Chemotaxonomic Monitoring of Genetically Authenticated Amomi Fructus Using High-Performance Liquid Chromatography–Diode Array Detector with Chemometric Analysis

Amomi Fructus is widely used to treat digestive disorders, and Amomum villosum, A. villosum var. xanthioides, and A. longiligulare are permitted medicinally in national pharmacopeias. However, there are a variety of adulterants present in herbal markets owing to their morphological similarities to the genuine Amomum species. Forty-two Amomi Fructus samples from various origins were identified using internal transcribed spacer and chloroplast barcoding analyses, and then their chromatographic profiles were compared using chemometric analysis for chemotaxonomic monitoring. Among the Amomi Fructus samples, A. villosum, A. longiligulare, A. ghaticum, and A. microcarpum were confirmed as single Amomum species, whereas a mixture of either these Amomum species or with another Amomum species was observed in 15 samples. Chemotaxonomic monitoring results demonstrated that two medicinal Amomum samples, A. villosum and A. longiligulare, were not clearly distinguished from each other, but were apparently separated from other non-medicinal Amomum adulterants. A. ghaticum and A. microcarpum samples were also chemically different from other samples and formed their own species groups. Amomum species mixtures showed diverse variations of chemical correlations according to constituent Amomum species. Genetic authentication-based chemotaxonomic monitoring methods are helpful in classifying Amomi Fructus samples by their original species and to distinguish genuine Amomum species from the adulterants.


Introduction
Amomi Fructus (Sa-In) has been used to treat digestive disorders associated with excessive dampness and originates from the ripe fruits or seed clusters of Amomum villosum Lour. and A. villosum var. xanthioides (Wall. ex Baker) T.L. Wu and S.J. Chen in the Korean pharmacopeia [1], whereas the ripe fruit of A. longiligulare T.L. Wu including those of the two species mentioned is additionally registered Molecules 2020, 25,4581 2 of 17 in the Chinese pharmacopeia [2]. However, the commercial status of Amomi Fructus in herbal markets does not correspond to official documents; that is, unofficial Amomum species or even Alpinia species are often sold and used as counterfeits or adulterants of Amomi Fructus [3,4]. This is because the fruits of A. villosum are recognized as being of high commercial grade among official Amomum species and consequently, are sold at a high price in herbal markets. Particularly, the indistinguishable organoleptic properties among official Amomum species and the adulterants exacerbate the misuse of Amomi Fructus. Owing to their morphological similarity, Amomi Fructus from different Amomum species cannot be clearly discriminated through macroscopic observation, which creates confusion in their medicinal application. Therefore, more accurate and precise techniques using genetic information have been developed to identify Amomi Fructus from other counterfeits. Internal transcribed spacer (ITS) and matK DNA sequencing [5], nuclear ribosomal DNA (nrDNA) ITS1 sequencing [6], genomic DNA barcoding using ITS [3] or single nucleotide polymorphism (SNP) genotypes [7], and chloroplast genome sequencing [8][9][10] have been used to identify various Amomum species at the genetic level.
Chemotaxonomic monitoring of herbal medicines using the chromatographic profiling method is another approach to identify and distinguish Amomi Fructus from various Amomum species. As Amomi Fructus is chemically characterized by abundant volatile oil, gas chromatography-mass spectrometry (GC-MS) has mainly been used to determine chemical constituents in the volatile oil from Amomum species [11][12][13][14]. Furthermore, Amomi Fructus samples originating from A. villosum and A. villosum var. xanthioides were chemically distinguished by the composition of their volatile oils using GC-MS combined with chemometric techniques [4]. However, the chemotaxonomic approach based on volatile constituents is insufficient to guarantee the chemical features of Amomi Fructus as these oils are easily vaporized when extracted at high temperature. Non-volatile constituents in Amomi Fructus remain after hot extraction and hence, these can possibly exert therapeutic activity when administered to patients. Previous studies reported the isolations of non-volatile constituents: flavonoids (quercetin, quercitrin, and isoquercitrin); phenolic acids (vanillic acid and 3,4-dihydroxy-benzoic acid), coumarins (flavanocoumarin and isoflavanocoumarin), and steroids (daucosterol, stigmasterol, and ergosterol) [15,16]. Consequently, chemical characterization using non-volatile constituents that are determined using high-performance liquid chromatography (HPLC) is also crucial to evaluate the chemotaxonomic discrimination of Amomum species. Several studies have reported the chemical discrimination of A. villosum and its adulterants using chromatographic fingerprinting with HPLC-based statistical analysis [17,18]. Although there is a limitation that Amomi Fructus samples used in those studies were not clearly authenticated before the HPLC experiment, chemical differences were observed.
To overcome the inaccuracy of herbal samples in chemotaxonomic analysis, a process of 'genetic authentication coupled chromatographic profiling' is required to support chromatography-based chemotaxonomic results. Currently, chemical analysis combined with DNA barcoding techniques is recognized as a powerful tool for identification and differentiation in the chemotaxonomy of herbal medicines [19,20]. Recent studies reported that the combination of DNA barcoding and HPLC fingerprinting was applied to identify Salvia species [21], Phellodendron species [22], and Daphne species [23]. Our research group also established a 'genetic authentication-coupled chromatographic profiling' process to chemically distinguish Atractylodes species using ITS sequencing, HPLC analysis, and chemometric tools [24,25]. Therefore, we collected 42 Amomi Fructus samples and genetically authenticated these using ITS sequences from nrDNA and chloroplast genome-based DNA barcoding analysis. A HPLC-diode array detector (HPLC-DAD) was used to construct the chromatographic profiles for chemical monitoring of Amomi Fructus samples.

ITS Regions of the Nuclear Ribosomal Cistron
To identify the species of the 42 Amomi Fructus samples, the nucleotide sequences of the ITS region were analyzed. Approximately 645-665 bases of amplified product sequence were identified based on the samples listed in Tables 1 and 2. As shown in our previous study [3], several adulterants were also observed in this study. Especially this time Amomi Fructus samples were collected from diverse countries where they were used for not only medical purposes but also as food products. Therefore, various adulterants of Amomum species were identified and some of these were mixed with each other, and especially, the mixture of species was mostly shown in the samples from Southeast Asia ( Table 2). Most of the samples from China were identified as A. villosum or A. longiligulare, except for two samples, AGS01 (A. ghaticum) and AMS04 (A. microcarpum). In contrast, NCBI BLAST analysis confirmed A. ghaticum, A. microcarpum, A. uliginosum, and A. longiligulare in the samples from Myanmar. In addition, A. microcarpum, A. longiligulare, A. villosum, A. tomrey var. tomrey, A. koenigii, A. chinense, and A. echinocarpum were also observed. The sequence identity matrix ranged from 0.989 to 0.883 in adulterants of Amomum, whereas it ranged from 0.885 to 0.841 in adulterants of Alpinia (Table S1, Figure S1).

Chloroplast Genome-Based DNA Barcode Sequence Analysis
According to the previous study, three chloroplast genome areas (rbcL, matK, and trnL-F intergenic spacer) were used as support for the identification result of the ITS region [3]. The psbA-trnH region was additionally analyzed to identify adulterant Amoum species in the work. Sequence characteristics of the four plastid regions and one nuclear region are presented in Table 3. The rbcL and trnL-F intergenic spacer regions showed higher variable sites among the four plastid regions (except the ITS region). The sequences of matK and psbA-trnH were the most conserved among the five regions analyzed when the aligned length and number of conserved sites were considered. Even though the variable sites of the four plastid loci were much lower than those of the ITS region and highly conserved, they could separate the Amomi Fructus species from those of the adulterants in Amomum and Alpinia and the results were consistent with the analysis of the ITS region. For more details regarding the four plastid loci, partial nucleotide sequences of the 933 bases in matK were determined using the 390F/1326R primer set. Furthermore, the sequence identity matrix between Amomi Fructus and the adulterants was considerably closer than the results with ITS; i.e., the minimum identity matrix result was 0.979 between Amomi Fructus and the adulterants including Alpinia (Table S1). Although the matK could not distinguish between A. villosum and A. uliginosum, it could classify the rest of species in this work.
For the rbcL gene sequence, we determined partial nucleotide sequences of 743 bases in all the samples listed in Tables 1 and 2 using the rbcL a-f/724R primer set. The number of variable sites in the nucleotide sequence was lower than that in the ITS results and some of the adulterants in Amomum shown identical results; however, the three genuine species of Amomi Fructus could still be distinguished as shown in the previous study [3]. The minimum sequence identity matrix of rbcL among samples in Tables 1 and 2 was nearly 0.900 (except for the Amomi Tsao-ko Fructus samples whose minimum sequence identity matrix was 0.897).
For the psbA-trnH intergenic spacer, partial nucleotide sequences of 800-836 bases were determined using the trnH2/psbAF primer set. Although the psbA-trnH intergenic spacer showed the most conserved sites with the matK area, the sequence length was different depending on the species. Thus, the sequence identity matrix range was 0.997-0.914 and well classified the species of Amomi Fructus from several adulterants as was observed with the ITS region.
In the case of the trnL-F intergenic spacer, amplified products with 395-415 bases were determined from the samples listed in Tables 1 and 2. The amplified product size was different depending on the species and the sequence identity matrix between Amomi Fructus and the adulterants ranged from 0.990 to 0.88.

Phylogenetic Analysis
The PhyML+SMS tree was constructed based on the concatenated nucleotide sequences of the ITS and the four chloroplast genome-based DNA barcode regions (Figure 1). The species in the phylogenetic tree were well separated by their original species, which supported the accuracy of the identification result based on the ITS region. The adulterants derived from Alpinia were divided into different clades distant from the plants in the Amomum group. Among the species in Amomum, three original species of Amomi Fructus, A. longiligulare, A. villosum, and A. villosum var. xanthioides were in group I and A. ghaticum, A. villosum var. villosum, and A. uliginosum were grouped very close together. A. microcarpum was also located close to group I. In addition, A. tomrey var. tomrey, A. chinense, A. echinocarpum, and A. koenigii were assembled close together in group II.

Optimization of Sample Extraction and Analytical Conditions
The extraction of the representative Amomi Fructus sample (ALS01 in Table 1) was performed using the rapid, efficient, and simple ultrasonic extraction method. Extraction times (10, 30, and 50 min) and solvent ratios (methanol:water = 1:0, 0:1, and 1:1) were compared and the extraction for 30 min with solvent mixture (methanol:water = 1:1) was selected to obtain a higher absolute area of the selected peaks in the chromatogram of ALS01.
HPLC analytical conditions were optimized in terms of the mobile phase modifier, mobile phase composition, and UV detection wavelength. Water with 0.1% formic acid (v/v), water with 0.1% TFA (v/v), and water without a modifier were compared with various mobile phase ratios of acetonitrile and water. Water containing 0.1% TFA as the aqueous mobile phase produced better interpeak separations and peak shapes with higher peak areas. Six detection wavelengths were selected in the DAD according to the optimal UV absorbance of each peak: 11 peaks at UV 225 nm, 7 peaks at UV 265 nm, 16 peaks at UV 280 nm, 9 peaks at UV 290 nm, 18 peaks at UV 310 nm, and 16 peaks at UV 320 nm (Table S2). Peak No. 18 and peak No. 35 were identified as vanillic acid and quercitrin, respectively, by comparison of their retention times and UV detection wavelengths with those of purchased standard compounds.

Chromatographic Profiling of Amomi Fructus Samples
Overlapped chromatograms of Amomi Fructus samples were mostly similar within interspecies and between interspecies of Amomum, with the exception of AVS02 within A. villosum samples, and A. microcarpum samples among other Amomum species, particularly after 40 min of retention time (Figure 2). Chromatograms of Amomi Fructus samples mixed with more than two Amomum species showed similar chromatographic patterns according to those of their original species; however, the species-mixed samples whose original species was not obtained in this study (SM07, -09, -11, and -13) showed distinguishable chromatographic patterns from those of the Amomum samples listed above (Figures S2 and S3). retention time (Figure 2). Chromatograms of Amomi Fructus samples mixed with more than two Amomum species showed similar chromatographic patterns according to those of their original species; however, the species-mixed samples whose original species was not obtained in this study (SM07, -09, -11, and -13) showed distinguishable chromatographic patterns from those of the Amomum samples listed above (Figures S2 and S3). There were a number of peaks that showed significant differences in average peak areas between two Amomum species: 6 peaks between A. villosum and A. longiligulare samples (peak 27, There were a number of peaks that showed significant differences in average peak areas between two Amomum species: 6 peaks between A. villosum and A. longiligulare samples (peak 27 Figure S4).

Clustering Analysis of Amomi Fructus Samples Using Chemometric Statistical Methods
Chemotaxonomic classification of Amomi Fructus samples was estimated using chemometric clustering tools with principle component analysis, hierarchical clustering analysis (HCA), and heatmap analysis, and Pearson's correlation coefficients were calculated to investigate the chemical relationship between the Amomi Fructus samples.
In the principal component (PC) plot, the samples of single Amomum species did not form distinct clusters separately by their original species; i.e., those samples were not clearly grouped, as single species samples and species-mixed samples were colocated within the middle range of the PC1 and PC2 scores. However, one A. villosum sample (AVS06) and four species-mixed samples (SM07, -09, -11, and -13) were exceptionally apart from the samples distributed together by their wide ranges in PC1 and PC2 scores (Figure 3).
The dendrogram from the HCA showed distinct clusters of Amomi Fructus samples according to their original Amomum species. One A. villosum sample (AVS01) and two A. longiligulare samples co-existed exclusively in the A. longiligulare group and A. villosum group, respectively. In contrast, A. ghaticum and A. microcarpum samples formed their original cluster groups without the coexistence of other species. Like in the PC plot, four species-mixed samples (SM07, -09, -11, and -13) were unambiguously clustered in a separate group (Figure 4).  In the cluster-based heatmap analysis, apparent clusters of Amomi Fructus samples were also observed by their species and the classification of single Amomum species samples were analogous to those in the dendrogram of HCA: two A. villosum + A. longiligulare groups, an A. ghaticum group, and an A. microcarpum group. Exceptionally, some samples of A. villosum and A. longiligulare were differently clustered in between the A. villosum + A. longiligulare groups, compared with the HCA dendrogram (AVS02 moved in, and ALS03 and -05 moved out). Such a shift was also observed in the species-mixed samples (SM05, -06, and -15). Consistent with the principle component analysis and HCA results, SM07, -09, -11, and -13 samples were also obviously distinguished ( Figure 5).

Similarity Evaluation of Amomi Fructus Samples Using the Pearson's Correlation Coefficient
Correlation between individual Amomi Fructus samples was measured using the Pearson's correlation coefficient (r; Table S4). The ranges of the mean and median values of the coefficients between each Amomi Fructus sample and the remaining samples (rest) were as follows:   (Table 4, Figures S5-S9).

Discussion
Amomum is currently recognized as the second largest genus in Zingiberaceae with approximately 150-180 species [5,26,27]. The distribution area of Amomum is from Sri Lanka to the Himalayas, China, Southeast Asia, Malaysia, and Northern Australia. Especially, the forests of Southeast Asia are treated as the center of endemism [28]. It suggests there is a possibility that another species of adulterant may be found, such as A. ghaticum, a newly found species within Amomi Fructus in our previous study [3].
In the 42 Amomi Fructus samples collected in this study, we also identified several adulterant species in Amomum and some of these were not previously mentioned as an adulterant of Amomi Fructus in the literature. The typical examples were A. uliginosum (SM04, SM05, and SM15), A. ghaticum (AGS01-AGS09, SM03, and SM04), A. tomrey var. tomrey (SM07 and SM09), and A. echinocarpum (SM11 and SM13). Additionally, seven A. microcarpum (AMS01-04, SM01, SM02, and SM14), four A. koenigii (SM07, SM09, SM11, and SM13), and three A. chinense (SM09, SM11, and SM13) were found in this study and they are recoded as 'Se-Sa-In (xi sha ren)', 'Ya-Cho-Gwa (ye cao guo)', and 'Hae-Nam-Ga-Sa-In (hai nan jia sha ren)' in the Flora of China. SM05 and SM15, identified as A. villosum var. villosum (KJ151892 and MH161417) through the NCBI BLAST analysis, are recorded as autonyms of A. villosum in the Flora of China. A. uliginosum is a widespread species in Laos, Cambodia, Vietnam, Thailand, Peninsular Malaysia, and the Sumatra area, and is very similar to A. villosum (continental Southeast Asia) in the shape of the flowering head and the small, dark-red, prickly fructus [28]. They showed similarities not only in the morphological feature, but also in the close sequence identity matrix in all five DNA barcode regions. The results of the phylogenetic analysis inferred through ITS nucleotide sequences showed that samples of genuine Amomum species formed a closer relationship and their groups were apparently distinguished from the other Amomum and Alpinia species, which were recognized as adulterants. Our results also support the clades of taxa in the Alpinioideae that include species of Amomum as proposed by the previous study [5].
Chemotaxonomic relevance among Amomum species was presented using chromatographic profiling combined with chemometric analysis and chemical characteristics of samples were obviously differentiated by their own species. The concordance rate of clustering single Amomum species samples into each independent species group in HCA was 87.5%, 66.7%, and 100% for samples of A. villosum, A. longiligulare, and both A. ghaticum and A. microcarpum, respectively. Particularly, A. villosum and A. longiligulare samples medicinally permitted in the Korean and Chinese pharmacopeias were chromatographically analogous to each other and they were distinctly separated from the other non-medicinal Amomum species in the cluster analysis. Although most A. villosum samples were distinguished from A. longiligulare samples, a few A. villosum or A. longiligulare samples interrupted the apparent distinction between the two species; i.e., AVS01, -02 and ALS01, -06 showed higher correlations with the opposite species. A. villosum and A. longiligulare samples are genetically divided into their own Amomum species in this study as well as in a previous study [8]; however, the chemical similarities between the samples of these two species makes for an ambiguous classification using chemometric cluster analyses, as reported previously [18].
A. ghaticum, a species mainly occurring in the Western Ghats of India, mostly originated from Myanmar and was confirmed as a major adulterant species among genuine Amomi Fructus in local herbal markets in this and a previous study [3]. The samples of A. ghaticum obviously formed their own species group in the cluster analysis, indicating that this species represented distinguishable chromatographic profiles from other Amomum species, notwithstanding their macroscopic similarities. However, the correlation analysis produced contradictory results that A. ghaticum samples showed comparatively high correlation coefficients (r) with A. villosum, and especially, A. longiligulare samples, meaning strong linear relationships between the chromatographic profiles of A. ghaticum and those of the above two species [29]. In contrast, A. microcarpum samples were genetically and chemically distant from the above three Amomum species, as shown from chromatographic patterns, cluster analysis, and correlation coefficients.
Interestingly, species-mixed samples, when sharing at least one of the original species in tested samples, were correspondently included into a single Amomum species group; for example, SM08, -10, and -12 were grouped in AVS-ALS clusters. In contrast, those samples without any sharing among the four Amomum species exclusively separated from the other samples, having extremely low correlations; for example, SM07, -09, -11, and -13, which included A. tomrey var. tomrey, A. koenigii, A. chinense, and A. echinocarpum, were grouped in their own cluster and their correlation coefficients in comparison to those of other samples were near '0 or even negative. These results demonstrated that these species-mixed samples are phylogenetically distant from the medicinal Amomun species and are also exclusively distinguishable by chemical analysis.
Taken together, we found that there were large quantities of Amomi Fructus mixed with non-medicinal adulterants of Amomum in herbal markets. Higher chemical correlations and morphological resemblance might cause serious confusion in the Amomi Fructus trade, and consequently, can result in unintended and undesirable effects on patients and health systems. Chemotaxonomic monitoring based on phylogenetic authentication can exclude medicinally non-available adulterants from the genuine Amomi Fructus. Moreover, it is effective in corroborating morphology-based inspection and is essentially required to improve the quality of medicinally available Amomum species.
Despite this distinct classification of Amomi Fructus, we had a few limitations in this study: (1) insufficient numbers of Amomi Fructus samples per single Amomum species owing to difficulties in collecting genuine Amomum species and a mixture of diverse adulterants; (2) the absence of another medicinally available Amomi Fructus, A. villosum var. xanthioides, owing to its scarcity in local markets or natural habitats; and (3) a lack of chemical research on Amomum species other than A. villosum.
Forty-two samples of Amomi Fructus were collected from their natural habitats, agricultural fields, and local markets in Korea, China, Myanmar, and Vietnam, and were also provided by the Korea Institute of Oriental Medicine (Table 1). For species identification of Amomum Fructus, the vouchers of Amomi Fructus and adulterants identified previously [3] were used for comparison with the newly collected samples in this study (Table 2). All the samples and isolated genomic DNA have been deposited at the herbarium of the college of Korean Medicine in Wonkwang University.

Preparation of Genomic DNA
The genomic DNA was extracted from Amomi Fructus samples according to the NucleoSpin ® Plant II kit manual (Macherey-Nagel, Düren, Germany) with PL1 lysis buffer. For some samples, 10% cetyltrimethyl ammonium bromide and 0.7 M NaCl were used to remove the phenolic compounds and polysaccharides.

Determination of DNA Sequences of PCR Product
PCR products separated from agarose gel were cloned using the TOPcloner™ TA Kit (Enzynomics, Daejeon, Korea) and the DNA sequences of the cloned PCR products were determined by Bioneer Sequencing Service (Daejeon, Korea). For accuracy, the DNA barcode analysis process was performed independently thrice from the genome DNA preparation stage.

Preparation of Samples for HPLC Analysis
Dried Amomi Fructus samples were pulverized and homogenized through a testing sieve. A 100 mg of sample powder was weighed accurately and sonicated with 1 mL of solvent mixture (methanol:water = 1:1) using an ultrasonic extractor (Power Sonic 520; Hwashin Tech, Daegu, Korea) for 30 min. Then, the extract was centrifuged at 10,770× g for 5 min and filtered through a 0.2 µm syringe filter (BioFact; Daejeon, Korea) prior to HPLC analysis.
The precision of selected peaks was determined by analyzing their retention times and absolute areas thrice within a day (intraday precision) and over three consecutive days (interday precision). Precisions are represented as relative standard deviations (RSDs), where RSD (%) = ((standard deviation/ mean value) × 100).

Chemometric Statistical Analysis
Forty-two samples that were genetically identified were recoded and the relevance between samples was determined using chemometric tools including principal component analysis, hierarchical clustering analysis, heatmap analysis, and Pearson's correlation analysis. In total, 77 peaks were selected as profiling peaks (the area of each peak > 1.0% of the total peak area) at their optimal UV absorption, and their absolute areas were calculated by peak area integration for chromatographic fingerprinting. A matrix comprising of the rows (Amomi Fructus sample) and columns (absolute area of each profiling peak) was used to construct the principle component (PC) plot, dendrogram, heatmap, and to calculate the Pearson's correlation coefficient. Tukey's test was used to compare the absolute area of each peak among the samples of single Amomum species. Chemometric analyses and Tukey's test were conducted using the open source software R (v. 4.0.2; The R Foundation for Statistical Computing).

Conclusions
In this study, forty-two Amomi Fructus from various locational origins were phylogenetically identified using ITS and chloroplast genome-based DNA barcoding analysis and thereafter, their chromatographic profiles were compared using chemometric and correlation analysis for chemotaxonomic monitoring of Amomi Fructus. Eight A. villosum and six A. longiligulare samples showed closer genetic and chemical relations than the other non-medicinal Amomum adulterants. A. ghaticum samples represented higher chemical correlations with medicinally available A. villosum and six A. longiligulare samples than A. microcarpum samples did. Fifteen samples of Amomum species mixtures showed a variety of chemical relations with genuine Amomum species and their adulterants with four samples having zero or negative correlations. Chemotaxonomic monitoring using chromatographic profiling with chemometric analysis provides species-specific classification of Amomi Fructus in accordance with their phylogenetic relations.  Table S1: identity matrix of DNA barcode regions; Table S2: retention times and detection wavelengths of profiling peaks; Table S3: precisions; Table S4: Pearson's correlation coefficients.