Absolute Quantification of Isoflavones in the Flowers of Pueraria lobata by qHNMR

Pueraria lobata (Willd.) Ohwi. is a widely used medicinal plant in Korea, China, and Japan. The flower of P. lobata (Puerariae Flos) contains various bioactive substances such as triterpenoidal saponins and isoflavonoids. In this study, we developed a quantitative analysis of the isoflavones of Puerariae Flos by quantitative proton nuclear magnetic resonance (qHNMR) spectroscopy using the internal calibrant (IC). From the qHNMR results, the isoflavone content was found to be 7.99% and 10.57% for the MeOH sonication extract (PLs) and the MeOH reflux extract (PLr) of Puerariae Flos, respectively. The quantified isoflavone content was validated using the conventional analytical method, high-performance liquid chromatography with ultraviolet detection (HPLC-UV). The present study shows that validated qHNMR spectroscopy is a reliable method for quantifying and standardizing the isoflavone content in Puerariae Flos.

qHNMR is a powerful analytical tool for purity determination and quantification of compounds in mixtures. There is a growing interest in the application of qHNMR methods to quantify various classes of components in natural products [22,23]. qHNMR is preferable over conventional chromatography-based quantitative methods because it allows for the simultaneous identification and quantification of the complex matrix in a single run. The integration of an individual resonance from different chemical markers depends on the number of protons and the molar concentration of the compound [22]. There are four methods for performing quantitative analysis: (i) relative 100% method without calibration, (ii) internal calibration (IC), (iii) external calibration (EC), and (iv) external calibration of the internal solvent signal (ECIC) [24]. Among them, IC is a direct method for the simultaneous measurement of standard and sample mixtures dissolved in solvents. IC does not require a calibration curve with an external calibrant, and it is the most common and easiest approach for absolute quantitation [24]. Because the samples and calibrants were subjected to identical experimental parameters, the analytical error is significantly smaller than that in other methods and the experimental time is shortened [22]. The objective of the present study was to develop a validated qHNMR method using proton signal selection criteria to accurately determine the isoflavone content in two extracts of Puerariae Flos: 5-methoxydaidzein (1) [25], tectorigenin (2) [1], genistin (3) [26], glycitin (4) [1], tectoridin (5) Figure 1). The quantitative results obtained by qHNMR were verified using conventional chromatography. The present method can be applied to the standardization and chemical profiling of the raw materials of dietary supplements and foods using Puerariae Flos.
performance liquid chromatography-photodiode array (HPLC-PDA), liquid chromatog raphy-electrospray ionization-tandem mass spectrometry (LC-ESI-MS) [18], liquid chro matography coupled with triple stage quadrupole tandem mass spectrometry (TSQ MS/MS) [19], ultra-performance liquid chromatography-electrospray ionization-tandem mass spectroscopy (UPLC-ESI-MS/MS) [20], and quantitative 1 H NMR (qHNMR) [14,21 qHNMR is a powerful analytical tool for purity determination and quantification o compounds in mixtures. There is a growing interest in the application of qHNMR meth ods to quantify various classes of components in natural products [22,23]. qHNMR is pre erable over conventional chromatography-based quantitative methods because it allow for the simultaneous identification and quantification of the complex matrix in a singl run. The integration of an individual resonance from different chemical markers depend on the number of protons and the molar concentration of the compound [22]. There ar four methods for performing quantitative analysis: (i) relative 100% method without ca ibration, (ii) internal calibration (IC), (iii) external calibration (EC), and (iv) external cal bration of the internal solvent signal (ECIC) [24]. Among them, IC is a direct method fo the simultaneous measurement of standard and sample mixtures dissolved in solvent IC does not require a calibration curve with an external calibrant, and it is the most com mon and easiest approach for absolute quantitation [24]. Because the samples and cal brants were subjected to identical experimental parameters, the analytical error is signif cantly smaller than that in other methods and the experimental time is shortened [22]. Th objective of the present study was to develop a validated qHNMR method using proto signal selection criteria to accurately determine the isoflavone content in two extracts o Puerariae Flos: 5-methoxydaidzein (1) [25], tectorigenin (2) (Figure 1). The quantitative results obtained by qHNMR were ver ified using conventional chromatography. The present method can be applied to th standardization and chemical profiling of the raw materials of dietary supplements an foods using Puerariae Flos.

1 H NMR Signal Assignment and Identification
The resonances for seven major isoflavones (1-7) of the Puerariae Flos extracts pre pared by two different methods [MeOH sonication (PLs) and MeOH reflux (PLr)] wer classified into three parts, as shown in Figure 2: (i) signals for the methoxy protons of th A-ring at 3.7-3.9 ppm, (ii) signals for an AA′XX′ spin system of the B ring and th H5/H6/H8 protons of the A ring at 6.3-7.5 ppm, (iii) H-2 singlets of the C-ring at 8.2-8. ppm. The first two areas (i and ii) showed intense signal overlaps, which can make precis signal assignment challenging and lead to inaccurate quantitation. However, relativel simple and well-separated singlets from the H-2 proton of the C-ring at 8.2-8.5 ppm ar appropriate to assign and quantify the isoflavones in the extracts [14,21].

1 H NMR Signal Assignment and Identification
The resonances for seven major isoflavones (1-7) of the Puerariae Flos extracts prepared by two different methods [MeOH sonication (PLs) and MeOH reflux (PLr)] were classified into three parts, as shown in Figure 2: (i) signals for the methoxy protons of the A-ring at 3.7-3.9 ppm, (ii) signals for an AA XX spin system of the B ring and the H5/H6/H8 protons of the A ring at 6.3-7.5 ppm, (iii) H-2 singlets of the C-ring at 8.2-8.5 ppm. The first two areas (i and ii) showed intense signal overlaps, which can make precise signal assignment challenging and lead to inaccurate quantitation. However, relatively simple and well-separated singlets from the H-2 proton of the C-ring at 8.2-8.5 ppm are appropriate to assign and quantify the isoflavones in the extracts [14,21]. The structures of isoflavones (1-7) have been identified as 5-methoxydaidzein (1) (7) [1] by analysis of the 1 H NMR and low-resolution electrospray ionization-tandem mass spectrometry (LR-ESI-MS) data and comparison with reported values. The H-2 signal of compound 2 was found to migrate in the 1 H NMR spectrum of the extract, and the shifted H-2 signal was confirmed by spiking experiments with pure standard tectorigenin (2) in the PLs extract as shown in Figure 3. This chemical shift inconsistency of the purified compound and the mixture state is caused by differences in the chemical environments, such as temperature, sample matrix, concentration, and pH. The structures of isoflavones (1-7) have been identified as 5-methoxydaidzein (1) [25], (7) [1] by analysis of the 1 H NMR and low-resolution electrospray ionization-tandem mass spectrometry (LR-ESI-MS) data and comparison with reported values. The H-2 signal of compound 2 was found to migrate in the 1 H NMR spectrum of the extract, and the shifted H-2 signal was confirmed by spiking experiments with pure standard tectorigenin (2) in the PLs extract as shown in Figure 3. This chemical shift inconsistency of the purified compound and the mixture state is caused by differences in the chemical environments, such as temperature, sample matrix, concentration, and pH.

Major Isoflavone Content Calculated by qHNMR
In the general qHNMR method, the relative and absolute contents of the target compounds were calculated using the peak integration method [28]. Because natural products are a mixture of various classes of compounds, when analyzing the NMR spectrum of the extract, it is difficult to separate and obtain the integration value of each peak due to signal overlapping. The precision and accuracy of qHNMR analysis depend on the accuracy of the integration of the target peak. In this study, although the resonances for H-2 of isoflavones were singlets and relatively well separated, the peak deconvolution method was applied to minimize the error in calculating the integral value. A global spectral deconvolution (GSD) function of MestReNova software was used to acquire peak integration. GSD is a fast and automatic algorithm that detects and deconvolutes all selected spectral peaks in a spectrum and extracts all impurities, noise, spikes, and signals from other untargeted molecules [29]. Methyl 3,5-dinitrobenzoate was selected as an internal calibrant for absolute quantitation, considering its stability and solubility [21]. A general criterion for selecting an internal calibrant is that it should be inactive to the sample, be of high purity, and show minimal signal overlap with the quantified signal [24]. The IC shows characteristic and well-defined signals at δH 8.91 (d, J = 2.2 Hz, 2H) and 9.04 (t, J = 2.2 Hz, 1H) ppm which were not overlapped with the constituent signals for the PLs and PLr extracts. Peak deconvolution was applied to the entire spectrum and quantified the H-2 singlet region at 8.2-8.5 ppm, resulting in black, blue, and red-colored lines representing the peak sum, peak curves, and peak residual, respectively. (Figure 4a). The chemical shifts at 8.45, 8.43, 8.39, 8.38, 8.33, 8.30, and 8.27 ppm were used to generate the areas and quantify the contents of compounds 5, 3, 7, 4, 6, 2, and 1, respectively. The areas for each deconvoluted peak were determined using the GSD option with five fitting cycles. The absolute content of the target compounds was calculated using the IC method [24]. The calculated content % w/w of the seven isoflavones (1-7) in the PLs extract were 0.95, 0.67, 0.43, 1.96, 1.86, 0.57, and 1.55, respectively (Table 1). All the quantified amounts were above the amount calculated by the limit of quantitation (LOQ) performed for system validation.

Validation Studies
The validation results proved that the developed method is suitable for quantification. The 1 H NMR method was validated in terms of sensitivity and accuracy based on the ICH guidelines [30].

Major Isoflavone Content Calculated by qHNMR
In the general qHNMR method, the relative and absolute contents of the target compounds were calculated using the peak integration method [28]. Because natural products are a mixture of various classes of compounds, when analyzing the NMR spectrum of the extract, it is difficult to separate and obtain the integration value of each peak due to signal overlapping. The precision and accuracy of qHNMR analysis depend on the accuracy of the integration of the target peak. In this study, although the resonances for H-2 of isoflavones were singlets and relatively well separated, the peak deconvolution method was applied to minimize the error in calculating the integral value. A global spectral deconvolution (GSD) function of MestReNova software was used to acquire peak integration. GSD is a fast and automatic algorithm that detects and deconvolutes all selected spectral peaks in a spectrum and extracts all impurities, noise, spikes, and signals from other untargeted molecules [29]. Methyl 3,5-dinitrobenzoate was selected as an internal calibrant for absolute quantitation, considering its stability and solubility [21]. A general criterion for selecting an internal calibrant is that it should be inactive to the sample, be of high purity, and show minimal signal overlap with the quantified signal [24]. The IC shows characteristic and well-defined signals at δ H 8.91 (d, J = 2.2 Hz, 2H) and 9.04 (t, J = 2.2 Hz, 1H) ppm which were not overlapped with the constituent signals for the PLs and PLr extracts. Peak deconvolution was applied to the entire spectrum and quantified the H-2 singlet region at 8.2-8.5 ppm, resulting in black, blue, and red-colored lines representing the peak sum, peak curves, and peak residual, respectively. (Figure 4a) 27 ppm were used to generate the areas and quantify the contents of compounds 5, 3, 7, 4, 6, 2, and 1, respectively. The areas for each deconvoluted peak were determined using the GSD option with five fitting cycles. The absolute content of the target compounds was calculated using the IC method [24]. The calculated content % w/w of the seven isoflavones (1-7) in the PLs extract were 0.95, 0.67, 0.43, 1.96, 1.86, 0.57, and 1.55, respectively (Table 1). All the quantified amounts were above the amount calculated by the limit of quantitation (LOQ) performed for system validation.

Sensitivity
According to the ICH guidelines, the limit of detection (LOD) and the limit of quantitation (LOQ) of a liquid chromatography system are determined by calculating the signal-to-noise (S/N) ratio. The acceptable S/N ratios were 3:1 and 10:1 for the LOD and LOQ, respectively. In this experiment, the sensitivity parameter of the 1 H NMR quantification method was determined by calculating the S/N ratio using the S/N ratio calculator function of the MestReNova software. The calibration curve was plotted based on linear regression analysis of the S/N ratio versus sample concentration, and the LOD (0.0014 mg in 600 μL) and LOQ (0.0036 mg in 600 μL) were calculated using a linear equation ( Figure  S15)

Validation Studies
The validation results proved that the developed method is suitable for quantification. The 1 H NMR method was validated in terms of sensitivity and accuracy based on the ICH guidelines [30].

Sensitivity
According to the ICH guidelines, the limit of detection (LOD) and the limit of quantitation (LOQ) of a liquid chromatography system are determined by calculating the signalto-noise (S/N) ratio. The acceptable S/N ratios were 3:1 and 10:1 for the LOD and LOQ, Plants 2022, 11, 548 6 of 10 respectively. In this experiment, the sensitivity parameter of the 1 H NMR quantification method was determined by calculating the S/N ratio using the S/N ratio calculator function of the MestReNova software. The calibration curve was plotted based on linear regression analysis of the S/N ratio versus sample concentration, and the LOD (0.0014 mg in 600 µL) and LOQ (0.0036 mg in 600 µL) were calculated using a linear equation (Figure S15).

Accuracy
The accuracy parameter was determined using the spike-recovery method [21]. Three replicates of PLs containing a known amount of methyl 3,5-dinitrobenzoate were quantified using the GSD method before and after spiking with tectoridin (5). The sample was prepared as mentioned in the Experimental section. 50 µL of the stock solution containing 2500 µg in 200 µL of DMSO-d 6 was spiked into the three samples. The average recovery % of spiked 5 was 98.4072%, which was within the acceptable range (Figures S16 and S17) [31] and the relative error was 0.7868%. The recovery % was calculated using the following Equation (1): conc. of (spiked sample − unspiked sample) spiking concentration × 100 (1)

Comparison of Two Different Sample Preparation Methods, PLr and PLs
Two different extraction methods were also explored: sonication and heat reflux. Sonication is a cavitation process that causes swelling and diffusion across the cell wall upon the application of sound energy, whereas heat reflux is accomplished by allowing hot solvent in solid tissue and leaching out the compounds. The extraction method is a crucial factor for varying the compositions and contents of the chemical compounds in an extract. Heat reflux was found to be a superior extraction method over the ultrasonic-assisted method because it gives high yields of bioactive chemical contents of polysaccharides, polyphenols, and flavonoids [32]. Similarly, as shown in the results (Table 1), the isoflavonoid content was higher in the MeOH reflux (PLr) extract than in the MeOH sonication (PLs) extract, indicating that the heat application method is better than the sound application method.

Plant Materials
The dried flowers of P. lobata (Puerariae Flos) were purchased from CK Pharm Co. Ltd., Seoul, Republic of Korea, in August 2019. The plant material was authenticated by one of the authors, Prof. Dae Sik Jang, and a voucher specimen (PULO4-2019) was deposited at the Laboratory of Natural Product Medicine, Kyung Hee University.

MeOH Reflux of Puerariae Flos (PLr)
The same plant material as PLs (50 g) was extracted with MeOH (500 mL × 2) under reflux for 2 h. The solvent was removed using a rotary vacuum evaporator (EYELA SB-1200) at 40 • C to yield a MeOH extract (PLr, 7.9 g).

Sample Preparation
Flos extracts (PLr and PLs, 10 mg each) and methyl 3,5-dinitrobenzoate (IC, 1 mg) were precisely weighed separately. Methyl 3,5-dinitrobenzoate (IC, 1 mg) was dissolved in 1 mL of DMSO-d 6 , and 150 µL of the IC solution was transferred to a glass vial containing 550 µL of DMSO-d 6 . Then, 650 µL of the resulting solution was added to 10 mg of the two extracts (PLs and PLr). Finally, 600 µL of the mixtures (IC with each extract) was transferred into a 5 mm NMR tube for the acquisition of the 1 H NMR spectrum. Samples were prepared and analyzed in triplicate.

NMR Data Acquisition and Processing
The NMR spectra were acquired in DMSO-d 6 on a 600 MHz Bruker AVANCE NEO spectrometer (Oxford magnet, Switzerland) at the Core Research Support Center for Natural Products and Medical Materials (CRCNM). 1 H NMR spectra were recorded using the following parameters: a calibrated 90 • pulse (P1), relaxation delay (D1), 60.0 s; acquisition time (AQ), 4.0 s; number of scans (NS), 64; free induction decay (FID) data points, 256 K; spectral width, 17,857.1 Hz; receiver gain (RG), 32; temperature, 298 K. NMR spectra were processed and analyzed with Mnova 12.0.4 software (Santiago de Compostela, Spain). Post-acquisition processing was performed by zero-filling to 256 K, Lorentzian-Gaussian apodization (line broadening = −0.3, Gaussian factor = 0.05), manual phasing, and baseline correction (fifth-order polynomial). The residual DMSO-d 6 peak was referenced at 2.50 ppm. Global spectral deconvolution (GSD) was performed by performing peak picking of the entire spectrum and the GSD function in the MestReNova software with five fitting cycles [21]. This created a peak table containing all the deconvoluted peaks with their Plants 2022, 11, 548 8 of 10 defining peak areas, which were assigned to the H-2 singlet signals at 8.2-8.5 ppm to quantify each isoflavone.

Conclusions
In this study, the qHNMR method was developed and successfully applied as a simple, rapid, sensitive, and reliable tool to quantify all seven isoflavones in the Puerariae Flos extracts obtained by two different methods. The proton signals of quantified isoflavones from H-2 singlets of C-ring at 8.2-8.5 ppm were used for quantification because they are distinct and do not overlap with each other or with the signals from other impurities. Our qHNMR quantification method was validated by the conventional HPLC method, and the quantification results of qHNMR closely corresponded to those of HPLC. This is the first report on a validated qHNMR method to quantify the secondary metabolites of Pueraria Flos and to compare the data with those procured by the conventional HPLC method. Hence, this qHNMR method can facilitate the quality control of Puerariae Flos for the preparation and botanical standardization of medicinal products, while overcoming all the limitations of HPLC. Furthermore, this study can be a guideline for establishing quantitative and chemical fingerprinting methods using NMR for various natural products containing isoflavonoids as the main components.