Molecular differentiation of Panax notoginseng grown under different conditions by internal extractive electrospray ionization mass spectrometry and multivariate analysis Author Zhang, Xiaoping * & Jiangxi Key Laboratory for Mass Spectrometry and Instrumentation, East China University of Technology, Nanchang, 330013, PR China Author Chen, Ze-Yan Author Qiu, Zi-Dong Author Liu, Mingxing Author Xu, Jiaquan Author Lai, Chang-Jiang-Sheng Author Frankevich, Vladimir Author Chingin, Konstantin text Phytochemistry 2022 113030 2022-02-28 194 1 9 http://dx.doi.org/10.1016/j.phytochem.2021.113030 journal article 53868 10.1016/j.phytochem.2021.113030 530e186e-a4a1-4586-a1cb-f48f003c9d1a 1873-3700 8235549 2.2. IEESI-MS analysis of Panax notoginseng samples grown under different conditions Under the optimal conditions, P. notoginseng samples grown under different conditions were collected for iEESI-MS analysis. The identification information of chemical components from the root of P. notoginseng samples is listed in Table 1 . In total, 35 components, including 23 saponins, were identified from P. notoginseng samples. The chemical assignments of the MS peaks were based on the exact mass measurement of [M+Cl] ¡ and [M ¡ H] ¡ and their corresponding product ions in MSn mode. The comparison with standard references and/or with the online references was used to further confirm the results. The more detailed characterization and chemical information are shown in Table 1 . Fig. 3 shows the mass spectrum fingerprints of different types of P. notoginseng samples analyzed by iEESI-MS. Dominant mass peaks including m/z 279.2339, m/z 377.0857, m/z 589.2709, m/z 835.4602, m/z 967.5034, m/z 981.5189, m/z 1107.5950, m/z 1143.5711, m/z 1193.5948, etc., were found in the mass spectra obtained from all Fig. 1. Schematic illustration of iEESI-MS for direct analysis of Panax notoginseng samples . (a) Analytical procedure of Panax notoginseng analysis by iEESI-MS. Approximately 0.1 mg of Panax notoginseng tissue was loaded into the sample chamber by punching without sample pretreatment, (b) Disposable iEESI device and its components, (c) Photo of iEESI-MS interface. Fig. 2. Chemical structures of eight ginsenosides used in this study as reference standards. G, ginsenoside; NG, notoginsenoside; glc, glucoside; rha, rhamnoside; xyl, xyloside; ara, arabinoside. Table 1 Chemical components in Panax notoginseng samples assigned by high-resolution iEESI-MS n .

No.	Identity	Formula	[M H] (error, ppm)	[M+Cl] (error, ppm)	Fragment ions	Ref
1	Fumaric acid	C4H4O4	115.0041 (3.6)	/	97, 87, 71	Chen et al. (2018)
2	Malic acid	C4H6O5	133.0146 (2.6)	/	115, 87, 71	Chen et al. (2018)
3	Succinic acid	C4H6O4	117.0197 (3.1)	/	99, 89, 73	Chen et al. (2018)
4	Dencichine	C5H7O5N2	175.0363 (1.5)	/	157, 113, 103	Chen et al. (2018)
5	Citric acid	C6H8O7	191.0202 (2.5)	/	173, 129, 111	Chen et al. (2018)
6	Fructose	C6H12O6	179.0566 (2.7)	215.0330 (1.0)	179, 161	Chen et al. (2018)
7	Palmitoleic acid	C16H30O2	253.2179 (2.3)	/	235, 209, 193	Kim et al. (2018)
8	Palmitic acid	C16H32O2	255.2338 (3.3)	291.2096 (0.1)	237, 211, 184	Kim et al. (2018)
9	Sodium dodecyl sulfate	C12H26O4S	265.1476 (1.1)	/	247, 229, 191, 97	/
10	Linoleic acid	C18H32O2	279.2339 (3.4)	/	265, 243, 235	Chen et al. (2018)
11	Oleic acid	C18H34O2	281.2488 (0.7)	/	263, 245, 237	Kim et al. (2018)
12	Sucrose	C12H22O11	341.1092 (0.8)	377.0857 (0.2)	341, 215,179,161	Zhu et al. (2021)
13	Raffinose	C18H32O16	503.1624 (1.3)	539.1388 (0.7)	503, 341, 323, 179, 161	Chen et al. (2018)
14	Rb1 a	C54H92O23	/	589.2709 (0.2)/	1143, 571, 553, 377	/
15	Rh2	C36H62O8	621.4374 (0.3)	/	475	Li et al. (2020)
16	Rh1	C36H62O9	637.4314 (1.1)	673.4090 (0.3)	475	Zhang et al. (2012)
17	2Sucrose	C12H22O11	683.2262 (2.0)	719.2023 (0.7)	377, 341	Zhu et al. (2021)
18	Rg2	C42H72O13	783.4891 (1.1)	819.4667 (0.0)	637, 619, 475	Wu et al. (2015)
19	Rg1/Rf	C42H72O14	799.4840 (1.2)	835.4602 (1.7)	799, 637, 475	Chen et al. (2018)
20	Malonyl-Rg1/malonyl-Rf	C45H74O17	885.4847 (0.7)	921.4611 (1.0)	841, 799, 781, 679, 619, 475	Li et al. (2010)
21	NG-R1	C47H80O18	931.5271 (0.1)	967.5034 (0.5)	799, 637, 619, 475	Chen et al. (2018)
22	Ro	C48H76O19	955.4899 (0.9)	/	793, 631, 613, 455	Sun et al. (2016)
23	Rd and Re	C48H82O18	945.5419 (1.0)	981.5189 (0.6)	799, 783, 765, 637, 621, 475, 459	Sun et al. (2016)
24	NG-R3	C48H82O19	961.5393 (1.6)	997.5142 (0.2)	799, 637, 475	Liu et al. (2015)
25	Malonyl-Rd and malonyl-	C51H84O21	1031.5423 (0.9)	1067.5214 (1.4)	945, 927, 621, 603, 475, 459	Li et al. (2010)
Re
26	3 Sucrose	C12H22O11	1025.3407 (0.6)	1061.3182 (0.1)	719, 377, 341	Zhu et al. (2021)
27	Rc	C53H90O22	1077.5850 (0.1)	1113.5623 (0.5)	945, 783, 621, 459	Li et al. (2010)
28	Rb1	C54H92O23	1107.5950 (0.6)	1143.5711 (1.1)	945, 783, 621, 459	Chen et al. (2018)
29	Malonyl-Rc	C56H92O25	1163.5854 (0.1)	1199.5662 (3.4)	1119, 1077, 1059, 945, 783	Sun et al. (2016)
30	Malonyl-Rb1	C57H94O26	1193.5948 (1.1)	1229.5706 (1.7)	1149, 1107, 1089, 945, 783, 621, 459	Fuzzati et al. (1999)
31	Ra1	C58H98O26	1209.6311 (3.1)	1245.6072 (2.5)	1077, 1047, 945, 915, 783, 621, 459	Sun et al. (2016)
32	Ra3	C59H100O27	1239.6371 (0.7)	1275.6165 (1.5)	1107, 945, 783, 621, 459	Wu et al. (2015)
33	Malonyl-Ra1	C61H100O29	1295.6332 (4.2)	/	1251, 1209, 1191, 1119, 1077, 1059, 945	Sun et al. (2016)
34	Malonyl-Ra3	C62H102O30	1325.6372 (0.8)	/	1281, 1239, 1221, 1149, 1107, 1089, 945, 621, 459	Wu et al. (2015)
35	4 Sucrose	C12H22O11	1367.4591 (1.1)	/	1061, 719	Zhu et al. (2021)

NG, notoginsenoside. P. notoginseng samples grown under different conditions. Fig. 3. Mass spectra of different types of Panax notoginseng samples analyzed by iEESI-MS. (a) Mass spectrum of Panax notoginseng from Kunming, (b) Mass spectrum of Panax notoginseng from Qujing, (c) Mass spectrum of Panax notoginseng from Hongjiaozhou, (d) Mass spectrum of Panax notoginseng from Wenshan (1 year), (e) Mass spectrum of Panax notoginseng from Wenshan (2 year), (f) Mass spectrum of Panax notoginseng from Wenshan (3 year), (g) Mass spectrum of Panax notoginseng from Hongjiaozhou (black soil), (h) Mass spectrum of Panax notoginseng from Hongjiaozhou (white soil), (i) Mass spectrum of Panax notoginseng from Hongjiaozhou (red soil). (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.) Ginsenosides are important quality indicators of P. notoginseng . In the negative ion mode, the predominant ions pairs [M+Cl] ¡ and [M ¡ H] ¡ with 36 Da difference were observed for most ginsenosides. The MSn spectra of [M+Cl] ¡ and [M ¡ H] ¡ exhibited a fragmentation pattern corresponding to the successive loss of the glycosidic units ( Table 1 ). A neutral loss of 132, 146, and 162 Da could indicate the elimination of pentose, rhamnose, and glucose residues, respectively ( Cao et al., 2019 ). [Aglycone ¡ H] ¡ ions at m/z 475, m/z 459 and m/z 455 corresponding to the (20 S)-protopanaxatriol aglycon moiety (PPT type ), the (20 S)-protopanaxadiol aglycon moiety (PPD type ), and the oleanolic acid aglycon moiety (OA type ), respectively ( Cao et al., 2019 ). In addition, the MSn spectra of malonyl-ginsenosides showed intense signals due to the loss of malonyl unit (CO 2 + CH 2 CO, 86 Da) together with peaks due to successive losses of glycosidic units ( Lai et al., 2015b ). Usingthisaboveinterpretation ofMS nspectraexperiments withstandard ginsenosides and literature search, a total of 23 saponins were revealed in P. notoginseng by iEESI-MSn analysis. Rb1 and malonyl-Rb1 were classified under the same category because the only difference observed between them was a malonyl group. Peaks at m/z 1143.5711 ([M+Cl] ¡ ) and m/z 1107.5950 ([M ¡ H] ¡ )were attributed to Cl adduct of Rb1 and deprotonated Rb1, respectively.In the MS 3 spectrum of m/z 1143 → 1107→, the fragment ions at m/z 945,783,621,and459,were observed ( Fig.4a ),corresponding to [M ¡ H ¡ glc] ¡ , [M ¡ H ¡ 2glc] ¡ , [M ¡ H ¡ 3glc] ¡ , [M ¡ H ¡ 4glc] ¡ , respectively. Therefore, m/z 1143.5711 was assigned to a PPD-type ginsenoside (Rb1). The peak at m/z 1193.5948 ([M ¡ H] ¡ ) with calculated molecular formula of C 57 H 94 O 26 was 86 m /z units higher than deprotonated Rb1 ( m/z 1107.5950).In the MS 2 spectrum of m/z 1193 the fragment ion at m/z 1149 was mainly observed ( Fig. S2a ). In the MS 3 spectrum of m/z 1193 → 1149 and the fragment ions at m/z 1107, 945, 783, 621 and 459, were observed ( Fig.4b ).The fragment ions at m/z 1149 and m/z 1107 indicated the ions of [M ¡ H–CO 2 ] ¡ and [M ¡ H–C 3 H 2 O 3 ] ¡ , which suggested the presence of a malonyl substituent. Other fragment ions at m/z 945, 783, 621 and 459 represented the ions of [M ¡ H–C 3 H 2 O 3 ¡ glc] ¡ , [M ¡ H–C 3 H 2 O 3 ¡ 2glc] ¡ , [M ¡ H–C 3 H 2 O 3 ¡ 3glc] ¡ , [M ¡ H–C 3 H 2 O 3 ¡ 4glc] ¡ , respectively.Therefore, m/z 1193.5948 was assigned to a PPD-type ginsenoside with a malonyl substituent (tentatively malonyl-Rb1). The peak at m/z 885.4847 ([ M ¡ H ] ¡ ) with calculated molecular formula of C 45 H 74 O 17 was 86 m /z units higher than deprotonated Rg1/Rf( m/ z 799.4840, C 42 H 72 O 14 ). In the MS 2 spectrum of m/z 885 the fragment ion at m/z 841 was mainly observed ( Fig. S2b ). In the MS 3 spectrum of m/z 885 → 841 and the fragment ions at m/z 799, 781, 637, 619 and 475, were observed ( Fig. 4c ). These signals represented the ions of [ M ¡ H – CO 2 ] ¡ , [ M ¡ H – C 3 H 2 O 3 ] ¡ , [ M ¡ H – C 3 H 2 O 3 – H 2 O ] ¡ , [ M ¡ H – C 3 H 2 O 3 ¡ glc] ¡ , [ M ¡ H – C 3 H 2 O 3 ¡ glc ¡ H 2 O ] ¡ and [ M ¡ H – C 3 H 2 O 3 ¡ 2glc] ¡ , respectively. Therefore, m/z 885.4847 was assigned to a PPT-type ginsenoside with a malonyl substituent (tentatively malonyl-Rg1/Rf). The peak at m/z 1031.5423 ([ M ¡ H ] ¡ ) with calculated molecular formula of C 51 H 84 O 21 was also 86 m /z units higher than deprotonated Rd/Re ( m/z 945.5419, C 48 H 82 O 18 ). In the MS 2 spectrum of m/z 1031 the fragment ion at m/z 987 was mainly observed ( Fig.S2c ). In the MS 3 spectrum of m/z 1031 → 987→ and the fragment ions at m/z 945, 927, 783, 621, 603, 475 and 459, were observed ( Fig. 4d ). These signals represented the ions of [ M ¡ H – CO 2 ] ¡ , [ M ¡ H – C 3 H 2 O 3 ] ¡ , [ M ¡ H – C 3 H 2 O 3 – H 2 O ] ¡ , [ M ¡ H – C 3 H 2 O 3 ¡ glc] ¡ , [ M ¡ H – C 3 H 2 O 3 ¡ 2glc] ¡ , [ M ¡ H – C 3 H 2 O 3 ¡ 2glc ¡ H 2 O ] ¡ , [ M ¡ H – C 3 H 2 O 3 ¡ 2glc ¡ rah] ¡ , and [ M ¡ H – C 3 H 2 O 3 ¡ 3glc] ¡ , respectively. Note that both [aglycone ¡ H ] ¡ ions at m/z 475 and m/z 459 were observed in the fragmentation of m/z 1031 → 987. Therefore, m/z 1031.5423 was assigned to a mixture of PPD-type and PPT-type ginsenoside with a malonyl substituent (tentatively malonyl-Rd and malonyl-Re). Peaks at m/z 1245.6072 ([ M +Cl] ¡ ) and m/z 1209.6311 ([ M ¡ H ] ¡ ) were attributed to Cl adduct of Ra1 and deprotonated Ra1, respectively. In the MS 3 spectrum of m/z 1245 → 1209→, the fragment ions at m/z 1077, 945, 783, 621 and 459, were observed ( Fig. 4e ), which represented the ions of [ M ¡ HCl ¡ xyl] ¡ , [ M ¡ HCl ¡ 2xyl] ¡ , [ M ¡ HCl ¡ 2xyl ¡ glc] ¡ , [ M ¡ HCl ¡ 2xyl 2glc] ¡ , and [ M ¡ HCl ¡ 2xyl 3glc] ¡ , respectively. Therefore, m/z 1209.6311 was assigned to a PD-type ginsenoside (tentatively Ra1).Peaks at m/z 967.5034 ([ M +Cl] ¡ ) and m/z 931.5271 ([ M ¡ H ] ¡ ) were attributed to Cl adduct of NG-R1 and deprotonated NG-R1, respectively ( Fig. S2d ). In the MS 3 spectrum of m/z 967 → 931→, the fragment ions at m/z 799,769, 637 and475, were observed( Fig.4f ), which represented the ions of [ M ¡ HCl ¡ xyl] ¡ , [ M ¡ HCl ¡ glc] ¡ , [ M ¡ HCl ¡ xyl ¡ glc] ¡ and [ M ¡ HCl ¡ xyl ¡ 2glc] ¡ , respectively. Therefore, m/z 931.5271 was assigned to a PT-type ginsenoside (tentatively NG-R1). Other types of ginsenosides were identified according to the similar pattern. Their MS 2 and MS 3 spectra are displayed in Figs. S3 and S 4 . Fig. 4. Tandem MS analysis of characteristic ions in Panax notoginseng samples . (a) MS 3 spectrum of m/z 1143 → 1107→, (b) MS 3 spectrum of m/z 1193 → 1149→, (c) MS 3 spectrum of m/z 885 → 841, (d) MS 3 spectrum of m/z 1031 → 987, (e) MS 3 spectrum of m/z 1245 → 1209, (f) MS 3 spectrum of m/z 968 → 931. 2.3. Differences of Panax notoginseng In general, P. notoginseng samples from four different production areas were broadly similar in chemical composition, being different only in relative content. Samples from Honghezhou held a lowest relative content of NG-R1. The separate and total relative contents of Rg1, Rb1, NG-R1, malonyl-Rb1were higher in Qujing samples than others. The fingerprints of Wenshan P. notoginseng of different growth years showed that the 1-year samples had the highest relative content of Rb1 ( Fig. 3 ). The relative content of Rg 1 in the 2-year sample increased significantly, and both NG-R1 and Rb1 showed a significant decrease, while the 3-year sample had more Rg1, and the content of NG-R1 and Rb1 was lower. Thus, Wenshan P. notoginseng with different growth years showed various fingerprints and chemical composition. Hongjiaozhou P. notoginseng from various growth soils ( Fig. 3 ) showed relative similarity in chemical fingerprints compared with those of different origins and growth years, nevertheless diversity also could be recognized on some compounds. 2.4. Comparison with other MS-based methods Table 2 compares several MS-based methods in the analysis of components in P. notoginseng samples. GC-MS and LC-MS are widely used for the analysis of components in ginseng, but these methods require multiple-step sample pretreatment (powder, extract, centrifuge and isolation), large amount of solvent, large amount of sample, and long analysis time ( Lai et al., 2015a ; Xie et al., 2007 ). Also, the multiple-step sample pretreatment may cause the loss of some key compounds. In advantage to GC/LC-MS, ambient MS methods facilitate sample pretreatment and increase the speed of analysis. To date, ambient MS methods such as desorption atmospheric pressure chemical ionization mass spectrometry (DAPCI-MS) and direct analysis in real time mass spectrometry (DART-MS) have been adopted to the direct detection of analytes on ginseng surface and ginseng extract, respectively ( Wang et al., 2014 ; Yue et al., 2013 ). Unfortunately, DAPCI-MS is mainly sensitive to the chemicals on the surface rather than inside of the ginseng samples. Time-consuming pretreatments such as powder, extract, centrifuge, etc. were demanded to make the ginseng extract ready for DART-MS analysis. In our iEESI-MS analysis, the P. notoginseng tissue sample (e.g., 0.5 mg , sized as 0.5 mm × 0.5 mm × 1 mm ) was directly loaded in the sampler with a single punch, requiring no other sample pretreatment. Note that less amount of solvent, less amount of P. notoginseng samples, and shorter analysis time were consumed for iEESI-MS analysis compared to other MS-based methods. Promisingly, iEESI-MS presents an alternative choice for the analysis of components in P. notoginseng tissue samples, with significantly improved analysis speed. 2.5. Multivariate statistical analysis To further characterize and visualize the chemical differences and correlations between different batches of P. notoginseng samples, the global iEESI-MS fingerprints data of 90 P. notoginseng samples was acquired, and the resulting three-dimensional data set comprising peak number, sample name and ion intensity was subjected to multivariate statistical analysis using SIMCA-p software. Principal component analysis (PCA) and orthogonal partial least squares discriminant analysis (OPLS-DA) mode containing 90 P. notoginseng samples with different growth conditions were established to screen differential metabolic chemical markers. Fig. 5a and b shows the PCA score plot obtained using mass spectra from 90 individual P. notoginseng samples. A total of 90 data points of P. notoginseng samples were clustered regularly in the PCA plots. Samples “notoginseng-qujing” and “notoginseng-kunming” occupied adjacent area in Fig. 5a , indicating that these two samples from different origin display similar chemical components. The first principal component (PC1) accounted for 65.8% of variance whereas PC2 and PC3 on the other hand accounting for 11.5% and 7.8%, respectively ( Fig. 5b ). It can be seen from the PCA loading plots that the peaks at m/z 377.0857, m/z 835.4602, m/z 265.1476, m/z 1143.5711, m/z 967.5034, m/z 279.2339, m/z 1193.5948, m/z 885.4847, m/z 341.1092, m/z 719.2023, m/z 133.0146 and m/z 981.5189 are the main contributors to the differentiation. Fig. 5c and d shows the OPLS-DA score scatter plot of all P. notoginseng samples using mass spectra from 90 individual P. notoginseng samples, in which the samples were separated from each other before analysis. It can be found that P. notoginseng samples with different origins, different soils, and different growth years displayed obvious difference, indicating that the different growing conditions had influence on the chemical composition of P. notoginseng samples. The results show that the parameters R 2 Y and Q 2 of the newly established OPLS-DA model were 0.939 and 0.875, respectively, indicating that the model showed good predictive ability and can be used to distinguish different types of P. notoginseng samples. Variable importance in the projection (VIP) can be used to measure the impact and interpret ability of each metabolite’ s expression pattern on the classification discrimination of each group. To identify the differentiated ginsenosides that contributed the most to the group separations, the feature selections were performed by a S plot derived from the constructed OPLS-DA. The VIP plots of the differential metabolic chemical markers of P. notoginseng samples are shown in Fig. 5d . Chemical identities of these markers are listed in Table 1. A total of 14 MS signals from P. notoginseng samples were screened by taking the VIP score > 1.0 as evaluation standard. Compared with references, these MS signals at m/z 377.0857, m/z 835.4602, m/z 265.1476, m/z 215.0330, m/z 1143.5711, m/z 967.5034, m/z 279.2339, m/z 1193.5948, m/z 885.4847, m/z 255.2338, m/z 341.1092, m/z 719.2023, m/z 981.5189 and m/z 133.0146 were as assigned to [sucrose +Cl] ¡ , [Rg1/Rf +Cl] ¡ , dodecyl sulfate anion, [fructose + Cl] ¡ , [Rb1+Cl] ¡ , [Noto-R1+Cl] ¡ , [linoleic acid ¡ H ] ¡ , [malonyl-Rb1 ¡ H ] ¡ , [malonyl-Rg1/Rf ¡ H ] ¡ , [palmitic acid ¡ H ] ¡ , [sucrose ¡ H ] ¡ , [2 sucrose +Cl] ¡ , [Rd/Re +Cl] ¡ , and [malic acid ¡ H ] ¡ , respectively. Note that the main contributors to the differentiation of OPLS-DA analysis are consistent with the results of PCA analysis described above. Thus, these results suggested that the difference in the content of sucrose, fructose, Rg1, Rf, Rb1, Noto-R1, malonyl-Rb1, malonyl-Rg1, malonyl-Rf, Rd, Re, linoleic acid, palmitic acid and malic acid in P. notoginseng samples can be used as key characteristic indicators to discriminate origin, commercial specifications, and cultivation conditions of P. notoginseng samples and can be considered as the metabolic markers of this Genus. Therefore, we conclude that the growth environment factors cause significant differences in the chemical quality of P. notoginseng . Table 2 3. Conclusions Several MS-based methods in the analysis of components in Panax notoginseng samples . In this work, iEESI-MS combined with multivariate techniques were used for the first time to discriminate and classify P. notoginseng samples under different growth conditions (e.g., place of origin, soil quality, growth season). Additionally, the chemical markers that could be used for the characterization and screening of P. notoginseng samples under different growth conditions were explored. A total of 35 compounds, including organic acids, saponins, sugars, etc., were revealed in P. notoginseng samples based on the accurate m/z values, tandem MS information, experiments with standard compounds and earlier reports. Nine batches of P. notoginseng samples grown under different conditions showed a clear separation in multivariate statistical analysis, and the significant differential components were revealed. In future, the developed iEESI-MS method could be broader applied to screen and characterize other plant-derived medicines and help understand the differences in pharmacological activities at the molecular level.