Quantitative Determination and Validation of Four Ketones in Salvia miltiorrhiza Bunge Using Quantitative Proton Nuclear Magnetic Resonance Spectroscopy

Salvia mltiorrhiza Bunge (SMB) is native to China, whose dried root has been used as medicine. A few chromatographic- or spectrometric-based methods have already been used to analyze the lipid-soluble components in SMB. However, the methodology of qNMR on the extracts of fresh SMB root has not been verified so far. The purpose of this study was to establish a fast and simple method to quantify the tanshinone I, tanshinone IIA, dihydrotanshinone, and cryptotanshinone in fresh Salvia Miltiorrhiza Bunge root without any pre-purification steps using 1H-NMR spectroscopy. The process is as follows: first, 70% methanol aqueous extracts of fresh Salvia Miltiorrhiza Bunge roots were quantitatively analyzed for tanshinone I, tanshinone IIA, dihydrotanshinone, and cryptotanshinone using 1H-NMR spectroscopy. Different internal standards were tested and the validated method was compared with HPLC. 3,4,5-trichloropyridine was chosen as the internal standard. Twelve samples of Salvia Miltiorrhiza Bunge were quantitatively analyzed by qNMR and HPLC respectively. Then, the results were analyzed by chemometric approaches. This NMR method offers a fast, stable, and accurate analysis of four ketones: tanshinone I, tanshinone IIA, dihydrotanshinone, and cryptotanshinone in fresh roots of Salvia Miltiorrhiza Bunge.

Modern pharmacological studies have shown that tanshinones have many pharmacological activities such as (1) antioxidation: CTS has antioxidant in vivo and in vitro pharmacological

Quantitative Analysis of SMB Using 1 H-qNMR
The sample solutions were optimized to obtain the best separation and stability for all the integrated signals in the 1 H-NMR spectrogram. Quantitation was performed using an NMR sampletube adapter at 298 K. Prior to all the measurements, the analyte and IS were qualitatively analyzed using routine 1 H experiments to determine the longest spin-lattice relaxation time (usually at least 5 times the T1, thus the optimized relaxation delay of 25 s was obtained). The 3,4,5-trichloropyridine was chosen as the IS due to its good solubility and stability.
Two basic requirements for an IS should be met.; one is that the signal of the IS and the target signal from the analyte should not overlap, and the other is that the resonance of the IS and the selected compound should not occur in crowded spectral regions. 3,4,5-trichloropyridine was chosen as an IS for the quantitative analysis.

Specificity and Selectivity
Specificity and selectivity are key prerequisites that must be evaluated to avoid possible interference from other components in the sample solution. We thus compared the 1 H-NMR spectra of the standard references and sample solution with IS, and IS and solvent individually ( Figure 2).

Quantitative Analysis of SMB Using 1 H-qNMR
The sample solutions were optimized to obtain the best separation and stability for all the integrated signals in the 1 H-NMR spectrogram. Quantitation was performed using an NMR sample-tube adapter at 298 K. Prior to all the measurements, the analyte and IS were qualitatively analyzed using routine 1 H experiments to determine the longest spin-lattice relaxation time (usually at least 5 times the T1, thus the optimized relaxation delay of 25 s was obtained). The 3,4,5-trichloropyridine was chosen as the IS due to its good solubility and stability.
Two basic requirements for an IS should be met.; one is that the signal of the IS and the target signal from the analyte should not overlap, and the other is that the resonance of the IS and the selected compound should not occur in crowded spectral regions. 3,4,5-trichloropyridine was chosen as an IS for the quantitative analysis.

Specificity and Selectivity
Specificity and selectivity are key prerequisites that must be evaluated to avoid possible interference from other components in the sample solution. We thus compared the 1 H-NMR spectra of the standard references and sample solution with IS, and IS and solvent individually ( Figure 2). We found that the resonances assigned to these protons are quite separated from the others, such as to have a good integration without interferences (Figure 3), compared with the data reported in the literature [19], signals for Tan  We found that the resonances assigned to these protons are quite separated from the others, such as to have a good integration without interferences (Figure 3), compared with the data reported in the literature [19], signals for Tan Ⅰ at 7.81 ppm (dd,1H), Tan ⅡA at 7.36 ppm (d,1H), DTS at 8.36 ppm (d,1H), CTS at 4.38 ppm (dd,1H) were selected as the quantitation signals.

Linearity, LOD and LOQ.
The intensity of the response signal is directly proportional to the amount of nuclei, as described in Equation (1). Consequently, the linearity regression yielded a good correlation coefficient (r 2 > 0.985). The concentration ratios of the four references ranged between 0.02 and 0.18 mg/mL (CTS), 0.02 and 0.24 mg/mL (Tan I), 0.03 and 0.33 mg/mL (Tan IIA) and 0.02 and 0.24 mg/mL (DTS), respectively.
The limit of detection (LOD) presents the lowest detectable analyte concentration, whilst the limit of quantitation (LOQ) represents the lowest quantifiable analyte concentration. These are two We found that the resonances assigned to these protons are quite separated from the others, such as to have a good integration without interferences (Figure 3), compared with the data reported in the literature [19], signals for Tan Ⅰ at 7.81 ppm (dd,1H), Tan ⅡA at 7.36 ppm (d,1H), DTS at 8.36 ppm (d,1H), CTS at 4.38 ppm (dd,1H) were selected as the quantitation signals.

Linearity, LOD and LOQ.
The intensity of the response signal is directly proportional to the amount of nuclei, as described in Equation (1). Consequently, the linearity regression yielded a good correlation coefficient (r 2 > 0.985). The concentration ratios of the four references ranged between 0.02 and 0.18 mg/mL (CTS), 0.02 and 0.24 mg/mL (Tan I), 0.03 and 0.33 mg/mL (Tan IIA) and 0.02 and 0.24 mg/mL (DTS), respectively.
The limit of detection (LOD) presents the lowest detectable analyte concentration, whilst the limit of quantitation (LOQ) represents the lowest quantifiable analyte concentration. These are two

Linearity, LOD and LOQ
The intensity of the response signal is directly proportional to the amount of nuclei, as described in Equation (1). Consequently, the linearity regression yielded a good correlation coefficient (r 2 > 0.985). The concentration ratios of the four references ranged between 0.02 and 0.18 mg/mL (CTS), 0.02 and 0.24 mg/mL (Tan I), 0.03 and 0.33 mg/mL (Tan IIA) and 0.02 and 0.24 mg/mL (DTS), respectively.
The limit of detection (LOD) presents the lowest detectable analyte concentration, whilst the limit of quantitation (LOQ) represents the lowest quantifiable analyte concentration. These are two fundamental elements of method validation defining the limitations of an analytical method. In qNMR, the LOD and LOQ cannot be determined by means of SNR (signal noise ratio) as the NMR response signals are Lorentzian lines. Hence, the LOD and LOQ were determined using the standard deviation of the response σ and the slope S of a calibration curve obtained in the linearity study by Equations (1) and (2): Molecules 2020, 25, 2043 The results for linearity, LOD and LOQ are shown in Table 1. To determine the optimal time window between sample collection and analysis, it is crucial to conduct a sample stability test. A solution with an RSD (relative standard deviation) value of less than 3.0% is considered stable. We analyzed the stability of the sample solution with IS of PS07 at room temperature (~25 • C) at 0, 6, 12, 18 and 24 h intervals and determined their RSD values. The content difference and RSD values of the sample solution are shown in Table 2. The accuracy of an analytical method is determined by measuring the correlation of theoretical value and the value found. To study the accuracy of our method, we carried out a recovery study by adding a known amount of Tan I, Tan IIA, DTS and CTS into the SMB extraction system, respectively. Each concentration level was repeated in triplicate. The accuracy was estimated by measuring the mean recovery and relative standard deviation (RSD). The recovery was determined by comparing the calculated value of added reference material and true value according to Equation (3): where m x is the calculated mass of analyte after reference addition, m o is the calculated mass of original analyte before reference addition and m s is the true mass of the reference addition. The results are summarized in Table 3.

Precision and Intermediate Precision
ICH guidelines recommend that precision is assessed using six repeated determinations. In this study, the intermediate precision was evaluated by a second analyst on an alternative probe (5 mm PABBO 600S3 BBF-H-D-05 Z SP (Z114607)). The calculated content of each analyte and statistical result were shown in Table 4. There were no significant differences in precision and intermediate precision between the results.  The robustness of an analytical procedure is measured in terms of its capacity to remain unaffected by small, but deliberate variations in method parameters listed in the procedure documentation, which indicates suitability during normal routine usage. In this study, PS07 was used to test for robustness. Three parameters were varied independently: (1) number of scans; (2) relaxation delay; (3) acquisition time. We observed that our method was largely unaffected by variations in all of the tested parameters (Table 5).

Quantitative NMR Analysis of Real Samples
PS01-PS12 were also analyzed using qNMR, and the results obtained from the two methods were compared. We observed no significant differences between the results obtained using HPLC and qNMR (Table 6). Therefore, qNMR represents an alternative method of quantifying compounds within complex matrices.

Sample Representation
PCA is often used to reveal the clustering behavior of samples. We observed different clustering behaviors for sample extracts obtained from different areas (PC1: 51.9%, PC2: 32.3%). As shown in the PCA scores plot (Figure 4), most samples clustered to the middle of the plot. However, PS07 clustered different from the other samples. The samples of PS08 and PS10 are relatively concentrated. PCA is often used to reveal the clustering behavior of samples. We observed different clustering behaviors for sample extracts obtained from different areas (PC1: 51.9%, PC2: 32.3%). As shown in the PCA scores plot (Figure 4), most samples clustered to the middle of the plot. However, PS07 clustered different from the other samples. The samples of PS08 and PS10 are relatively concentrated. To further investigate our observation, the cluster analysis, which measures the distances or correlation between objects [23][24][25][26], was carried out. Consistent with the PCA results, we found that PS07 clustered the furthest away from the other samples ( Figure 5). In the dendrogram, PS08 and PS10 were belong to the same node. To further investigate our observation, the cluster analysis, which measures the distances or correlation between objects [23][24][25][26], was carried out. Consistent with the PCA results, we found that PS07 clustered the furthest away from the other samples ( Figure 5). In the dendrogram, PS08 and PS10 were belong to the same node. Our results were consistent with the geographic information of these samples ( Figure 6). Among them, the collection area of PS07 has the highest latitude, while the collection areas of PS08 and PS10 have similar latitudes, which indicates that the latitude of the collection area may have some impact on the cluster results of some sample. Our results were consistent with the geographic information of these samples ( Figure 6). Among them, the collection area of PS07 has the highest latitude, while the collection areas of PS08 and PS10 have similar latitudes, which indicates that the latitude of the collection area may have some impact on the cluster results of some sample.

Instruments and Parameters
Fresh SMB roots samples were milled using a Q-250B1mill (Shanghai, China). A KH5200DE ultrasonic bath (Nanjing, China) was used for extraction.
The samples for qNMR analysis were measured on a 600 MHz Avance III HD spectrometer with a TXI probe at 298K (Bruker Corporation, Faellanden, Switzerland). 1 H-NMR experimental parameters were shown as follows: zg30 pulse sequence with 32 scans of 32 K data points in a spectral width of 12019.2 (20 ppm), acquisition time 5.0 s, relaxation delay 25 s. All of the data processing was performed by using MestReNova11.0 and SPSS21.0: first, we use MestReNova's superposition function to superimpose the internal standard, solvent, sample, and 4 standards, and select the appropriate signal peak. Then, the sample is integrated according to the selected signal peak, and the content of the four components in the sample is calculated by combining with the standard curve. After the above process is completed, SPSS is used for principal component analysis and cluster analysis.

Preparation of Sample Solution
Approximately 2.0 g of the SMB root powder was added into 20 mL 70% methanol aqueous solution and settled for 12 h at room temperature, then extracted for 45 min in an ultrasonic bath. The supernatant was cooled to room temperature and filtered. Next, 1500 µL of filtered liquor was transferred into a clean Eppendorf tube and nitrogen at 35 • C. The 600 µL of CDCl 3 (containing 0.03% v/v TMS) was then added to dissolve the sample for NMR analysis. The NMR sample tube was assembled for analysis at 298 K and sealed prior to the 1 H-NMR measurement. For HPLC analysis, the filter liquor was directly injected into the HPLC system. All of the extractions and subsequent NMR measurements were performed in triplicate.

Quantitative NMR Analysis
The most important fundamental relation of qNMR is the signal response (integrated signal area) I x in a spectrum that is directly proportional to the number of nuclei N x generating the corresponding resonance line [27,28]: K s is an unknown spectrometer constant, which is a constant for all resonance lines in the same 1 H single-pulse NMR spectrum. Accordingly, the determination of the relative area ratios I x /I y is the most efficient way to obtain quantitative results by using Equation (5) when K s cancels for the ratio: For the purity determination of a substance, an internal standard with known purity is needed. Based on Equation (5), the component purity can be calculated from the NMR intensity via Equations (6) and (7): W x and P x represent the mass and purity of the analyte. M x and M Std are the molar masses of the analyte and the standard (3,4,5-trichloropyridine: 182.44 g/mol). m is the weighed mass of the investigated sample. m Std and P Std are the weighed mass and the purity (99.5%) of the standard. N Std and I Std correspond to the number of protons for the standard (in this experiment is 2) and the integrated signal area of a typical NMR line (which was 2 in this experiment). N x and I x correspond to the number of protons for the analyte 1 H.

HPLC Analysis
The qNMR results were verified using HPLC. The HPLC experiments were performed using an Agilent 1260 HPLC system (Agilent, Palo Alto, CA, USA) equipped with a Diode-array detector and a Sunfire C 18

Chemometrics Methods
Principal component analysis (PCA) and hierarchical cluster analysis were used to analyze the NMR data.

Conclusions
A selective and accurate qNMR method was established for quantitatively determining and validating four lipid-soluble ketones in SMB extracts. The results of the analysis of linearity, precision, stability, accuracy, LOD and LOQ demonstrated that 1 H-NMR can be used to accurately determine the content of ketones. We found that the results generated using the qNMR method were consistent with those obtained using HPLC analysis. Whilst HPLC took more than 60 min to complete, qNMR took only 19 min for each sample. Due to the good linearity between IS and the references, qNMR can be performed without any standard references. Although future studies will be required to maximize its application potential, NMR profiles can be combined with chemometric approaches to give insights into the relationship between the quality, properties and geographic origins of plant-based materials. In summary, qNMR presents a rapid and effective method of analyzing various plant extracts and their by-products.