A comprehensive strategy to monitor quality consistency of Weibizhi tablets based on integrated MIR and UV spectroscopic fingerprints, a systematically quantified fingerprint method, antioxidant activities and UPLC-Q-TOF-MS chemical profiling

Abstract

As complicated mixture systems, traditional Chinese medicines (TCM)/herbal medicines (HM) are very difficult to comprehensively investigate with regard to their quality consistency by chromatographic fingerprinting using a single detection technique. Therefore, finding a rapid, effective and comprehensive quality control method is of great importance for guaranteeing TCM/HM safety and efficacy in clinical applications. In this research, a novel combination strategy of mid-infrared (MIR) and ultra violet (UV) spectroscopic fingerprinting using Fourier transform mid-infrared (FTMIR) spectrometry and flow injection analysis (FIA) was developed and applied to monitor the quality consistency of HM in popular patent drug Weibizhi tablets (WBZT). In order to completely identify saturated and unsaturated chemical bonds for HM components in WBZT, an integrated assessment method based on MIR and UV spectroscopic fingerprinting was set up. The quality grades of 27 batches of WBZT samples from the same manufacturer were successfully discriminated by means of a systematically quantified fingerprint method (SQFM), in which both qualitative and quantitative evaluation parameters were scientifically improved on the basis of ‘Similarity’ conventionally used and ‘simple quantified ratio fingerprint method’ previously proposed. In addition, the HM chemical profiling of WBZT samples was obtained by ultra performance liquid chromatography coupled with quadrupole time-of-flight mass spectrometry (UPLC-Q-TOF-MS) in positive ion mode, providing an important chemical structure foundation for further bioactivity and quality control studies. The relationship between high performance liquid chromatography (HPLC) fingerprints and antioxidant activities of WBZT samples was established using the partial least squares regression (PLSR) method, which offers a robust predictive ability of the antioxidant activities of WBZT samples. This study demonstrates that integrated MIR and UV spectroscopic fingerprints combined with antioxidant activities can monitor TCM/HM quality consistency rapidly, effectively and comprehensively.

---

1. Introduction

The therapeutic effects of TCM/HM derived from botanical materials are principally based on synergistic efficacies of multi-constituents toward multi-targets, in contrast to modern pharmacology and synthetic drugs that often focus on a single chemical component.¹ Therefore, the method of employing a limited number of compounds to assess the quality of TCM/HM consisting of hundreds of chemically different constituents is faced with severe challenges.² Fingerprinting techniques systemically characterizing complicated TCM/HM, as a rational and powerful method, have been widely accepted and adopted for quality control by many major authorities.³

Nowadays, although chromatographic fingerprints are frequently used for TCM/HM, the approaches based on chromatographic separation techniques combined with detection techniques such as UV, diode array detector (DAD), evaporative light scattering (ELSD) and mass spectrometry (MS),^1,4–7 can only provide information about compounds with corresponding chromatographic and detection properties, and cannot give us information about all the chemical components in a TCM/HM system, which restricts the comprehensive quality control of multi-constituents acting synergistically. In this situation, spectroscopic techniques including UV, near-infrared (NIR) and MIR spectroscopy have been proposed to make up for this limitation for guaranteeing the TCM/HM safety and efficacy in clinical applications. Recently, owing to clear analytical advantages such as, simple sample preparation, short measurement time and low experiment costs, IR/UV spectroscopic fingerprinting in combination with chemometrics have been used for TCM/HM quality control.^8–11

MIR spectra based on molecular vibration and rotation¹² give not only characteristic absorption bands at specific, narrow frequency ranges due to the presence of specific chemical groups,¹³ but also fingerprint absorption bands arising from chemical bonds such as C–C, C–N, C–O, C–H, N–H and O–H widely present in natural products in the 1300–400 cm⁻¹ region.¹⁴ This provides us with a great deal of structural information about all compounds present in TCM/HM, especially saturated bonds. UV spectra of the chromophores and auxochromes in compounds mainly reveal information about unsaturated bonds of components with conjugated or aromatic systems, such as flavonoids, lignans, coumarins and polyphenolics.¹⁵ However, the information about saturated bonds of components lacking chromophores, such as terpenoids, can be compensated for by using MIR spectra. Since UV and MIR spectra qualitatively and quantitatively reflect the features of saturated and unsaturated bonds of components,^8,14 the combination of these techniques can offer an appealing and effective method for comprehensive TCM/HM quality control.

TCM/HM components are chemical substances exhibiting a variety of therapeutic efficacies and,¹⁶ thus, it is necessary to carry out the chemical profiling of TCM/HM for detailed bioactivity and quality control studies. Although the complexity of natural constituents in TCM/HM presents great challenges in terms of their separation and identification, UPLC-Q-TOF-MS with a high separation efficiency and mass resolution provide us with a powerful method to investigate TCM/HM compounds.

Antioxidants can reduce the risk of a variety of conditions such as cancer, cardiovascular disease and peptic ulcer, and this is closely related to the harmful effects of free radicals.^17,18 Accordingly, research concerning natural antioxidants and their antioxidant properties is attracting increasing attention,¹⁹ and this has prompted us to investigate TCM/HM antioxidant activities and correlate them with chromatographic fingerprints.

WBZT has been used internationally for over 20 years as a popular patent drug for the treatment of gastric ulcers, and has been officially documented in the Chinese Pharmacopeia since 1995. It is composed of medicinal herbs, Extract Licorice, Cortex Frangulae, Fructus Foeniculi and synthetic drugs, bismuth aluminate, heavy magnesium carbonate, sodium hydrogen carbonate in a mass ratio of 30 : 2.5 : 1 : 20 : 40 : 20.²⁰ Currently, analytical methods available for HM quality control in WBZT are primarily based on quantifying a single bioactive component using HPLC^20–22 as well as few chromatographic fingerprinting.²³ In addition, there are no published studies on HM chemical profiling in WBZT.

In the present study, MIR and UV spectroscopic fingerprinting were developed for characterization of the quality of 27 batches of WBZT using FTMIR spectrometry and FIA. In order to obtain comprehensive information about HM constituents in WBZT, an integrated assessment method based on these two fingerprinting techniques was set up. In fingerprint assessments, SQFM was established for scientific quality analysis of TCM/HM from qualitative and quantitative perspectives, by which the quality grades of the 27 samples from the same manufacturer were well differentiated. In addition, a correlation analysis between HPLC fingerprints and antioxidant activities of WBZT was performed using the PLSR method. A UPLC-Q-TOF-MS method was developed to systematically investigate HM constituents in WBZT in positive ion mode, which provided information about chemical substances for bioactivity and quality control studies. It has been demonstrated that integrated UV and MIR spectroscopic fingerprints combined with antioxidant activities offers a comprehensive and efficient method for HM quality control in WBZT.

2. Theory of SQFM

We assume that is the sample fingerprint vector (SFPV) and is the reference fingerprint vector (RFPV), where x_i and y_i are the peak area of the ith ingredient in the sample and reference fingerprints, respectively. The ‘Similarity’ (S_F) between these two fingerprints can be expressed by eqn (1), as recommended by the State Food and Drug Administration (SFDA) of China.²⁴ Compared with S_F, the ratio similarity (S′_F) shown in eqn (2) effectively eliminates the effect of major peaks masking minor ones,²⁵ however, the action of minor peaks is also exaggerated by taking into consideration all the peaks in equal weights. Thus, an improved qualitative parameter, termed macro qualitative similarity (S_m), is shown by eqn (3), which compensates for the above limitations and thus can accurately describe the resemblance in terms of the distribution and number of fingerprints between SFPV and RFPV.

On the other hand, for more scientifically quantitative similarity assessments, macro quantitative similarity (P_m), an improved quantitative parameter, is defined by eqn (4) and is used to examine the total content similarity for all ingredients between SFPV and RFPV. It should be noted that P_m is revised by a mass factor (f_wi), which is defined as the weight ratio (m_RFP/m_i) of RFPV and the ith SFPV. The fingerprint variation coefficient (α), as shown in eqn (5), is also a qualitative parameter which reflects the fingerprint dissimilarity between SFPV and RFPV. Accordingly, the quality evaluation method in terms of S_m, P_m and α is given the name SQFM, by which TCM/HM quality can be classified into 8 grades (Table 1). In the evaluation system, grade 1 represents the highest quality and grade 8 the lowest one, and the grades in the range 1–5 are recognized as qualified.

Table 1 The TCM/HM quality grades classified by SQFM

Parameter	1	2	3	4	5	6	7	8
Sm ≥	0.95	0.9	0.85	0.8	0.7	0.6	0.5	Sm < 0.5
Pm% ∈	95–105	90–110	80–120	75–125	70–130	60–140	50–150	0–∞
α ≤	0.05	0.1	0.15	0.2	0.3	0.4	0.5	α > 0.50
Quality grade	1	2	3	4	5	6	7	8

---

3. Experimental

3.1 Chemicals and reagents

A total of 27 batches of WBZT samples (labeled S1–S27) were manufactured by Tonglian Pharmaceutical Co., Ltd. (Shenyang, China). Individual herbs, including Extract Licorice, Cortex Frangulae, and Fructus Foeniculi, were obtained from Tonglian Pharmaceutical Co., Ltd. (Shenyang, China). Seven reference standards, liquiritin, isoliquiritin, liquiritigenin, isoliquiritigenin, liquiritin apioside, isoliquiritin apioside and glycyrrhizic acid mono-ammonium salt were purchased from Shanghai Winherb Science and Technology Inc. (Shanghai, China). In the UPLC-Q-TOF-MS analysis, the used formic acid and acetonitrile were HPLC-grade (Fisher Scientific, USA). In the HPLC and UV spectroscopic analyses, HPLC-grade methanol, acetonitrile and glacial acetic acid were purchased from Yuwang Industry (Shandong, China). Unless otherwise noted, all reagents were analytical grade and supplied by Yuwang Industry (Shandong, China). Deionized water was purified using a Milli-Q system (Millipore, USA).

3.2 Sample preparation

To obtain WBZT solutions, each test sample was pulverized into powder with a mortar, and then each powder sample (1.2 g) was accurately weighed and extracted by refluxing three times with 25 mL methanol for 30 min. The combined filtrates were concentrated in a rotary evaporator under vacuum, and the sample solution for HPLC, UPLC-Q-TOF-MS and antioxidant activity analyses was prepared in methanol in a 25 mL volumetric flask. Then, this solution was further diluted 10-fold with methanol to provide a solution for UV analysis. All the solutions were stored at 4 °C and passed through a 0.22 μm Millipore filter prior to analyses.

To obtain WBZT tablets, each test sample was ground into powder, and then the powder (1 mg) and KBr powder (120 mg) were, respectively, accurately weighed. After thorough mixing with a mortar, they were pressed into a sample tablet for FTMIR spectroscopic analysis. To ensure that moisture was not an interfering factor, all samples were dried in an oven of 60 °C, until the weight variation was less than 0.1%.

3.3 Data analysis

All the UPLC-Q-TOF-MS data were processed using MassLynx 4.1 software (Waters, USA) and all the fingerprint evaluations were performed using independently developed software ‘Digitized Evaluation System for Super-Information Characteristics of TCM/HM Chromatographic Fingerprints 4.0’ (Software certificate no. 0407573, China). In addition, SPSS 19.0 and SIMCA 13.0 were also used for data analyses.

3.4 Experimental conditions

3.4.1 UPLC-Q-TOF-MS conditions. An Acquity UPLC system (Waters, USA) coupled with a Micromass-Q-TOF Premier mass spectrometer (Waters, USA) was used to carry out HM chemical profiling in WBZT samples. Chromatographic separation was performed on an Acquity UPLC column (HSS C₁₈ 100 × 2.1 mm, 1.8 μm) at 35 °C. Samples were maintained at 15 °C in the auto sampler. Water–formic acid (A; 100 : 0.1, v/v) and acetonitrile (B) were used as the mobile phase. The gradient elution program with a flow rate of 0.2 mL min⁻¹ was set as follows: 5–25% B at 0–2 min, 25–60% B at 2–15 min, 60–90% B at 15–18 min, 90% B at 18–20 min. The sample injection volume was 2 μL. The parameters of the mass spectrometer by the electrospray ionization (ESI) source operating in positive ion mode were as follows: ion source temperature 130 °C, desolvation gas temperature 400 °C, cone gas flow 50 L h⁻¹, desolvation gas flow 700 L h⁻¹, capillary voltage 3.0 kV, cone voltage 35 V, and extraction cone voltage 4.0 V. Full-scan MS data were collected from 50 to 1500 Da with a collision energy of 20–40 eV.

3.4.2 HPLC chromatographic conditions. Agilent 1100 HPLC series equipped with a DAD, a quaternary pump, an on-line degasser and an autosampler (Agilent, USA) was used to analyze the HPLC fingerprints of WBZT samples. Separation was achieved on a Century SIL C₁₈ BDS column (250 × 4.6 mm, 5.0 μm) (Century, China) at 35 °C. The mobile phase consisting of water–glacial acetic acid (A; 100 : 1, v/v) and acetonitrile–glacial acetic acid (B; 100 : 1, v/v) using a gradient elution of 5–8% B at 0–8 min, 8–25% B at 8–25 min, 25–30% B at 25–45 min, 30–50% B at 45–65 min, 50–70% B at 65–85 min and 70–77% B at 85–90 min. The flow rate and sample injection volume was 1.0 mL min⁻¹ and 5 μL, respectively. The detection wavelength was set at 250 nm.

3.4.3 UV spectroscopic conditions. UV spectra were recorded on an Agilent 1100 HPLC system equipped with a DAD (Agilent, USA) over the range 190–400 nm by replacing the chromatographic column with a hollow polytetrafluoro-ethylene (PTFE) pipe (6500 × 0.12 mm) and using methanol–water (55 : 45, v/v) as a carrier, with FIA being used as a analytical principle as shown in Fig. 1. The remaining parameters were as follows: temperature of the PTFE tube 35 °C, flow rate 0.4 mL min⁻¹, sample injection volume 1 μL, wavelength interval 1 nm, and slit width 1 nm.

Fig. 1 The analytical principle plot for sample UV spectra.

3.4.4 MIR spectroscopic conditions. A Bruker IFS-55 FTMIR spectrometer equipped with a DTGS detector (Bruker, Germany) was used. MIR spectra were collected by accumulating 20 scans each over the range 4000–400 cm⁻¹ with a resolution of 8 cm⁻¹.

3.4.5 Antioxidant activity conditions. The classical Fenton reaction was employed to investigate the antioxidant activity of WBZT extracts. The reaction between ferrous iron and hydrogen peroxide produces hydroxyl radicals (OH˙) with outstanding oxidant ability, which attack the carbon–carbon double bond causing color fading of crystal violet. However the color fading can be inhibited due to the addition of antioxidant agent in the system. Thus, OH˙ scavenging ratio (SR) of a sample was indirectly deduced from the absorbance changes of crystal violet solutions.

The reaction solution was prepared by adding successively 6.0 mL of 1.0 mmol L⁻¹ FeSO₄·10H₂O solution, 1.0 mL of 0.4 mmol L⁻¹ crystal violet solution, 0.09 mL of potassium hydrogen phthalate buffer solution (pH = 4.0) and 1.1 mL of 1% H₂O₂ solution to a 25 mL volumetric flask and diluting to the volume with water, and then absorbance (A_b) was measured at 588 nm. The blank sample solution was identical to reaction solution except without added H₂O₂, and absorbance (A₀) was measured. The negative control samples were prepared by adding different volumes of methanol (0.10 mL, 0.20 mL, 0.25 mL, 0.50 mL, 0.75 mL) to blank sample solution, and absorbance values (A_B) were measured. The positive control samples were identical to negative control samples except replacing methanol with corresponding volumes of sample solution as described in Section 3.2.

These solutions were shaken and stored at 4 °C for 30 min, and then absorbance values (A_S) were measured. The SR of OH˙ was calculated using eqn (6):

The sample concentrations in positive control samples were calculated according to added sample volumes, and calibration curves were obtained by plotting SR values vs. sample concentrations. Finally, the sample effective concentration that scavenged 50% of OH˙ (EC₅₀) could be calculated by interpolation.

4. Results and discussion

4.1 UPLC-Q-TOF-MS analysis of WBZT samples

Under the UPLC-MS conditions described in Section 3.4.1, reference standards, a WBZT sample and three individual herbal extracts were analyzed in positive ion mode. The base peak ion chromatogram of the WBZT sample is shown in Fig. 2A. A total of 20 components, including flavonoids, saponins and coumarin, were identified (Table S1†), and their sources were confirmed by comparing the base peak ion chromatograms of the WBZT sample and its three individual herbal extracts. For reference standards, their structures (Fig. 3A) were unambiguously identified by comparing corresponding retention time, accurate mass and MS/MS data. For those components without reference standards, their structures (Fig. 3B) were tentatively characterized based on accurate mass measurements within a 5 ppm error and tandem MS behavior and related literature data.²⁶

Fig. 2 Base peak chromatogram of WBZT samples in positive ion mode (A), the MS/MS spectra and proposed fragmentation pathways of liquiritin apioside (B), isoliquiritin apioside (C), glycyrrhizic acid (D) and inflacoumarin A (E) in positive ion mode.

Fig. 3 The chemical structures of reference compounds (A) and constituents tentatively identified by UPLC-Q-TOF-MS (B) in WBZT.

4.1.1 Identification of flavonoids in WBZT samples. Flavonoids might be present in WBZT samples as they are major components of licorice. Peaks 1, 2, 3, 4, 7 and 9 were unambiguously attributed to liquiritin apioside, liquiritin, isoliquiritin apioside, isoliquiritin, liquiritigenin and isoliquiritigenin by comparison with authentic compounds. Liquiritin apioside, a typical flavonoid in WBZT samples, was used to characterize the fragmentation pathways (Fig. 2B). An [M + H]⁺ at m/z 551.1754 was found in positive ion mode, which could successively lose one apiosyl group (Api, 132 Da) and one glucosyl group (Glc, 162 Da) to form fragments at m/z 419.1326 and m/z 257.0805, and then retro Diels–Alder (RDA) cleavage of the ion at m/z 257.0805 would gave rise to fragments at m/z 137.0235 and m/z 119.0490, and the loss of C₆H₆O₂ (110 Da) from the ion at m/z 257.0805 would generate a fragment at m/z 147.0438. According to these fragmentation patterns, peaks 6 and 19 were identified as ononin and licoflavone A, respectively.

Isoliquiritin apioside, an isomer of liquiritin apioside, was used to characterize the fragmentation pathways (Fig. 2C). Since it showed the same MS and MS² pattern as liquiritin apioside, its fragmentation pathways was similar to that of liquiritin apioside. There was an [M + H]⁺ at m/z 551.1754 in positive ion mode, which could successively lose one apiosyl group (132 Da) and one glucosyl group (162 Da) to form fragments at m/z 419.1326 and m/z 257.0805, and then cleavage of the ion at m/z 257.0805 would give rise to fragments at m/z 137.0235 and m/z 119.0490, and the loss of C₆H₆O₂ (110 Da) from the ion at m/z 257.0805 would generate a fragment at m/z 147.0438. Based on these cleavage patterns, peaks 5 and 16 were identified as neoisoliquiritin and inflacoumarin A, respectively.

4.1.2 Identification of saponins in WBZT samples. Saponins might be present in WBZT samples as they are important constituents of licorice. Peak 10 was unambiguously assigned to glycyrrhizic acid by comparison with the authentic compound.

Glycyrrhizic acid, an important saponin in WBZT, was used to characterize the fragmentation pathways (Fig. 2D). There was an [M + H]⁺ at m/z 823.4111 in positive ion mode, and fragments at m/z 647.3806, m/z 471.3479 and m/z 453.3375 corresponding, respectively, to [M + H–GlcA]⁺, [M + H–2GlcA]⁺ and [M + H–2GlcA–H₂O]⁺. According to these fragmentation patterns, peaks 8, 11, 12, 14, 15, 17, 19 and 20 were identified as licorice-saponin G₂, licorice-saponin H₂, licorice-saponin K₂, licorice-saponin B₂, 18-α-glycyrrhetinic acid 3-O-glucuronide, 18-β-glycyrrhetinic acid 3-O-glucuronide, 18-α-glycyrrhetinic acid and 18-β-glycyrrhetinic acid, respectively.

4.1.3 Identification of coumarins in WBZT samples. Coumarins might be present in WBZT samples as they are ingredients of licorice. Peak 13 was tentatively presumed to be inflacoumarin A. The proposed fragmentation pathway of peak 13 in positive ion mode is illustrated in Fig. 2E. There was an [M + H]⁺ at m/z 323.1290 in positive ion mode and fragments at m/z 267.0660, m/z 239.0706 and m/z 121.0291 corresponding to [M + H–C₄H₈]⁺, [M + H–C₄H₈–CO]⁺ and [M + H–C₄H₈–CO–C₇H₂O₂]⁺, respectively.

According to above studies, it was concluded that HM components in WBZT samples mainly come from principal individual herb, i.e. Extract Licorice and consist of flavonoids, saponins and coumarin.

This is in agreement with previous studies of licorice ingredients.²⁶ This information about the structures of HM components in WBZT will help in developing research strategies for bioactivity and quality control studies.

4.2 MIR/UV spectroscopic fingerprint analyses

4.2.1 Methodology validation of fingerprint analysis. All the samples as described in Section 3.2 were used to perform the following experiments. Method repeatability was assessed by analyzing six independently prepared samples (S1) using the same experimental procedure. Instrument precision was determined by analyzing the same sample (S1) six times consecutively. UV sample stability was examined by analyzing the same sample (S1) stored at 4 °C for 0, 3 and 6 hours, while MIR sample stability was examined by analyzing the same sample (S1) stored in a dryer for 1, 2 and 3 hours.

In the UV fingerprint analysis, unseparated chromatograms at 250 nm and UV spectra of samples in the region 190–400 nm were recorded, and a typical 3D chromatogram plot of a WBZT sample is shown in Fig. 4. The retention time (RT) and peak area (PA) of a sample with an unseparated chromatogram at 250 nm were used to estimate the repeatability, precision and stability, and the obtained results showed that, for repeatability, the relative standard deviations (RSD) of RT and PA were less than 0.4 and 1.2%, respectively; for precision, the obtained values did not exceed 0.4 and 1.0%, respectively; for stability, the obtained values were less than 0.3 and 1.0%, respectively.

Fig. 4 3D chromatogram plot (min × nm × absorbance) for a WBZT sample.

In the MIR fingerprint analysis, MIR spectra of samples were recorded in the region 4000–400 cm⁻¹. The quantitative evaluation parameter (P_m) of sample fingerprints was used to estimate the repeatability, precision and stability, and the obtained results showed that, for repeatability, the RSD of P_m was less than 0.2%; for precision, the obtained value did not exceed 0.1% and, for stability, the obtained value was less than 0.4%. Thus, these results demonstrated that the developed UV and MIR spectroscopic methods met the fingerprint analysis requirements for WBZT samples.

4.2.2 Analysis of MIR/UV spectra. Typical MIR/UV spectra for 27 WBZT samples, as shown in Fig. 5, were obtained in the region 4000–400 cm⁻¹/190–400 nm using the methods described in Sections 3.4.4/3.4.3 and 3.2. For complex TCM/HM systems, the MIR spectra clearly show an overlap of the absorption spectra of various components. Based on structural analyses obtained by UPLC-Q-TOF-MS, the main MIR peak positions of the WBZT samples were preliminarily assigned as follows. The peaks at 3446 and 1423 cm⁻¹ belong to stretching and in-plane bending vibrations of O–H groups, respectively, which are widely present in flavonoids, saponins and coumarin. The peaks at 2922 and 2851 cm⁻¹ are, respectively, asymmetric and symmetric C–H stretching vibrations of CH₂ and CH₃ groups mainly present in saponins, while the peak at 1617 cm⁻¹ corresponds to the C=O stretching vibration in flavonoids, coumarin and certain saponins. The peak at 1485 cm⁻¹ is the C=C stretching vibration of the aromatic skeleton present in flavonoids and coumarin. The peak at 1020 cm⁻¹ is attributed to the C–OH stretching vibration of glycosides, and the peaks at 884, 804, 675 and 594 cm⁻¹ belong to the C–C stretching vibration and O–H out-of-plane bending vibration, which are mainly present in flavonoids and saponins.

Fig. 5 Typical MIR spectra (A) and UV spectra (B) for 27 WBZT samples.

Regarding the UV spectra, the absorption bands at around 200 nm and 270 nm are likely to be mainly due to aromatic rings, involving π–π* electronic transitions of aromatic rings in flavonoids and coumarin. Moreover, the absorption bands at nearby 270 nm and 320 nm might be, respectively, attributed to π–π* and n–π* electronic transitions of conjugated systems involving aromatic rings and C=O in flavonoids.

4.2.3 Analysis of MIR fingerprints. The parameters S_m, α and P_m (as listed in Table 2) for 27 samples were calculated by using all MIR data points (about 4600) in the region 4000–400 cm⁻¹. In order to investigate the effect of these data points on above three parameters, analysis of variance (ANOVA) of MIR fingerprints for 27 samples in the region 4000–400 cm⁻¹ was carried out. From the analytical results shown in Fig. 6A, it was clear that the 11 fingerprint peaks at the marked wave numbers (3446, 2922, 2851, 1617, 1485, 1423, 1020, 884, 804, 675 and 594 cm⁻¹) had greater variance maximum values, indicating that they were the most important factors for computing three parameters (S_m, α and P_m) and subsequent assessment of the quality grades of drug samples.

Table 2 The evaluation results for 27 WBZT samples based on MIR fingerprints, UV fingerprints and integrated method by SQFM

Sample	MIR				UV				Integrated
	Sm	Pm	α	Grade	Sm	Pm	α	Grade	Sm	Pm	α	Grade
S1	0.999	113.2	0.002	3	0.994	86.7	0.044	3	0.997	100.0	0.023	1
S2	0.997	89.7	0.005	3	0.998	95.7	0.006	1	0.998	92.7	0.006	2
S3	0.996	74.3	0.001	5	0.998	99.8	0.004	1	0.997	87.1	0.003	3
S4	0.999	80.7	0.002	3	0.998	101.7	0.004	1	0.999	91.2	0.003	2
S5	0.999	79.8	0.000	4	0.999	96.4	0.010	1	0.999	88.1	0.005	3
S6	0.996	72.3	0.010	5	0.999	93.9	0.008	2	0.998	83.1	0.009	3
S7	0.999	115.5	0.004	3	0.998	90.1	0.001	2	0.999	102.8	0.003	1
S8	0.998	118.7	0.005	3	0.999	94.5	0.001	2	0.999	106.6	0.003	2
S9	0.994	96.5	0.013	1	0.999	102.1	0.008	1	0.997	99.3	0.011	1
S10	0.998	117.9	0.004	3	0.999	101.0	0.007	1	0.999	109.5	0.006	2
S11	0.997	113.6	0.006	3	0.999	93.1	0.012	2	0.998	103.4	0.009	1
S12	0.996	91.0	0.009	2	0.999	96.8	0.015	1	0.998	93.9	0.012	2
S13	1.000	123.2	0.002	4	0.999	102.1	0.012	1	1.000	112.7	0.007	3
S14	1.000	114.2	0.001	3	0.988	107.8	0.063	2	0.994	111.0	0.032	3
S15	0.995	167.4	0.004	8	1.000	101.2	0.002	1	0.998	134.3	0.003	6
S16	1.000	123.4	0.001	4	0.999	101.5	0.009	1	1.000	112.5	0.005	3
S17	0.994	83.4	0.017	3	0.997	121.1	0.022	4	0.996	102.3	0.020	1
S18	0.995	93.9	0.012	2	0.998	96.6	0.035	1	0.997	95.3	0.024	1
S19	0.992	94.6	0.021	2	0.702	38.4	0.350	8	0.847	66.5	0.186	6
S20	0.989	87.8	0.004	3	0.996	95.9	0.046	1	0.993	91.9	0.025	2
S21	0.997	80.7	0.013	3	0.999	113.6	0.001	3	0.998	97.2	0.007	1
S22	0.993	96.5	0.017	1	0.999	92.1	0.027	2	0.996	94.3	0.022	2
S23	0.990	107.8	0.000	2	0.703	37.0	0.341	8	0.847	72.4	0.171	5
S24	0.997	97.1	0.014	1	0.994	122.5	0.025	4	0.996	109.8	0.020	2
S25	0.991	98.2	0.008	1	0.704	53.8	0.288	7	0.848	76.0	0.148	4
S26	0.995	113.2	0.005	3	0.999	88.9	0.018	3	0.997	101.1	0.012	1
S27	0.995	136.6	0.001	6	0.999	102.8	0.009	1	0.997	119.7	0.005	3

---

Fig. 6 ANOVA plot of MIR fingerprints for 27 WBZT samples (A), PCA score plot (B) and loading plot (C) for 27 WBZT samples on the basis of the transmittance values of the 11 marked MIR fingerprints.

In order to carry out a detailed investigation of the distinguishing ability of the transmittance values of the marked 11 MIR fingerprints, a principal component analysis (PCA), a well-known chemometrics method,²⁷ was performed using software SIMCA 13.0 and the transmittance values of the 11 fingerprint peaks were used as input data to construct two-dimensional matrices (27 × 11), with 27 and 11 representing the sample number and wave number type, respectively.

A two-component PCA model was obtained which cumulatively accounted for 88.6% of the variation. The total variance explained for the first principal component was 54.5% and that for the second principal component was 34.1% (Fig. 6B). In the loading scatter plot (Fig. 6C), the coordinate positions of the wave numbers of 11 fingerprint peaks showed their corresponding weights in principal components and, so, the farther away from coordinates-origin the wave number, the greater the correlation between the principal component and the wave number. By observing score plot and loading scatter plot, it was found that 1423, 884, 804, 1485 and 594 cm⁻¹ had a greater correlativity with PC1, the same as 3446 cm⁻¹ with PC2, while 2922, 2851, 1020, 1617 and 675 cm⁻¹ had less influence on PC1 and PC2. In the score plot, 27 drug samples from the same manufacturer could be clearly divided into three clusters marked as group 1, 2 and 3, respectively. Because the samples in group 3 had the greatest positive correlativity with 3446 cm⁻¹, the absorbance values at 3446 cm⁻¹ (in the range 0.17726–0.4005) in group 3 samples were all higher than those in group 1 and group 2 (in the ranges 0.04395–0.13306 and 0.03136–0.12611, respectively). Also, the samples in group 1 exhibited the highest positive correlativity with 1423, 884, 804, 1485 and 594 cm⁻¹, and the absorbance values at 1423, 884, 804, 1485 and 594 cm⁻¹ (in the ranges 0.13685–0.30329, 0.79674–0.84764, 0.72682–0.80244, 0.21065–0.36264 and 0.69749–0.76675, respectively) in the group 1 samples were all higher than the corresponding values (in the ranges 0.03645–0.07219, 0.77155–0.78375, 0.68577–0.7061, 0.11128–0.15665 and 0.63383–0.68035, respectively) in the group 2 samples. Consequently, 15 products were clustered in group 3, mainly because of their higher absorbance values at 3446 cm⁻¹, while the difference between group 1 and group 2 was mainly due to the absorbance values at 1423, 884, 804, 1485 and 594 cm⁻¹. Therefore, we were able to conclude that the two-component PCA model based on the marked 11 MIR fingerprints exhibited a greater ability to discriminate among the 27 batches of samples from the same manufacturer.

4.2.4 Sample quality evaluation based on MIR/UV fingerprints. MIR/UV reference fingerprint profiles (RFP) were constructed by taking the average of 27 corresponding spectra. As shown in Fig. 5, the MIR/UV fingerprints for the 27 samples were very similar and, thus, the subtle spectral differences need to be described qualitatively and quantitatively by means of SQFM. In order to evaluate the quality grades, the S_m, P_m and α values of the MIR/UV spectroscopic fingerprints of the 27 samples were computed by importing all MIR/UV spectral data in the region 4000–400 cm⁻¹/190–400 nm into the TCM/HM fingerprint software, and the obtained results are summarized in Table 2.

For the MIR fingerprints, the S_m and α values for the 27 samples were not below 0.989 and above 0.021, respectively, demonstrating that all samples had similar chemical compositions. Although all samples should have the highest quality based on S_m and α from a qualitative perspective, in fact, only 4 samples (S9, S22, S24 and S25) were judged as grade 1, and the remaining ones were in range grade 2–8 (grade 2: S12, S18, S19 and S23; grade 3: S1, S2, S4, S7, S8, S10, S11, S14, S17, S20, S21 and S26; grade 4: S5, S13 and S16; grade 5: S3 and S6; grade 6: S27; grade 8: S15.) in the combination of P_m from a quantitative perspective. Similarly, for UV fingerprints, since the qualitative parameters S_m and α of the drug samples were, respectively, not below 0.988 and above 0.046, except for S14, S19, S23 and S25, the 23 samples should be grade 1. However, in fact, only 13 samples (S2, S3, S4, S5, S9, S10, S12 S13, S15, S16, S18, S20 and S27) were judged as grade 1, and the remaining 10 ones were in the range grade 2–8 in combination of the quantitative parameter P_m (grade 2: S6, S7, S8, S11, S14 and S22; grade 3: S1, S21 and S26; grade 4: S17 and S24; grade 7: S25; grade 8: S19 and S23.). The above results indicated that the qualitative evaluation should be performed first and then further quantitative assessment should not be ignored. P_m, as a parameter describing the overall ingredient content in samples, has a great potential to be associated with medicinal efficacy in clinical situations.

4.2.5 Integrated evaluation based on both MIR and UV fingerprints. It is noteworthy that the quality grades evaluated by MIR and UV spectroscopic fingerprints for the same samples may be different, and this can be attributed to two different analytical principles, and the MIR and UV fingerprints mainly reflect the features of saturated and unsaturated chemical bonds in compounds, respectively.

To synthesize two types of bond features and achieve comprehensive quality control for HM components in WBZT, an integrated assessment method based on UV and MIR spectroscopic fingerprints was set up in equal weights. Three integrated assessment parameters, S_m, P_m and α values, were calculated according to eqn (7)–(9), where S_m-IR, P_m-IR and α_IR as well as S_m-UV, P_m-UV and α_UV represent three parameters of MIR and UV spectroscopic fingerprints, respectively. The integrated quality grades (Table 2) could be assessed according to TCM/HM quality grades classified by SQFM (Table 1).

From the integrated results of the 27 samples (Table 2), we found that: because acceptable S_m and α are, respectively, not below 0.7 and above 0.3, while acceptable P_m values are set in the range 70.0–150.0%, S15 and S19 had unqualified integrated grades (grade 6), while the remaining 25 samples had qualified ones (grade 1–5, grade 1: S1, S7, S9, S11, S17, S18, S21 and S26; grade 2: S2, S4, S8, S10, S12, S20, S22 and S24; grade 3: S3, S5, S6, S13, S14, S16 and S27; grade 4: S25; grade 5: S23). Compared with the quality grades assessed by the individual MIR/UV fingerprint method, the integrated results exhibited some fluctuations and even greater differences. For example, S15 and S19 had unqualified integrated quality (grade 6), despite their having better individual ones (grade 1 in the UV method and grade 2 in the MIR method, respectively.), which illustrated that our integrated assessment strategy was very comprehensive and necessary to avoid a bias caused by a single fingerprint method. In addition, the integrated quality differences among the 27 samples might be due to the variability in the raw herbs associated with a wide range of factors or variability in the manufacturing processes.²⁸

4.3 Correlation analysis between HPLC fingerprints and antioxidant activities in vitro

According to the chemical profiling of HM components in WBZT, it appeared that drug samples possess excellent antioxidant abilities due to the presence of many flavonoids. Thus, sample antioxidant activities in vitro, using EC₅₀ as the measurement indicator, were investigated using the method described in Section 3.4.5, and their HPLC fingerprints were measured under the conditions described in Section 3.4.2. Typical HPLC chromatograms with 39 co-possessing fingerprints for 27 WBZT samples are shown in Fig. 7A and the RFP was generated from the mean of 27 sample chromatograms. A correlation analysis between the EC₅₀ values and HPLC fingerprints was performed by the PLSR method, in which the EC₅₀ values and areas of 39 co-possessing fingerprints for 27 samples were used as Y variables and X variables, respectively.

Fig. 7 Typical HPLC chromatograms for 27 WBZT samples at 250 nm (A), measured versus predicted values of antioxidant activities of samples used for the calibration model (B), standardized regression coefficient plot based on Hotelling's T² (0.05) for PLSR analysis (C).

The PLSR model was validated by means of full cross-validation for rationality, as shown by the two principal components score plot in PLSR analysis (Fig. S1†), two samples, i.e. S17 and S25, were identified as singular points, which should be removed when final mathematical modeling is carried out. After omitting S17 and S25, the remaining 25 samples were randomly divided into two groups of calibration and validation sets (Table S2†), which were, respectively, used to describe the relationship between the two groups of variables and evaluate the predictive ability of the established model. A calibration model with two latent variables was chosen in terms of the cross-validation, reaching an explained variance (R²) of 93.3% for Y variables, a predictive ability (Q²) of 86.2% and a root mean square error of estimation value (RMSEE) of 0.0172 (Fig. 7B), indicating that the performance of the calibration model was excellent. As shown by the standardized coefficients plot of the calibration model (Fig. 7C), 21 peaks (1, 5, 7, 9, 10, 11, 12, 14, 18, 19, 22, 25, 26, 27, 29, 30, 31, 32, 33, 36 and 39) out of 39 fingerprints in the HPLC chromatogram were negatively correlated, while the remaining peaks were positively correlated with EC₅₀. The peak 3, 7, 9, 10, 14 and 18 in HPLC fingerprints were identified as liquiritin apioside, liquiritin, isoliquiritin apioside, isoliquiritin, liquiritigenin and isoliquiritigenin by comparing the retention times and on-line UV spectra with reference standards.

In order to examine the predictive ability of the established model, the co-possessing fingerprint areas of samples in the validation set were substituted into the model, and the obtained predicted values of the EC₅₀ values are listed in (Table S2†). A root mean square error of the predicted values (RMSEP) of 0.0211 was obtained, indicating that the calibration model has a robust predictive ability. No significant difference was observed between the measured and predicted EC₅₀ values of 25 samples except for S17 and S25.

5. Conclusions

In the present study, the HM chemical profiling in WBZT was performed by UPLC-Q-TOF-MS. As a result, a total of 20 compounds including 9 flavonoids, 10 saponins and 1 coumarin were identified or tentatively characterized, which guided the direction of studies with respect to bioactivity and quality control strategies. MIR and UV spectroscopic fingerprints were integrated in equal weights to reflect overall characterizations of HM constituents in WBZT from chemical bonds perspective. SQFM with qualitative and quantitative assessment advantages was established for TCM/HM quality differentiation, which not only overcame the defect of ‘Similarity’ lack of quantitative judgment, but also improved scientifically three evaluation parameters on the basis of ‘simple quantified ratio fingerprint method’. According to integrated MIR and UV spectroscopic fingerprints as well as SQFM, the quality consistency of 27 batches of WBZT samples from the same manufacturer were well differentiated. Two out of 27 drug samples (S15 and S19) were judged as unqualified quality grades (grade 6), and the rest samples were judged as qualified ones (grade 1–5; grade 1: S1, S7, S9, S11, S17, S18, S21 and S26; grade 2: S2, S4, S8, S10, S12, S20, S22 and S24; grade 3: S3, S5, S6, S13, S14, S16 and S27; grade 4: S25; grade 5: S23). Moreover, the correlation analysis between HPLC fingerprints and antioxidant activities of samples was conducted using PLSR, and the established model had robust predictive ability, which could provide important supplemental bioactivity information for WBZT quality control. This study offered a scientific, rapid and comprehensive analytical strategy for HM quality control which would have a great potential to play an important role in WBZT practical production.

Acknowledgements

This work was supported by National Natural Science Foundations of China (81403094 and 81560695) and the general education department of Scientific Research Project of Liaoning Province (L2013393).

References

1. P. S. Xie, S. B. Chen, Y. Z. Liang, X. H. Wang, R. T. Tian and R. Upton, J. Chromatogr. A, 2006, 1112, 171–180.
2. J. Liang, W. Wu, G. Sun, D. Wang, J. Hou, W. Yang, B. Jiang, X. Liu and D. Guo, J. Chromatogr. A, 2013, 1294, 58–69.
3. M. Goodarzi, P. J. Russell and Y. Vander Heyden, Anal. Chim. Acta, 2013, 804, 16–28.
4. L. Peng, Y. Wang, H. Zhu and Q. Chen, Food Chem., 2011, 125, 1064–1071.
5. Y. Ding, E. Q. Wu, C. Liang, J. Chen, M. N. Tran, T. C. H. Hong, Y. Jang, K. L. Park, K. Bae, Y. H. Kim and J. S. Kang, Food Chem., 2011, 127, 755–760.
6. A. Liu, Y. Lin, M. Yang, H. Guo, S. Guan, J. Sun and D. Guo, J. Chromatogr. B: Anal. Technol. Biomed. Life Sci., 2007, 846, 32–41.
7. G. Alaerts, M. Merino-Arevalo, M. Dumarey, D. B. N. Noppe, N. Matthijs, J. Smeyers-Verbeke and Y. V. Heyden, J. Chromatogr. A, 2010, 1217, 7706–7716.
8. G. Sun, L. Li, Z. Liu and Y. Li, Ann. Transl. Med., 2013, 1, 30 , http://www.atmjournal.org/article/view/2387/3253.
9. Y. Wu, S. Sun, Q. Zhou and H. Leung, J. Pharm. Biomed. Anal., 2008, 46, 498–504.
10. W. Li, Z. Cheng, Y. Wang and H. Qu, J. Pharm. Biomed. Anal., 2013, 72, 33–39.
11. X. Wang, D. Sheng, Z. Zhu, F. Xu, D. Huang and C. Yu, Spectrochim. Acta, Part A, 2015, 141, 94–98.
12. M. Diem, Introduction to Modern Vibrational Spectroscopy, Wiley Press, New York, 1993.
13. M. Guillen and N. Cabo, J. Sci. Food Agric., 1997, 75, 1–11.
14. J. R. J. Pare and J. M. R. Belanger, Instrumental Methods in Food Analysis, Elsevier Press, New York. 1997.
15. H. A. Gad, S. H. El-Ahmady, M. I. Abou-Shoer and M. M. Al-Azizi, Phytochem. Anal., 2013, 24, 1–24.
16. J. van der Greef, S. Martin, P. Juhasz, A. Adourian, T. Plasterer, E. R. Verheij and R. N. McBurney, J. Proteome Res., 2007, 6, 1540–1559.
17. O. Blokhina, E. Virolainen and K. V. Fagerstedt, Ann. Bot., 2003, 91, 179–194.
18. S. Dudonné, X. Vitrac, P. Coutière, M. Woillez and J. M. Mèrillon, J. Agric. Food Chem., 2009, 57, 1768–1774.
19. S. M. Jachak, R. Gautam, C. Selvam, H. Madhan, A. Srivastava and T. Khan, Fitoterapia, 2011, 82, 173–177.
20. National Pharmacopoeia Committee, Pharmacopoeia of People's Republic of China, The Chinese Medicine Science and Technology Press, Beijing, 2010, vol. 1, pp. 587–588.
21. H. S. Feng and G. F. Tian, China Pharm., 2007, 16, 32–33.
22. Y. C. Wu, S. X. Zhuang, Y. B. Xun, L. G. Jiang and Z. Z. Bai, Acta Chin. Med. Pharmacol., 2009, 37, 56–57.
23. Y. Liu, G. Sun, Y. Wang, L. Yang and F. Yang, J. Sep. Sci., 2015, 38, 2174–2181.
24. F. Gong, B. T. Wang, F. T. Chau and Y. Z. Liang, Anal. Lett., 2005, 38, 2475–2492.
25. G. X. Sun, Y. S. Hu and K. S. Bi, Acta Pharmacol. Sin., 2009, 44, 401–405.
26. Q. Zhang and M. Ye, J. Chromatogr. A, 2009, 1216, 1954–1969.
27. L. Xu, Chemometrics—Principles and Applications of Essential Methods, Science Press, Beijing, 2004, pp. 14–16.
28. M. Wichtl and N. G. Bisset, Herbal Drugs and Phytopharmaceuticals, Medpharm Scientific Publisher, Stuttgart, 1994.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c5ra21468h

---