Comparative metabolic fingerprinting of Gentiana rhodantha from different geographical origins using LC-UV-MS/MS and multivariate statistical analysis

Backgrounds Gentiana rhodantha, a rich source of iridoids and polyphenols, is a traditional ethnomedicine widely used in China. Metabolic fingerprinting based on a LC-UV-MS/MS method was applied to explore the chemical markers for discrimination of G. rhodantha from different geographical origins. Results Targeted compounds were separated on a Shim-pack XR-ODS III (150 × 2.0 mm, 2.2 μm), with a mobile phase consisted of acetonitrile and 0.1% formic acid in water, under gradient elution. In quantitative analysis, all of the calibration curves showed good linear regression (R2 < less than 0.9991) within the tested ranges, and accuracy ranged from 97.8% to 104.2% and the %RSD of precision (less than 3%) were all within the required limits. The most abundant mangiferin (82.21 mg/g) found in sample from Zunyi, Guizhou province. Furthermore, 64 samples according to their geographical origins, could be classified by partial least-squares discriminate analysis (PLS-DA) and nine compounds including two new compounds identified by mass spectrometry could be regarded as characteristic compounds for discriminating samples from different geographical origins. Conclusions The developed method appears to be a useful tool for analysis of G. rhodantha, which could provide potential indicators for differentiation of different geographical origins. Electronic supplementary material The online version of this article (doi:10.1186/s12858-015-0038-5) contains supplementary material, which is available to authorized users.


Backgrounds
Metabolic fingerprinting based on modern separation science was an effective tool for analyzing bio-samples with complex chemical information [1,2]. Currently, it showed strong potential for analysis of food and medicine as well as monitoring metabolic variation under different surrounding conditions when combined with multivariate statistical analysis [3][4][5][6][7]. Herbal medicines utilized as folk medicines since the ancient times are increasingly rising worldwide attention owing to its effect for maintaining health. The therapeutic effects of herbal medicines, to a large extend, were derived from synergistic effect of its metabolites. However, metabolites in herbal medicines vary with the geographical origins, which may have influence on the quality and effectiveness of herbal medicines [8][9][10][11][12][13][14][15].
Several chromatographic methods including TLC, HPLC and UPLC-MS were developed for determination of mangiferin in G. rhodantha [15,22,23]. To the best of our knowledge no report is available on discrimination of G. rhodantha with different source by metabolic fingerprinting combined with multivariate analysis.
In present study, metabolic fingerprinting based on LC-UV-MS/MS method was developed for simultaneous determination of five compounds and evaluating metabolic similarities and differences in 64 batches of G. rhodantha from different geographical origins. Similarly analysis was applied for measuring correlations between samples from different sites. Then, principal component analysis (PCA) was used to find the resemblance and pre-classify these samples. According to PCA results, partial least squares discriminant analysis (PLS-DA) was designed for discrimination of selected samples and determined the characteristic compounds, which could provide potential indicators of G. rhodantha origins.

Optimization of UFLC-UV-MS/MS analysis conditions
To achieve a desirable resolution and separation in both chromatographic fingerprint and full scanning mode, several analytical parameters including mobile phases and elution mode, flow rate, column temperature, detection wavelength critical parameter and dwell time were optimized.
Several types of solvent systems, including methanolwater and acetonitrile-water in various elution modes were tested to reach an optimum separation on the Shim-pack XR-ODS III (150 × 2.0 mm, 2.2 μm). A desirable separation performance was obtained in acetonitrile-water system. Then, 0.1% formic acid was added to the mobile phase to enhance resolution and eliminate peak tailing of the metabolites in fingerprints while enhancing the intensity of adducted molecular ions [M + HCOO] − and protonated molecular ions [M + H] + in mass spectrometer. Moreover, flow rate and column temperature was set at 0.35 min/ml and 40°C respectively, which could also improve separation efficiency. The UV wavelength was set at 242 nm, because the entire standard compounds had adequate absorptions and fingerprints also exhibited satisfactory performance ( Figure 1). The optimization of mass conditions was performed in both positive and negative ionization mode. The mass range in full scanning mode was set as follow: 2-17 min at m/z 110-900 Da, 17-22 min at m/z 900-1800 Da. The critical parameter (CE) in product ions scanning mode was optimized for improving the signal of product ions according to different precursor ion. The MRM settings were auto optimized.

Method validation and quantification
The developed method for determination of the five compounds was validated in terms of linearity, repeatability, stability, precision and accuracy. These calibration curves revealed good linear relationship and the correlation coefficients for the five compounds were greater than 0.9991. The LOD and LOQ were less than 0.007 and 0.035 μg/ml, respectively ( Table 1). The validation was performed on QC sample based on retention times (Rt) and peak areas (Pa). The RSD% of repeatability and stability for Pa were less than 1.67% and 2.61%, respectively. The intra-along with inter-day %RSD of Rt were less than 1.21% and %RSD of peak areas were less than 4% ( Table 2). The Recovery for all analytes in the range of low to high concentration was in the range of 97.8-104.2% (Table 3).
The five compounds were identified by comparison of their retention times and precursor/product-ion pairs obtained by MRM acquisition mode. The established method was applied for simultaneous determination of the five compounds in 64 samples of G. rhodantha. Among them, swertiamarin and gentiopicroside with low concentration was determined by MRM acquisition mode. Each sample was analyzed in triplicate to determine a mean content (mg/g) and standard deviation (SD). As shown in Figure 1 Chromatograms at 242 nm for QC sample. Table 4, there were significant differences in compounds contents of G. rhodantha from different geographical origins. Mangiferin was found to be predominant in G. rhodantha, ranging from 45.19 to 82.21 mg/g. The yield of mangiferin followed in the order: GZ > YM > GXL > GAS > YK > GX > YD > GAL > YW > YL > GK. In addition, the highest yield of sweroside (6.10 ± 1.50 mg/g) was found in Guangxi province.

Similarly analysis
The similarly analysis of chromatographic fingerprint was performed by Chinese Pharmacopoeia Committee (Version 2004 A) and SPSS 20.0. The common peaks confirmed by the RSD% of all peaks' relative retention time (<1%) were matched automatically. All fingerprints were classified into 11 groups according to their origins. Then, standard fingerprint of each group was obtained by comparing chromatograms of samples from the same origin. The correlation coefficients of fingerprint of samples from the same and different geographical origins are listed in Additional file 1: Table S1. The results showed correlation coefficients of fingerprint of samples from the same origin were greater than 95%. As shown in Additional file 2: Figure S1, samples from different origin had the same peak numbers, but peak areas and correlation coefficients were visually different in these 11 groups (Additional file 1: Table S1). For example, correlation coefficients (0.891) implied samples from Dali and Lijiang were more similar when compared with other samples. Moreover, peak area (No.20) in their fingerprints was greater than others (Additional file 2: Figure S1). These results indicated that geographical origins might have influence in metabolites accumulation.

PCA and PLS-DA
PCA, an unsupervised classification approach, was applied for dimensions of complex data obtained from chromatographic fingerprint and discrimination of different sample. The peak areas without any pre-processing obtained from chromatographic fingerprints were subjected to PCA analysis. In Additional file 3: Figure 2S, 64 samples from 11 sites could not achieve a good classification performance. However, these samples from geographical close cities in Yunnan and Guizhou province could be clustered together, which would provide a basis for classification of these samples for further analysis.
According to PCA results, PLS-DA, a supervised classification approach, was designed for further discrimination of these samples. As shown in PLS-DA scores plot (Figure 2a), 64 samples from 6 main geographical origins (northwest, east and southeast of Yunnan, northeast and southwest of Guizhong as well as Guangxi) were well separated. Among these origins, Linxian, Guangxi province is located between southeast Yunan and Guizhou. The relative location of samples from the three regions in PLS-DA scores plot conformed to the actual geographic location. Moreover, samples from different origins in Yunnan were significant different, which might owe this  difference to the complicated climatic and geographical conditions in Yunnan.
In loading-bi plot (Figure 2b), the important principal components and chemical compounds in separating the samples could be exhibited. PC1 have potent impact on discrimination of samples in Yunnan from others. Moreover, peak 34 and 27 might have more influence on discrimination of samples from southwest Guizhong; Peak 17 and 15 play important roles for the discriminating GXL samples from other places; Peak (2, 18 and 22) and Peak (20, 24 and 30) likely to have more contribution for classification of samples from northwest Yunnan. Interestingly, these results could correspond to the results of similar analysis and characteristic in their corresponding fingerprints.

Marker metabolites screening and identification
Variable importance plot (VIP) based on PLS-DA result was used for screening the characteristic compounds according to contribution values for discrimination of different samples. In this study, peak 32, 27, 4, 30, 31, 9, 10, 29 and 20 (descending order) were likely to be considered as characteristic compounds of G. rhodantha in accordance with their VIP values (>1.2). These characteristic compounds were identified using LC-MS/MS operated at multiple scanning modes and literature data on the fragment ions patterns of these compounds. Marker metabolites were tentatively identified ( Table 5) and their structures are displayed in Figure 3.
First of all, Peak 7, 12, 14 and 15 were identified as loganic acid, swertiamarin, mangiferin and sweroside through comparing by retention times and mass data of reference compounds.
According to the mass data analysis of peak 27-32, the major product ions and fragmentation pattern of the revealed their derivatives with the same basic skeleton.  [17,18]. In negative mode, their fragment ions patterns were highly similar. As shown in Table 6 and Figure 4a     at m/z at 223 and 209. According to matching the fragment ions pattern from published work [19], peak 9 and 10 were tentatively assigned as secologanoside and secoxyloganin, respectively. The fragmentation pathways of secologanoside and secoxyloganin were exhibited in Figure 4b and c.

Conclusion
In present study, a simple and reliable method was developed for simultaneous determination of the main compounds in G. rhodantha. Metabolic fingerprint based on LC-UV-MS/MS was design for elucidating the change in samples from different geographical origins. The multivariate analysis demonstrated that metabolites of G. rhodantha were clearly dependent on geographical. Furthermore, five characteristic compounds were considered as the most important variables for distinguishing different samples according to their origins. Metabolic fingerprint based on LC-UV-MS/MS could be a useful tool for simple discrimination of G. rhodantha from different geographical origins when combined with multivariate analysis. The results could be consistent with the practical geological setting, which may reflect potential relationship between geographical origins and accumulation of characteristic compounds for further study.

Materials and reagents
The aerial parts of G. rhodantha were collected from southwest China including 11 sites ( Figure 5) in January, 2014. In total of 64 samples were authenticated by Professor Hang Jin (Institute of Medicinal Plants, Yunnan Academy of Agricultural Sciences). Fresh plants were dried at 60°C and then ground into fine powder around 60 meshes. The information of samples is listed in Table 6. No specific permits were required for the described field studies, as no endangered or protected species were sampled, and the localities where the samples came from are not protected in any way.
The LC grade solvents (acetonitrile and formic acid) purchased from Fisher Scientific and Dikmapure (USA),    Multiple reaction monitoring (MRM) acquisition mode and products ion scanning was used for quantification purposes of chemical constituents with low content and qualitative analysis, respectively.