 17.6-270.8 ng/µL and 1.4-2.1, respectively. Before the G. biloba specific amplification of target region, all extracts were tested by PCR targeting a universal eukaryotic region (EG-F/EG-R) to check the capacity of DNA amplification and avoid false negatives [23]. All DNA extracts used for reactivity testing were amplified (Table S1, Supplementary Material).
So far, different regions have been assessed, either as a single locus or in combination, for their adequacy as barcode markers in plant species, which include matk, rbcL, ITS and ITS2, among others [27]. In this work, the non-coding ITS region of nuclear ribosomal DNA was selected due to its high power of species discrimination over plastid regions, allowing the differentiation of closely related species [27][28][29]. This region has been previously proposed for the development of PCR assays using species-specific primers aiming at identifying medicinal plant species, with high specificity and sensitivity [30,31]. The specificity of the newly designed primers (Gkb2-F/Gkb2-R) was initially in silico verified and subsequently assayed experimentally against different DNA extracts from several plant species (n = 73). As expected, the primers proved to be specific since only the DNA extracts from G. biloba were amplified (Table S1, Supplementary Material). Afterwards, the optimized species-specific PCR assay, using a 10-fold serially diluted G. biloba DNA extract (20 ng), showed a sensitivity down to 0.002 ng ( Figure S1, Supplementary Material) and was further applied in the analysis of the commercial samples ( Table 1). The achieved sensitivity was much higher than that obtained by the RPA-lateral flow strip device reported by Liu et al., which was approximately 1 ng of purified DNA. Moreover, only a few plant species were used for cross-reactivity testing (Crataegus pinnatifida, Epimedium brevicornu, Selaginella tamariscina and Arisaema heterophyllum) by those authors. In the same work, a species-specific PCR assay targeting G. biloba DNA was also developed, but again with very limited specificity testing (only against S. japonica).

Method Development
Following the demonstrated suitability of the proposed primers for G. biloba specific detection, a real-time PCR method was developed using a newly designed hydrolysis probe (Gkb2-P), increasing the sensitivity and specificity of the assay. Figure 1 presents the real-time PCR amplification curves and respective calibration curve using a 10-fold serially diluted ginkgo DNA extract. The average parameters of PCR efficiency (101.4%), slope (−3.284) and R 2 (0.988) were all within the acceptance criteria ( Figure 1B), suggesting a high performance of the assay [25,26]. The dynamic range covered six orders of magnitude of the target analyte (20 ng to 0.02 pg of ginkgo DNA) and the absolute LOD of the real-time PCR assay was established as 0.02 pg of G. biloba DNA, corresponding to 0.285 genomic DNA copies (using the mean value of the Plant DNA C-value database [32]) and considering the amplification of all replicates (n = 6 from two independent assays). Since the LOD value was within the linear dynamic range of the calibration curve, the LOQ value was set at the same value (0.02 pg) [25,26]. For establishing a quantitative model of ginkgo in herbal material, a normalized real-time PCR assay using the ∆Ct method was developed. This approach accounts with amplification variations due to inconsistent DNA recovery and quality/degradation among extracts as a result of processing [24,[33][34][35]. It relies on the construction of a normalized calibration curve using the cycle of quantitation (Cq) values from the target region (ITS1) and a reference endogenous gene (nuclear 18S rRNA) by applying the expression ΔCq = Cq (ginkgo)-Cq (universal gene). The normalized calibration curve was obtained by plotting the calculated ΔCq values versus the logarithm of the gingko concentration, using the binary mixtures with known quantities of G. biloba in C. sinensis For establishing a quantitative model of ginkgo in herbal material, a normalized real-time PCR assay using the ∆Ct method was developed. This approach accounts with amplification variations due to inconsistent DNA recovery and quality/degradation among extracts as a result of processing [24,[33][34][35]. It relies on the construction of a normalized calibration curve using the cycle of quantitation (Cq) values from the target region (ITS1) and a reference endogenous gene (nuclear 18S rRNA) by applying the expression ∆Cq = Cq (ginkgo)−Cq (universal gene). The normalized calibration curve was obtained by plotting the calculated ∆Cq values versus the logarithm of the gingko concentration, using the binary mixtures with known quantities of G. biloba in C. sinensis (50.0%, 10.0%, 5.0%, 1.0%, 0.5%, 0.1%, 0.05%, and 0.01%, w/w) (Figure 2). The choice of C. sinensis, also commonly known as the "tea plant", to prepare the reference mixtures was based on the high frequency of its use in mixed herbal infusions. The developed normalized real-time PCR approach exhibited high performance, as inferred from the obtained parameters of PCR efficiency (96.2%), R 2 (0.982) and slope (−3.417) (mean values from 6 independent assays), covering 7 magnitude orders, which were all within the acceptable criteria. The approach enabled an LOD and LOQ down to 0.01% (w/w) (n = 12 from 3 independent assays), corresponding to 0.1 g of G. biloba per 1 kg of C. sinensis.
Foods 2020, 9, x FOR PEER REVIEW 8 of 12 (50.0%, 10.0%, 5.0%, 1.0%, 0.5%, 0.1%, 0.05%, and 0.01%, w/w) (Figure 2). The choice of C. sinensis, also commonly known as the "tea plant", to prepare the reference mixtures was based on the high frequency of its use in mixed herbal infusions. The developed normalized real-time PCR approach exhibited high performance, as inferred from the obtained parameters of PCR efficiency (96.2%), R 2 (0.982) and slope (−3.417) (mean values from 6 independent assays), covering 7 magnitude orders, which were all within the acceptable criteria. The approach enabled an LOD and LOQ down to 0.01% (w/w) (n = 12 from 3 independent assays), corresponding to 0.1 g of G. biloba per 1 kg of C. sinensis. Compared with the recent report of Dhivya et al. [21], describing a species-specific real-time PCR with a hydrolysis probe targeting the matk gene, the present approach achieved similar performance parameters in terms of PCR efficiency and R 2 using serially diluted leaf DNA of ginkgo. However, the proposed real-time PCR method provides a much wider dynamic range (seven orders of magnitude) and a higher sensitivity (0.02 pg of ginkgo DNA) than that obtained by Dhivya et al. [21] (five orders of magnitude and 10 pg of ginkgo DNA). Regarding specificity, the proposed primers and probe targeting the ITS region do not provide any cross-reactivity with any of the known potential adulterants (Sophora japonica and Fagopyrum esculentum Moench) ( Figure S2), while the method of Dhivya et al. [21] was reactive with S. japonica at late amplification cycles, which compromised its sensitivity, and the potential reactivity with F. esculentum Moench was not verified by the referred authors. Therefore, the proposed method demonstrated full specificity and high sensitivity for gingko detection, with the important achievement of providing, for the first time, a normalized quantitative real-time PCR approach to enable a determination of the proportion of ginkgo in herbal products.

Method Validation
To proceed with the validation of the method, the precision and accuracy should also be evaluated [25,26]. Therefore, blind mixtures containing 20.0%, 8.0%, 2.0%, and 0.2% (w/w) of G. biloba in C. sinensis were used. The results regarding the estimated values (%) of ginkgo and the comparative analysis with the real values are presented in Table 3. The obtained values exhibited adequate coefficients of variation (CV), which were between 5.6-17.9% and, therefore, lower than the maximum acceptable (25%), demonstrating the high precision of the method over the considered Compared with the recent report of Dhivya et al. [21], describing a species-specific real-time PCR with a hydrolysis probe targeting the matk gene, the present approach achieved similar performance parameters in terms of PCR efficiency and R 2 using serially diluted leaf DNA of ginkgo. However, the proposed real-time PCR method provides a much wider dynamic range (seven orders of magnitude) and a higher sensitivity (0.02 pg of ginkgo DNA) than that obtained by Dhivya et al. [21] (five orders of magnitude and 10 pg of ginkgo DNA). Regarding specificity, the proposed primers and probe targeting the ITS region do not provide any cross-reactivity with any of the known potential adulterants (Sophora japonica and Fagopyrum esculentum Moench) ( Figure S2), while the method of Dhivya et al. [21] was reactive with S. japonica at late amplification cycles, which compromised its sensitivity, and the potential reactivity with F. esculentum Moench was not verified by the referred authors. Therefore, the proposed method demonstrated full specificity and high sensitivity for gingko detection, with the important achievement of providing, for the first time, a normalized quantitative real-time PCR approach to enable a determination of the proportion of ginkgo in herbal products.

Method Validation
To proceed with the validation of the method, the precision and accuracy should also be evaluated [25,26]. Therefore, blind mixtures containing 20.0%, 8.0%, 2.0%, and 0.2% (w/w) of G. biloba in C. sinensis were used. The results regarding the estimated values (%) of ginkgo and the comparative analysis with the real values are presented in Table 3. The obtained values exhibited adequate coefficients of variation (CV), which were between 5.6-17.9% and, therefore, lower than the maximum acceptable (25%), demonstrating the high precision of the method over the considered dynamic range.
Regarding the accuracy, three out of the four blind mixtures presented bias values in the range of 5.6-17.9%, being within the recommended range (±25%) [26]. Although the mixture with 0.2% (w/w) presented a slightly higher error (−27.4%), this is the lowest tested level, not likely to occur due to adulteration, but rather from contamination. Besides, according to Kang [36], bias within 25-30% have been considered as acceptable in real-time PCR methods for food analysis. Table 3. Results of the validation assays using the normalized quantitative PCR system applied to blind mixtures of G. biloba in C. sinensis.

Analysis of Commercial Herbal Infusions
For assessing the applicability of the method, the normalized real-time PCR system was used to analyze and further verify the authenticity and labelling compliance of several commercial herbal products (herbal infusions). The analyzed herbal infusions were all labelled as containing ginkgo, wholly or partially (Table 1). All the samples produced amplifiable DNA extracts, which were positive for the ginkgo-specific PCR assay. The samples of mixed herbal species were further assayed by quantitative real-time PCR to assess their ginkgo content. The quantitative results demonstrated that, out of five samples of herbal mixtures with labelled ginkgo contents, four samples (#4, #10, #11 and #16) declared 15% of ginkgo, but the obtained contents were within 0.01-2.98%. In particular, sample #10 had only trace amounts (0.01%) of gingko, suggesting its complete substitution with other plant(s). Sample #12 declared 30% of ginkgo, but the obtained content was 9.95%. Consequently, the results of samples #4, #11, #12, and #16 suggest the partial substitution of ginkgo with other plant(s). The other two mixed herbal samples (#14, #15) did not provide any quantitative information regarding gingko, having low estimated amounts (<3%), suggesting again its reduced use. Therefore, the results of mixed herbal products strongly suggest the practice of adulterations, probably due to the high market price of G. biloba and its increasing demand, with the industries using less quantity than they declared to raise their profits.

Conclusions
In the herein presented work, a new molecular marker of the ITS region was identified for the species-specific detection of G. biloba by both qualitative PCR and real-time PCR with a TaqMan probe, providing high specificity and sensitivity, down to 0.02 pg of DNA (0.285 genomic DNA copies). For the effective quantification of ginkgo in herbal products, a novel normalized real-time PCR system based on the ∆Cq method was successfully developed using reference herbal mixtures. The method exhibited high performance parameters, namely PCR efficiency, coefficient of correlation and covered dynamic range (50-0.01%), achieving a LOD and LOQ of 0.01% (w/w) of ginkgo in tea plant. The quantitative approach was further validated with blind mixtures, demonstrating accuracy, repeatability, and trueness within the range of 20-2%. The applicability of the PCR approaches was demonstrated using a set of commercial ginkgo-containing herbal infusions (n = 20), confirming the presence of ginkgo in all the products. However, the obtained quantitative results regarding the estimated ginkgo content of seven herbal mixture samples suggest adulterations due to reduction or almost elimination of ginkgo. The proposed system was demonstrated to be a powerful and robust tool for control laboratories and regulatory authorities to ensure labelling compliance of ginkgo-containing herbal products. Since it was demonstrated that the developed method has a high specificity and sensitivity, it can potentially be useful for further detecting G. biloba in other processed herbal products or foods.