Identification of constituents in the samples by HPLC and Ultra-high-resolution MS
Vanquish horizon HPLC system combined with Q Exactive Plus ultra-high resolution mass spectrometry (Thermo Fisher Scientific, MA, USA) and compound discovery structure identification software was used to detect the main chemical components in DXT extraction solution samples. The precursor ion selectivity of the conjugate hyperboloid quadrupole combined with the high resolution accurate mass detection technology based on orbitrap had the advantages of ultra-high resolution (R = 140, 000 @ m/z 200, could quickly switch between positive and negative at the same time to obtain comprehensive sample information), ultra-high quality accuracy and stability, high sensitivity (ag-fg level) and ultra-wide linear range (104—105), which could greatly increase the credibility and efficiency of identification.
The chromatographic conditions were as follows: Column, Hypersil gold aQ (2.1 * 100 mm, 1.9 μm); Mobile phase, 0.04% acetic acid water (A) - 0.04% acetonitrile acetate (B); Flow rate, 0.35 mL/min; The injection volume, 2 μL. The elution gradient was shown in Additional file 1.
In the Q Exactive Plus MS system, an electrospray ion source was used for detection in the positive and negative ion scanning modes, and the conditions are as follows: Spray voltage: 3.8 kV(+) / 3.2 kV(-); Vaporiser temp: 350 ℃; Sheath gas: 40 arb; AUX gas: 10 arb; Capillary Temp: 320°C;s-Lens: 60; General Method: Fullms-ddms2;Scan range: 120—1500;Resolution (MS1): 70, 000; MS/MS resolution: 17, 500; Stepped NCE: 20, 40, 60.
The total ion chromatograms in negative ion and positive ion modes were shown in Fig.1 and 2, and the mass spectrum details of the measured component obtained from the detector were shown in Table 1. A total of twenty-one components were tested and identified. Based on preliminary experiments, nine components with high activity or high content were selected for subsequent content determination. The nine ingredients were protocatechuic acid, hydroxytyrosol, caffeic acid, epicatechin, salidroside, chlorogenic acid, desrhamnosyl isoacteoside, rutin, and liriodendrin. And the chemical formulas were shown in Additional file 2.
Table 1 Mass spectrum information of tested components
Detection mode
|
Peak
|
Retention time / tR (min)
|
Excimer ion
|
Component attribution
|
Negative mode
[M-H]-(M/Z)
|
a
|
3.41
|
167.03
|
Vanillic acid
|
b
|
3.52
|
153.01
|
Protocatechuic acid
|
c
|
3.96
|
153.05
|
Hydroxytyrosol
|
d
|
4.38
|
299.11
|
Salidroside
|
e
|
4.63
|
353.08
|
Chlorogenic acid
|
f
|
4.81
|
137.02
|
p-hydroxybenzoic acid
|
g
|
5.36
|
179.03
|
Caffeic acid
|
h
|
5.48
|
289.07
|
Epicatechin
|
i
|
6.67
|
477.14
|
Desrhamnosyl isoacteoside
|
j
|
8.00
|
193.05
|
Ferulic acid
|
k
|
8.29
|
312.12
|
Feruloyltyramine
|
l
|
12.20
|
487.34
|
Madasiatic acid
|
m
|
5.17
|
577.16
|
Procyanidin B2
|
n
|
5.86
|
741.71
|
Liriodendrin
|
o
|
8.67
|
283.06
|
Physcion
|
p
|
8.73
|
609.14
|
Rutin
|
q
|
9.55
|
685.38
|
Rosamultin
|
r
|
17.79
|
17.79
|
Oleanolic acid
|
Positive mode
[M+H]+(M/Z)
|
a
|
2.03
|
167.07
|
Apocynin
|
b
|
5.48
|
291.07
|
Epicatechin
|
c
|
5.68
|
271.05
|
Emodin
|
d
|
6.61
|
303.04
|
Quercetin
|
Method Validation
The specificity, calibration curves, limits of detection (LOD), quantitation (LOQ), precision, stability, repeatability, and recovery were investigated.
The specificity was investigated by taking the mixed reference solution and 50% methanol-water blank solvent, and performing sample analysis according to the optimized LC/MS conditions. The calibration curve for each constituent was established by plotting the peak area (y) versus the concentration (x) of each analyte. The LOD and LOQ for nine analytes were estimated at S/N of 3 and 10, respectively, by injecting a series of dilute solutions with known concentration. The intra-day and inter-day precision for each analyte was investigated by determining the nine analytes in six replicates during a single day and three consecutive days. Variations of the peak area were taken as the measures of intra-day and inter-day analysis precision to calculate the RSD. Stability was investigated by analyzing the solution at 0, 2, 4, 8, 12, 24 h, respectively. To assess the repeatability, six solutions prepared from sample were analyzed. Variability was expressed in RSD (%). The recoveries of the analytes were determined by using the method of standard addition within the same day. Three different concentrations of mixed standard solutions (50%, 100%, and 150% of the known amount in sample) were spiked into sample. The recovery results were calculated by comparing the difference between the spiked and the un-spiked sample that were analyzed under the same conditions.
The extracted ion chromatogram obtained from the specificity validation was shown in Fig.3, indicating that the method had good specificity. The calibration curve had good linearity, and LOD and LOQ also met the requirements (as shown in Table 2). During the investigation of the precision, stability, repeatability and sample recovery rate of the nine components, the RSD values were all ≤ 3.2%, which met the requirements. These results were shown in Additional file 3, which showed that the method was stable and reliable.
Table 2 Calibration curve, LOD, and LOQ
Analytes
|
y = Ax + B
|
Linear range
(μg/mL)
|
R2
|
LOD
(μg/mL)
|
LOQ
(μg/mL)
|
Protocatechuic acid
|
y = 244691x - 2895.5
|
1.553-0.025
|
0.9991
|
0.0036
|
0.012
|
Hydroxytyrosol
|
y = 20152x - 635.69
|
2.997-0.047
|
0.9997
|
0.0050
|
0.015
|
Caffeic acid
|
y = 534896x - 6034.8
|
0.753-0.012
|
0.9997
|
0.0026
|
0.006
|
Epicatechin
|
y = 28186x + 766.44
|
6.108-0.096
|
0.9992
|
0.0080
|
0.024
|
Salidroside
|
y = 30434x - 183.7
|
2.295-0.036
|
0.9990
|
0.0030
|
0.018
|
Chlorogenic acid
|
y = 217299x - 29966
|
25.500-0.398
|
0.9994
|
0.1250
|
0.025
|
Desrhamnosyl isoacteoside
|
y = 122310x - 2488.7
|
1.536-0.024
|
0.9991
|
0.0030
|
0.010
|
Rutin
|
y = 9600.2x - 131.34
|
0.813-0.013
|
0.9988
|
0.0026
|
0.006
|
Liriodendrin
|
y = 13136x - 1540.9
|
6.004-0.094-
|
0.9973
|
0.0060
|
0.023
|
Apparatus and optimal chromatographic and MS/MS conditions
An ACQUITY I-Class UPLC system (Waters Corp, MA, USA) with a conditional auto-sampler and an Acquity I-Class UPLC BEH C18 Column (2.1 mm × 100 mm, internal diameter 1.7 µm) was used for the analyses. The system was also equipped with a Waters VanGuard BEH C18 (2.1 mm × 5 mm, 1.7 µm) column. The column and auto-sampler were maintained at 40 °C and 25 °C, respectively. The injection volume was 1 µL. The gradient mobile phase system consisting of 0.1% formic acid in acetonitrile (B) and 0.1% aqueous formic acid (A) was applied at a flow rate of 0.3 mL/min and run time of 6 min. The mobile phase consisted of 0.1% formic acid in water as solvent A and 0.1% formic acid in acetonitrile as solvent B. The gradient program was as follows: 0 – 0.5 min, 10% ~ 10% B; 0.5 – 2.5 min linear gradient 10% ~ 40% B; 2.5 – 4 min, 40% ~ 90% B; 4 – 4.5 min, 90% ~ 90% B; 4.5 – 6 min, 90%~10% B. The flow rate was 0.3 mL/min. The injection volume was 1 µL with partil loop mode. The temperature of the sample manager was maintained at 15 ℃.
Mass spectrometric detection was performed with an XEVO TQS Triple-Quadrupole Tandem Mass Spectrometer (Waters Corp, Milford, MA, USA) equipped with an electrospray ionization (ESI) source. The mass spectrometer parameters were: Capillary voltage, 3.0 kV; Capillary ionization voltage, 3 kV; ion source temperature, 120 °C; spray gas and backflush gas, N2; desolvation gas flow rate, 650 L/h; and desolvation gas temperature, 350 °C. Multiple reaction monitoring (MRM) mode was used for quantification. The optimal parameters for the analytes in the MRM mode are listed in Table 3. All data were obtained using MasslynxTM V4.1 software and processed using the QuanlynaTM V4.1 (Waters Corp., Millford, MA, USA) workstation.
Table 3 Mass spectrometric parameters of nine determination components
Component
|
Molecular formula
|
Retention time / tR (min)
|
Quantitative ion pair
|
Cone voltage
|
Collision energy
|
Protocatechuic acid
|
C7H6O4
|
0.89
|
152.9-109
|
30
|
15
|
Hydroxytyrosol
|
C8H10O3
|
0.90
|
153.5-123
|
20
|
15
|
Caffeic acid
|
C9H8O4
|
1.50
|
178.8-134.8
|
35
|
15
|
Epicatechin
|
C15H14O6
|
1.57
|
289.1-244.6
|
35
|
15
|
Salidroside
|
C14H20O7
|
0.95
|
299.1-119
|
35
|
20
|
Chlorogenic acid
|
C16H18O9
|
1.24
|
353-191
|
30
|
15
|
Desrhamnosyl isoacteoside
|
C23H26O11
|
2.05
|
477-161
|
30
|
15
|
Rutin
|
C27H30O16
|
1.91
|
609.2-300
|
30
|
15
|
Liriodendrin
|
C34H46O18
|
1.81
|
741.7-417.5
|
30
|
25
|
Determination of UPLC-ESI-TQS MS/MS content of nine components in thirty batches of DXT
Nine components with higher content or stronger activity from the tested components were selected, and UPLC-ESI-TQS MS/MS was used to determine the content of different batches of DXT medicinal materials in thirty different regions under the above optimized conditions. The content determination results were shown in Table 4.
Quality assessment
Hierarchical clustering analysis
In order to analyze the changes in the content of each component between DXT medicinal materials in different regions as a whole, SPSS software (IBM SPSS Statistics 20.0 Developer, IBM Corp., NY, USA) was used to perform hierarchical clustering analysis (HCA) on the content determination data of nine components in thirty batches of DXT medicinal materials. Using the ward method, the squared Euclidean distance was selected as the metric to perform HCA. The HCA dendrogram was shown in Fig.4. It could be seen that thirty samples were divided into two major clusters (I and II) based on the contents of nine compounds. Samples from Guizhou, Sichuan, Yunnan, and Guangxi were included in the cluster I, while samples from eastern provinces such as Anhui, Hubei, and Jiangxi were clustered into the cluster II, which reflected the similar content of chemical composition in the same cluster of samples. In addition, it was also found that the areas of cluster I was located in the southwest of China, while the area of cluster II belongs to east of China. Obviously, the differences between the east and the southwest of China affect the components of DXT, such as geographic location, climate, and precipitation [23].
Cluster I was further divided into another two branches (group ① and ②). It could be found that all samples from Sichuan, and Bijie and Zunyi of Guizhou were grouped into group ①, while samples from Guangxi and Yunnan, as well as Anshun, Duyun, Xingyi, Liupanshui of Guizhou were gathered in the group ②. Geographically, Bijie and Zunyi are adjacent to Sichuan, while Anshun, Duyun, Xingyi, Liupanshui is adjacent to Yunnan and Guangxi. And Guiyang is located in the central part of Guizhou, with similar distances from Sichuan, Yunnan, and Guangxi. In the dendrogram, samples from Guiyang were distributed in both group ① and ②, which could explain to a certain extent that the quality of the DXT in southwest of China was not completely consistent.
Principal components analysis
In order to analyze the relationship between the content of the component and the region, principal components analysis (PCA) was performed using SPSS software (IBM SPSS Statistics 20.0 Developer, IBM Corp., NY, USA). And 2D and 3D plots of loading plot and score plot were obtained as shown in Fig.5.
The loading plot was used to describe data features, it could be known that the relationship between variable attribute features. And the distinction between samples could be got from the score plot. It can be seen from the calculation of the principal component (PC) that the larger the absolute value of the loading, the greater the influence on the PC. Three PCs were extracted (PC 1, 2 and 3), and the cumulative contribution rate was 77.18%.
Combining the distribution characteristics (Additional file 4) and the loading plot (Fig.5 (a) and (b)) of the content, in the PC1 axis, protocatechuic acid and rutin were negatively correlated and other components were positively correlated. The contents of the first two component variables in the southwestern samples were higher than that in the eastern samples, while other components in the southwestern samples were less than that in eastern samples, which indicated that the differences in the contents of the component variables in different regions were mainly reflected in this PC. Additionally, hydroxytyrosol and desrhamnosyl isoacteoside were mainly moved on PC1 axis, while not obvious on other axis, indicating that these two components were almost completely affected by PC1, that is, the content of the two components in the eastern region was significantly higher than that in the western region, which was consistent with the quantitative results.
As shown in score plots (Fig.5 (c) and (d)), the samples in the eastern and southwest regions except sample 19 were almost completely separated, which was approximately consistent with HCA result. On the 2D graph, sample 25 from Guizhou Bijie was moved in the second quadrant, and other samples were distributed near by the PC1 axis. Additionally, the samples derived from eastern regions were shifted in the positive direction of PC1 axis, while the samples in the southwest region were mainly distributed in the negative direction of the axis. Analyzed with 2D load plot, it could be speculated that the medicinal materials from eastern region were mainly affected by the component variables desrhamnosyl isoacteoside, while the medicinal materials from southwestern region were mainly affected by the component variable protocatechuic acid. When the third PC was added, the sample 4 from Jiangxi Jiujiang moved significantly. The quantitative results showed that desrhamnosyl isoacteoside was significantly less than other samples in the same region, while the content of salidroside was significantly higher. Combining the results of loading plots and score plots, the main variable that affects sample 25 was rutin, and the main variable that made sample 4 moved was salidroside.
To sum up, the components that made the quality difference of DXT medicinal materials of the east and the southwest might include desrhamnosyl isoacteoside and protocatechuic acid.
Partial least squares discriminant analysis
In order to further screen the key components that caused the differences in the quality of DXT from various regions, the raw data of the nine measured components of DXT from different regions was input as variable values into SIMCA 14.1 software (Umetrics, Sweden) for partial least squares discriminant analysis (PLS-DA), the variable importance of projection (VIP) value of each index component was obtained. As shown in Fig.6, when VIP > 1 as the screening criterion, it could be concluded that protocatechuic acid, desrhamnosyl isoacteoside, liriodendrin, and hydroxytyrosol had significant differences in the DXT from different regions.