Introduction

Licorice refers to roots and rhizomes of Glycyrrhiza glabra; one of the about 18 accepted congeneric species in Leguminosae family (IPNI, 2008; WCSP, 2012), originated in Mediterranean region. They are distributed and grow wild throughout the northern hemisphere. Six species of the genus were reported for the Flora Iranica by Rechinger (1984): G. bucharica Regel, G. aspera Pall., G. uralensis Fisch. ex DC., G. glabra L., G. echinata L. and G. macedonica Boiss. & Orph. Glycyrrhiza glabra L. (Syn.: G. violacea Boiss.) is represented in Iran by two resembling varieties: G. glabra var. glabra (the autonym) and
G. glabra var. glandulifera (WALDST. & KIT.) Boiss. (Rechinger, 1984).

Licorice species have been considered as the most important taxa in the genus; they have long been used as medicinal plants; their most important constituent (glycyrrhizin, C₄₂H₂₂O₁₆) is widely used as a natural sweetener and a pharmaceutical agent due to its anti-inflammatory and hepatoprotective properties (Hayashi and Sudo, 2009). Hayashi et al. (2005a) studied the relationships between 10 strains of so called economically important species of the genus Glycyrrhiza and divided them into two types (consisting GA and AT genotypes). The study failed to correlate chemotypes to genotypes, however, in other phylogenetic studies by Hayashi et al. (2000; 2005b), G. glabra was considered as a close relative to G. inflata and G. uralensis; a clade that produced glycyrrhizin as the major constituent. Other recent studies considered intra-specific variation of licorice s. l. species, quality control and authentication methods (Yao et al., 2008; Khan et al., 2009; Daei et al., 2010; Zhang et al., 2011). Variations in glycyrrhizin contents of licorice roots were evaluated by Haji-Mehdipour et al. (2008) in a number of wild populations of the species in Iran. Fars province population was reported among the three higher glycyrrhizin producing sites.

Flavonoids (also referred to as Vitamin P) are a class of plant secondary metabolites. The term is the general name of the compounds based upon a fifteen-carbon skeleton. At the simplest level, the skeleton consists of two phenyl rings (A- and B-rings) connected by a three-carbon bridge (C-ring). In general, plants alone possess the biosynthetic ability of the flavonoids (Mobh, 1939). This study was aimed to assess phenetic relationships between wild populations of the species G. glabra (licorice) in the central Zagros region of Iran (including Fars province), using quantitative morphological data, flavonoid spots and glycyrrhizin content profiles.

Material and Method

Plant material was collected from wild populations of G. glabra throughout Zagros Mountain. chain of Iran with the emphasis on central Zagros region. Locations of the studied specimens are shown on the map (Figure 1) and the corresponding list is presented in Table 1. Rhizome samples for flavonoid extraction and glycyrrhizin content analyses were selected from those rhizomes with 1-2 mm diameter collected in August-September, from the 0-300 mm soil layer (Douglas et al., 2004; Bolouri-Moghaddam et al., 2009).

Figure 1. Locations of populations where licorice specimens were collected

Table 1. Sample code, scientific names, locality, soil texture and GPS coordinates of specimens were studied.

Morphological analysis

Twenty-one quantitative morphological characters (Table 2) were measured on 27 population accessions. Each character measured up to 6 times on collected materials. Data, including length, perimeter and seed area, legumes, leaflets, etc. were obtained from calibrated digital images and entered into a raw data matrix. Averages measurements of each character were calculated and P-value was defined as (Max-Min)/Average to determine variability of each character across all samples. P- and CV (coefficient of variation) values for each character are presented in Table 2. Then, data were encoded as qualitative data and new data matrix were formed on the basis of midpoints of distribution histograms. Qualitative data matrix was used for multivariate analyses (PCA, NMDS and Cluster analysis). Multivariate analyses were performed using NTSYS-pc software package (Rohlf, 2000). Cluster analysis based on SM similarity coefficient for qualitative data and dendrogram were constructed using ME sorting method of Tamura et al. (2007).

Table 2. Quantitative morphological characters

*minimum and **maximum values of P (see text) and cv.

Profile of flavonoid spots

Flavonoids were extracted from 5 gr of dried and grinded rhizomes of each sample using methanol (80%) according to Markham (1982). Flavonoids were separated from water-insolubles by vacuum drying then dissolved in water and then further extraction by dissolving in n-butanol. Extractions were vacuum-dried and dissolved in 5 ml pure methanol. Flavonoids were separated on UV254F silica-gel thin-layers using an optimized solvent system [water: 50, ethanol: 15, butanol: 20, acetic acid: 10, chloroform: 5], then visualized under UV254nm. Chromatograms were photographed digitally to inspect the spots and score. Presence or absence of each flavonoid spot was scored as 1/0; data were entered into a raw data matrix. Multivariate analyses (PCA and Cluster analysis) were performed in NTSYS-pc software package. Cluster analysis was performed using DICE similarity coefficient as the coefficient of choice for qualitative data (Duarte et al., 1999) and dendrogram was constructed using Tamura’s ME sorting method (Tamura et al., 2007).

Purification and identification of flavonoids

Close spots of total flavonoid were put on a horizontal line on a thin layer to separate constituent flavonoids. Skeleton of purified flavonoids were identified using UV-spectrophotometry (200-500 nm). Major substitutions on flavonoid skeleton were identified using NaOAc, H₃BO₃, HCl, AlCl₃ shift-reagents according to Markham (1982).

Extraction and quantitation of glycyrrhizin

Glycyrrhizin was extracted and purified from 2 gr of dried and grinded rhizomes of each accession using the method as described by Shabani et al. (Shabani et al., 2009). Thickened roots (rhizomes) of each accession were dried and grinded to fine powder and 2gr of each material was used for extraction. Ten µl of each extract was placed on UV254F silica-gel and run using a solvent system consisting of [chloroform: 64, methanol: 50, water: 10]. Chromatogram was photographed digitally (Figure 11) and the image was processed using ImageJ software package (Rasband, 2011) to quantify the amount of glycyrrhizin. Pure glycyrrhizin standard (1 mg/ml) was used for calibration. Measures were reported as percent of dry weight.

Results and discussion

Morphology

Twenty-one quantitative morphological characters were measured across twenty-seven accessions of two varieties (G. glabra var. glabra and G. glabra var. glandulifera). Each character was measured up to 6 times to calculate the averages and coefficients of variation (CV). Most variable character was seed area (S-S) and least variable characters were i) Widest portion of legume (Lg-W) and ii) Length of sub-terminal leaflet at left (Le-L3) based on CV- and P-values, respectively.

Figure 2. Exemplar photos in analysis of morphological variation in leaflets, legumes and seeds quantitative characters. Leaflets (1-4), Legumes (5-8), Fuzz on legumes (9-12) and Seeds (13-16).

Principal coordinate analysis of qualitative (0/1) morphological data was implemented using Jaccard coefficient (Jaccard, 1908), scattered accessions on the PCO 3-D plot. Relationships between accessions were evaluated by superimposition of a minimum-length spanning tree (Rohlf, 1975) and rotating the plot (Figure 3A). Cluster analysis of qualitative (0/1) morphological data using ME sorting method (Tamura et al., 2007) of Simple Matching coefficient (SM) similarities, grouped 27 accessions based on geographical location of populations from which the accession was collected (Figure 3B). Resultant groupings were not clear cut; however, meaningful clusters were obtained from both analyses. Two main clusters were observed on the unrooted tree (dendrogram). One consisted of accessions S15 (Kuhmareh), S16 (Pasargad), S07, S08 (Marvdasht), S09 (Sepidan), S11 (Darab), S12 (Darab) and S13 (Firuzabad) and S14 (Kuhmareh). The other cluster consisted of the rest of accessions. Members of this cluster belonged to the variety glabra, except S11 and S15. Intermix of accessions belonging to the two varieties studied were also observed in other clusters on the phenogram.

Close phenetic relationship between accessions S05, S06 (collected from Beyza), S04, S02 (collected from Eghlid) and a sub cluster consisted of accessions S27 (Nourabad), S28 (Yasouj), S34 (Eghlid) and S35 (Shahrekord), all belonging to the variety glandulifera is shown on the phenogram. Cluster analyses of both quantitative and qualitative (0/1) morphological data failed to separate accessions based on taxonomic rank, nor “clearly” grouped accessions based on their population geographic location, although, interpretable clusters were obtained. A North-South partitioning of populations may be deduced. Populations in the first cluster were located in southern part of study area (except for S08 and S09) and populations in the second cluster were located in northern part (except for S29, S30 and S32).

Figure 3. Multivariate analysis of qualitative (0/1) morphological data showing phenetic relationships between 27 accessions. A: Principal Coordinate Analysis using Jaccard coefficient (Jaccard, 1908). A minimum-length spanning tree (Rohlf, 1975) is superimposed on PCO plot, showing the relationships between objects (accessions) . B: Unrooted tree obtained from cluster analysis based on ME sorting method (Tamura et al., 2007) of Simple Matching coefficient similarities calculated by using NTSYS-pc software package (Rohlf, 2000). Open boxes are var. glandulifera and solid boxes are var. glabra (branch lengths proportional to distances).

Results of K-means-Clustering and MDS (Multi-Dimensional Scaling) showed that six clusters of accessions could be defined. Internal similarity tended to be reduced for K>6 or K<6. Each obtained cluster for K=6 consisted of from 2 to 8 accessions. Results were compatible to those of cluster analysis using SM coefficient and ME sorting method (Figure 3). Clusters obtained in K-means-MDS analysis (K=6) were identical to corresponding clusters in the unrooted tree (Figure 3), confirming the robustness of analyses and further confirmed the partitioning of accessions into two major groups consisting of a total of six subgroups.

The first cluster consisted of accessions S02, S04-S06, S17-S19, S21, S25-S30 and S32-S35. The rest of accessions fell into the second cluster. Membership of S29, S30 and S32 in the first cluster and the membership of S08 and S09 in the second cluster were inconsistence with partitioning of accessions along NW-SE. However, the grouping of the rest of accessions was consistence to geographical partitioning.

Figure 4. K-means-MDS clustering. Six clusters were defined based on maximized internal similarities in groups. For membership of each accession in each cluster refer to text and table 3.

Table 3. Membership of accessions in each cluster (K=4 to K=7). K=6 was the best solution based on maximized internal similarities in defined groups.

Flavonoids spot profiles

A total of 95 spots (bands) were scored for 25 accessions (TLC-33 using solvent system [chloroform (25%), acetic acid (25%), butanol (25%), methanol (25%)]). Number of scored bands in TLC-34 (for the 24 accessions analyzed using solvent system [water (20%) ethanol (20%) butanol (10%) chloroform (5%) acetic acid (10%) acetonitryl (10%) metanol (20%) aceton (5%)] was 89 bands, while the number of scored bands in TLC-35 (for 24 accessions analyzed using solvent system [water (50%) ethanol (15%) butanol (20%) acetic Acid (10%) chloroform (5%)] was 135 bands. TLCs were visualized under UV254nm and UV366nm. Data obtained from TLC-35 were adopted and profiles for 2-dimensional TLC for selected accessions was checked to make sure that all the possible bands were separated in TLC-35 (Figure 5).

The solvent system used for TLC-35 was a polar system, lacking non-polar components used in TLC-34. It also had increased proportions for polar solvents compared to the solvent system used for TLC-33. Therefore, it was expected that TLC-35 effectively separate the polar flavonoid constituents.

Figure 5. Flavonoid spots on chromatogram of TLC-35. Numbers beneath chromatogram correspond to accessions. For accession names and the solvent system used, refer to text.

Figure 6 shows phenetic relationships between accessions based on flavonoids spots. Four main clusters (groups1-4) of accessions were identified in the resultant dendrogram. Group 1 and group 2 consisted of a mix of 19 accessions from both varieties. Two subgroups were identified in group 1 while members of group 2 were chained. Members of the second subgroup in group 1 were arranged based on their population distance. Populations of accessions S11 and S12 (both from Darab), S13 and S14 (from Firouzabad and Kuhmareh) were close together with similar soil textures. Accessions S17 and S18 made in small cluster that was also observed in the dendrogram of morphological data. They were geographically located in close distances and were related to different varieties. The fourth group consisted of accessions S01 to 04; all located in Eghlid (NE of Shiraz). Members of this group were also closely related in the dendrogram obtained from morphological data.

Clustering of accessions in groups 1 and 2 were not exactly based on the geographical locations of their populations, however, analyses were robust when different qualitative similarity coefficients were utilized (only results of DICE similarity coefficient are presented).

Accessions studied here could be assigned to two groups; NE populations in Eghlid, Marvdasht and Sepidan and the rest of populations in NW and South of the studied area.

Hayashi et al. (2000) in their phylogenetic study on licorice species using rbcL sequence claimed that It was difficult to distinguish the variation in GL-producing species by rbcL sequence, since they were very similar in all of the Glycyrrhiza species. Our results from morphological and flavonoids spots profiles showed that although a clear-cut grouping was not achieved, variations at the infra-specific level could be elucidated. The variation of flavonoids in leaves of Glycyrrhiza species was reported to be higher than that of rhizomes (Hayashi et al., 2003a; 2003b), which makes them proper markers for further investigation of phenetic relationships in central Zagros region populations. However, flavonoid spot profiles of rhizomes of the studied accessions in this study were also good enough to reveal phenetic relationships among them.

Figure 6. Phenetic relationships between licorice accessions based on flavonoid spots profile. Dendrogram based on data from TLC-35 visualized by UV 254nm and UV 366nm.

Flavonoid Identification

Flavonoid constituents of rhizomes were separated by using TLC chromatography. The solvent system used for separation of flavonoid spots (see material and method) was also used for purification of each flavonoid. Inspection of chromatograms under UV254nm showed that six bands could be extracted and purified for each variety. Chromatograms of flavonoid constituents of each variety (A: variety glandulifera, B: variety glabra) are presented in Figure 7 and Rf values are reported in Table 4. Rf values ranged from 0.454 to 0.735 for mentioned solvent system and TLC type (refer to Material and Method) and 25^oC Temperature.

Figure 7. Chromatograms for separation and identification of major flavonoids in A: var. glandulifera and B: var. glabra. Rf values are reported in Table 4. Twelve bands were recovered from two TLC plates from which seven different flavonoids were identified (Table 7 and Figure 9).

UV spectrophotometry of each constituent in the range of 200-500 nm before and after applying shift reagents (Markham, 1982) showed that three flavonoid skeletons were involved in the separated flavonoids. Variety glandulifera consisted of 3 flavonoids, while variety glabra consisted of four flavonoids. Properties of each skeleton are presented in Table 4 and skeletons are themselves shown in Figure 10.

Figure 8. Exemplar UV spectra of methanolic and shift reagents for band#4 in G. glabra var. glabra. 1: MeOH spectrum, 2: AlCl₃ spectrum, 3: MeOH spectrum, 4: MeOH spectrum, 5: NaOAc spectrum, 6: H₃BO₃ spectrum.

Both varieties consisted of Flavones and Flavonols. Spots number 1, 5, 6 in variety glandulifera and spots number 1, 6 in variety glabra were not flavonoids, but phenolic compounds. Those spots were excluded from further identification. The flavonoid skeletons of spots number 2 and 4 in variety glandulifera were flavones. Spots number 2, 4 and 5 in variety glabra were also flavones. Spot number 3 in variety glandulifera and the same number in variety glabra were flavonols, although they were not the same. The flavonol in variety glandulifera was characterized by an additional hydroxyl group on the carbon number 7 of aromatic ring A.

Table 4. Major flavonoid skeletons identified from licorice rhizomes.

Figure 9. Major flavonoid skeletons identified from licorice. Three flavonoids with skleton 1, 3 were identified from var. glandulifera and four flavonoids with skeleton 1, 2 were identified from var. glabra.

Hayashi et al (2003b) reported Glabridin as the major flavonoid in underground parts of G. glabra collected from Kazakhstan. They also reported Rutin (RT), Isoquercitrin (IQ), Pinocembrin (PN), Licoflavanone (LF) as the four major flavonoids identified from the leaves of same specimens.

In this study, seven more flavonoids with different Rf values were separated in this species; three flavonoids in rhizomes of the var. glandulifera and four flavonoids in rhizomes of the var. glabra.

Glycyrrhizin contents of rhizomes

Glycyrrhizin contents of rhizomes in 24 accessions belonging to the two studied varieties were measured using image processing technique. Results showed that accessions were highly variable (from 0.03 to 0.23 percent of dry weight), so that the most glycyrrhizin rich accession (S15, Kuhmareh, 0.23% DW) had more than seven folds glycyrrhizin than the glycyrrhizin-poor accession (S20, Khafr, 0.03% DW). Both the richest and poorest accessions belonged to variety glandulifera and both were collected from locations of similar soil texture (clay). Quantities of glycyrrhizin in rhizomes of accessions collected from sandy soils (S4, S10, S12, S17, S18, S21 and S28) were also diverse, suggesting that neither soil texture nor variety (taxonomic rank) were main factors affecting the glycyrrhizin amount in rhizomes. Hayashi and co-workers claimed that glycyrrhizin contents of rhizomes of Glycyrrhiza glabra were 10.5% of dry weight (Hayashi et al., 2000). In another report by the same author, glycyrrhizin contents in the underground parts of 3-years-old cultivated
G. uralensis (China type) was 2.08 to 5.12% of dry weight; relatively higher than those of the Kazakhstan type (0.75 - 2.55% of dry weight) and both were much higher than those of Iranian natural populations studied here (Hayashi et al., 2005a).

Figure 10. Quantification of glycyrrhizin contents of rhizomes using image analysis. Control lane contained 4µg of glycyrrhizin standard (pure).

Conclusions

The two varieties of G. glabra studied here, grew together forming mixed populations. Intermixture of closely related taxa in licorice plants was already reported for G. glabra and G. uralensis in a previous study by Hayashi et al. (2003b). Cluster analysis of morphological data neither separated the two varieties nor accessions separated based on their geographical location of populations, although some meaningful groups were obtained. K-means clustering and NMDS analyses confirmed the groupings. Populations from which the accessions were obtained, could be grouped into two Northern and Southern groups in central Zagros region, with a weak support from morphological data.

Hayashi et al. (2005a) divided strains of G. uralensis into two types based on their cp-rbcL sequences (the GA type and AT type). He reported that there were no correlations between the chemotype and the rbcL genotype. Accessions of the two varieties of
G. glabra in our study were not separate in cluster analysis as both produced similar flavonoid spot profiles.

Separation of major polar flavonoids of bulked extractions from each variety resulted in identification of seven flavonoids which had not been reported before. The major non-polar flavonoid of this species without considering the variety of samples was reported as glabridin (Hayashi et al., 2003b).

Our solvent systems efficiently extracted and purified seven flavonoid skeletons which were not reported before for roots and rhizomes of G. glabra varieties. These flavonoids shared the prenyl group on aromatic ring-A (position at carbon 6) and differed in Rf values and other substitution properties.

Glabridin which was a major non-polar constituent of underground parts of licorice was not a major flavonoid constituent in our results.

Both varieties consisted of flavones and flavonols. It is expected that G. glabra var. glandulifera exhibit more pharmacological properties due to the presence of an extra hydroxyl group on carbon number 7 of skeleton number 3 (Figure 10), although this claim must be examined at the variety level.

Closest relatives of G. glabra which was studied by Hayashi et al. (2005b) based on rbcL sequences, were G. inflata and G. uralensis. These taxa represented a clade that produced glycyrrhizin (an oleanane-type triterpene saponin) as the major constituents in rhizomes. Their sister group to G. glabra clade consisted of G. echinata, G. macedonica and G. pallidiflora; which did not produced glycyrrhizin as the major constituent (they produced macedonoside C as the major constituent in rhizomes). Our results provided more detailed information about G. glabra populations in Fars province at the variety level and refined the results of a previous study by Haji-Mehdipour at al. which reported Fars populations among top three populations regarding glycyrrhizin contents of rhizomes (Haji-Mehdipour et al., 2008).

Finally, while the phylogenetic relationship between glycyrrhiza spp. at species level is clear, the infra-specific relationships at population level are still not known. Phenetic relationships between varieties of G. glabra which was studied here claimed that their classification under the putative species was accurate, as they shared similarities in amounts of glycyrrhizin produced in rhizomes, morphological characters and flavonoid constituents. However, the varietal rank of these taxa may be changed to forma according to suggestions made by Brummitt (1990). Results of this study showed intermix of both populations and their characteristics regarding morphology, flavonoids and glycyrrhizin contents, support this suggestion.