Characterization of main components in Xiao’er Xiaoji Zhike oral liquid by UPLC-MS and their taste evaluation

snowzhan@bucm.edu.cn This paper provided a method for determining the potential quality markers in Xiaoji Zhike oral liquid (XXZOL) based on the concordant tastes of compounds with their respective originated Chinese medicinal pieces. UPLC-Q-Exactive-Orbitrap-MS technology was used to identify the main chemical constituents in XXZOL. The electronic tongue collected the electronic responses of the sour, bitter, sweet, pungent, and salty sample solutions, and the discriminant factor analysis (DFA) model was established to recognize the taste characteristics of 23 liquid samples. Fifteen high content ingredients in XXZOL were identified by UPLC-MS, and the established DFA model recognized their respective tastes. The accurate recognition rate of the DFA model was 73.33%, the false rate was 6.67%, and the unrecognized rate was 20%. The concordance rate of their authentic tastes with the tastes of their respective originated Chinese medicinal pieces was 78.57%. Trigonelline, malic acid, citric acid, and caffeic acid were the potential sour material bases of XXZOL. Mannitol was the potential sweet material basis of XXZOL. L-phenylalanine, sinapine, forsythoside I, pinoresinol-4-O-β-D-glucopyranoside, naringin, and neohesperidin were the potential bitter material bases of XXZOL. So the above 11 main compounds were the potential quality markers of marker. Practical Application: A method for determining the potential quality markers in Xiao’er Xiaoji Zhike oral liquid based on the concordant tastes of compounds with their respective originated Chinese medicinal


Introduction
The electronic senses electronically reproduce the responses similar to the five senses presenting in human beings, such as electronic eye (Orlandi et al., 2018), electronic ear, electronic skin (Zhang et al., 2020b), electronic tongue (e-tongue) (Toko,1998) and electronic nose (Mohd Ali et al., 2020). These electronic sensory technologies can be used alone (Gan et al., 2019;Pascual et al., 2018;Nategh et al., 2021) or combined (Orlandi et al., 2019;Banerjee et al., 2019;Xu et al., 2019) to achieve experimental purposes. The e-tongue is a kind of liquid analytical instrument consisting of three parts, including sensor array, signal acquisition system, pattern recognition system (Toko, 1998), and taste sensations of the sensor array can usually be classified into sweet, sour, salty, bitter, and umami. Sensor array imitates the taste cells in taste buds that interact with the flavorful substance and transmit the taste information to nerve fibers and the relevant areas in the central nervous system (CNS) (Pandurangan & Hwang, 2015). The non-specific and poorly selective sensors in the sensory array are immersed into a sample solution and acquire a global characteristic response signal of the substances in the sample (Jiang et al., 2018). Then the corresponding signals are sent to a signal processing system equivalent to CNS, which can analyze the acquired signals using appropriate pattern recognition methods. The e-tongue can realize the qualitative and quantitative analysis of simple or complex liquid samples and obtain the results reflecting the taste characteristics of samples (Legin et al., 2000). Based on the working principle of the sensor array, the e-tongue can be classified into electrochemical (Wei et al., 2018;Ciosek & Wróblewski, 2011), optical and enzymatic e-tongue. Besides the advantages of low cost and simple operation, sensitivity, reliability, and robustness, e-tongue has its particular advantage to analyze the taste of some toxic substances (Wadehra & Patil, 2016). The e-tongue has been widely used in different fields of food (Wadehra & Patil, 2016;Ghasemi-Varnamkhasti et al., 2018;Peris & Escuder-Gilabert, 2016), medicine (Wasilewski et al., 2019;Guedes et al., 2021), or environmental detection (Shimizu et al., 2019;Magro et al., 2018). The e-tongue system has good applications in quality monitoring of Chinese medicinal pieces (Shi et al., 2018), origins identification (Wu et al., 2018) and harvest period determination of herbal medicines, and the evaluation of the pharmaceutical process of traditional Chinese medicine (TCM) (Bi et al., 2020), such as correcting or masking the taste of the decoction from TCM (Lin et al., 2016;Feng et al., 2019).
Xiao' er Xiaoji Zhike oral liquid (XXZOL), a prescription preparation of TCM, with the effect of clearing heat in lungs, relieving cough due to lung heat, and helping digestion, usually is applied to treat children's respiratory diseases such as mycoplasma pneumonia (Zhang et al., 2020a), asthma (Zhou et al., 2020) and cough (Liang et al., 2018). Because of its safety and definite effects, the market sales of XXZOL have been increasing steadily. There are 10 Chinese medicinal pieces in the prescription of XXZOL according to their taste deployment to reduce the toxicity and bias of a single Chinese medicinal piece and achieve the overall synergy of all Chinese medicinal pieces. This prescription is the compatibility of stir-baked Crataegi Fructus, Arecae Semen, honeyed Eriobotryae Folium, Trichosanthis Fructus, Platycodonis Radix, Forsythiae Fructus, Aurantii Fructus Immaturus, stir-baked Raphani Semen, stir-baked Descurainiae Semen Lepidii Semen, and Cicadae Periostracum (China, 2020). The taste of a single herbal decoction or a prescription preparation decoction is stable only when the type and quantity of the chemical compounds in the sample solution do not alter. According to the main tastes of Chinese medicinal pieces from the prescription, to understand the recipe-construction rule of XXZOL, high content components in XXZOL were identified. Their tastes were recognized respectively by the e-tongue system. Comparing with the main tastes of Chinese medicinal pieces from the prescription, components with consistent tastes are more likely to be the material bases of XXZOL.
In our work, the UPLC-Q-Exactive-Orbitrap-MS platform was used to analyze the chemical constituents of XXZOL, and the main components were identified by the high-resolution mass spectrometry (MS) technology and control substances. The e-tongue collected the sour, bitter, sweet, pungent, and salty samples, which were 1.0 mg/mL solutions of the reference components with above tastes, and the discriminant factor analysis (DFA) model to recognize the taste characteristics of liquid samples was established to predict the taste of every main component in XXZOL. Considering the main tastes of Chinese medicinal pieces from XXZOL, we obtained several main components in XXZOL with similar tastes, which were likely to be the important quality markers of the prescription, providing a basis for the determination of the quality markers of XXZOL.
XXZOL was purchased from Lunan Hope Pharmaceutical Co., Ltd (Linyi, Shangdong province, China). Take the oral liquid 0.25 mL precisely to 10 mL volumetric flask, water was added to the mark. Then, the solution was filtered with a membrane filter (0.22 µm) to collect the successive filtrate as the sample solution.
The powder (25 mg) of each reference component was dissolved in 25 mL of 50% v/v aqueous ethanol with ultrasonication at room temperature for 10 minutes. The obtained solutions were collected for electronic tongue analysis.

The electrochemical e-tongue device and data collected parameters
The ASTREE e-tongue (Alpha M.O.S., France) equipped with the fifth set of sensor systems was used, and the sensor system had 7 electrochemical sensors (SRS, GPS, STS, UMS, SPS, SWS, BRS) and an Ag/AgCl reference electrode. After 7 electrochemical sensors were activated and calibrated, these sensors were put into a liquid sample to collect the data, including taste characteristics of the liquid sample. The parameters for the e-tongue data collected were as follows, the data acquisition time 120 s, acquisition period 1.0 s, acquisition delay 0 s, stirring rate 1.0 r·s -1 . Nine replicates for each sample, only the average of the 100-120 s data of the last 3 replicates were kept as 3 results for each sample to ensure the stability of the sensor response values, indicating that the sensor responses were stable when the RSD value of 3 results for each sample was less than 5%.

Data processing
The Xcalibur 4.0 software (Thermo Fisher Scientific, USA) was used to calculate the high-resolution accurate mass of the compounds. Based on error less than 5 ppm and MS/MS fragment matching, we identified main compounds in XXZOL, and the identified substances were used to verify the identified results.
Principal component analysis (PCA) and DFA were used to recognize the taste characteristics of liquid samples of reference components and establish the classification model of five kinds of tastes. The taste of every main component in XXZOL was discriminated by the FDA model.

Identification of main chemical compounds in XXZOL
The negative and positive total ion chromatograms of XXZOL were shown in Figure 1. Due to their high responses in the mass spectrometer and UV spectrometer, high content compounds in XXZOL were focused on, and 15 high content compounds in XXZOL were chosen considering the compounds' solubility in 50% ethanol. The high-resolution accurate mass and major MS/MS fragments of the above 15 main compounds were presented in Table 2, and the mass relative errors between the theoretical mass and measured mass of 15 main compounds were all smaller than 5 ppm, and the MS/MS fragments measured of each compound existed rationally and accorded with the MS/ MS fragments reported in the works of literature. Furthermore, 12 main compounds were identified with the corresponding control substances in the DAD spectrometer at the detection wavelength of 254 nm, (A) the sample chromatogram, (B) the mixed control substances chromatogram, as shown in Figure 2, UV absorption spectrum and retention time for each compound in the sample were consistent with those in control substances.

Analysis of taste resolution ability of electronic tongue
Prepare 50% ethanol solutions with high (8.0 mg/mL), medium (5.0 mg/mL), and low (1.0 mg/mL) concentrations of citric acid monohydrate (sour), quinine (bitter), sucrose (sweet), zingerone (pungent) and sodium chloride (salty), and collect the data of the above 15 solutions by the e-tongue, 3 replicates for each solution. PCA and DFA were applied to process the e-tongue data of 15 samples, as shown in Figure 3, 3 collected data of the same solution well-replicated, which indicated the stability of the e-tongue platform and the robustness of the e-tongue data.
In the two-dimensional PCA scatter plot of 5 reference components with different tastes, the accumulative contribution rate for the two PCs was 91.264%, in which PC1 accounted for 69.075%, and PC2 contributed 22.189%, and five reference components with different tastes were differentiated ( Figure 3A). Three different concentration samples from the same reference component had small changes in e-tongue data and were all clustered into the same category in the two-dimensional PCA plot. The results showed that the e-tongue combined with PCA could distinguish the reference components with different tastes, and concentration of the sample solution had little effect on the classification of the PCA model.
In the two-dimensional DFA scatter plot of 5 reference components with different tastes, the accumulative contribution rate with the two DFs was 97.395%, in which DF1 accounted for 91.131%, and DF2 contributed 6.264%. Five reference components with different tastes were better differentiated in Figure 3B when compared with the results of the PCA model ( Figure 3A), the projection distances between different concentration samples from the same reference component, i.e., the projection distances in the same group were more concentrated, which showed that concentration of sample solution also had little effect on the classification of the DFA model. The distances between groups were further increased in DFA scatter plot, indicating that the DFA model of the e-tongue data had better classification results for five reference components with different tastes than the PCA model.
The classification model of the e-tongue data could effectively distinguish the reference components with different tastes. The different concentration samples of the same reference component overlapped together in the two-dimensional PCA or DFA scatter plot, indicating that the concentration of the sample solution had little effect on the classification ability of the classification model. So 1.0 mg/mL reference component solutions were prepared in the following research, and DFA was applied to the sensor data of these solutions collected by the e-tongue platform.

DFA classification model for distinguishing components with different tastes
Prepare 1.0 mg/mL 50% ethanol solutions of reference components in Table 1, collect data according to the parameters of the e-tongue in section 2.3, and DFA was used to recognize the taste characteristics of reference component solutions with five different tastes.
In the two-dimensional DFA scatter plot of these reference component samples with 5 kinds of tastes, the accumulative contribution rate with the two DFs was 75.023%, in which DF1 was 44.389%, and DF2 was 30.634%. Reference components with different tastes were differentiated, and reference components with the same taste were clustered into one category (Figure 4). The results showed that the e-tongue combined with DFA could effectively distinguish reference components with different tastes.

Taste judgments of 15 main components in XXZOL
Prepare 1.0 mg/mL 50% ethanol solutions of 15 main components in XXZOL. The established DFA model was used to discriminate the tastes of these solutions. The taste information of 15 main components was shown in Table 3. Authentic taste of each component was evaluated by three assessors, predicted taste of each component was judged by the DFA model of the reference components in Table 1, the originated Chinese medicinal piece was obtained by the reported works of literature, and the piece taste was obtained by the Chinese Pharmacopoeia (2020 version a). Table 3, the tastes of 12 main components were effectively recognized by the established DFA model. Except for forsythoside I, the predicted tastes of mannitol, trigonelline, sucrose, malic acid, citric acid, L-phenylalanine, forsythoside E, caffeic acid, forsythoside A, naringin, and neohesperidin were The authentic tastes of mannitol, trigonelline, malic acid, citric acid, L-phenylalanine, caffeic acid, sinapine, forsythoside I, pinoresinol-4-O-β-D-glucopyranoside, naringin, and neohesperidin were similar to the main tastes of their respective originated Chinese medicinal pieces. The concordance rate of authentic tastes of 14 compounds with the main tastes of their respective originated Chinese medicinal pieces was 78.57%, the false rate was 21.43%.

As shown in
The originated Chinese medicinal pieces were the main herbal decoctions in the prescription of XXZOL, including Stir-baked Crataegi Fructus, Trichosanthis Fructus, Aurantii Fructus Immaturus, Platycodonis Radix, Forsythiae Fructus, consistent with their authentic tastes. The accurate recognition rate for the above 15 main components was 73.33%, the false rate was 6.67%, and the unrecognized rate was 20%. When the number and representativeness of the reference components for the established DFA model increase, the unrecognized rate will further decrease. So the e-tongue combined with classification model can be used for predicting the tastes of different components in Chinese medicinal pieces.
For XXZOL, exclude sucrose as the additive of the preparation, the remaining 14 ingredients with high responses in the mass spectrometer, were all from Chinese medicinal pieces in XXZOL.  Table 1 (1.0 mg/mL).

Original Article
Stir-baked Descurainiae Semen Lepidii Semen and Stir-baked Raphani Semen. The high content compounds of the above Chinese medicinal pieces included 14 compounds in Table 3, from which 11 compounds were chosen because of consistent tastes with the main tastes of their respective originated Chinese medicinal pieces, and these compounds were more likely to be the quality markers of XXZOL. Trigonelline, malic acid, citric acid, and caffeic acid were the potential sour material bases of XXZOL. Mannitol was the potential sweet material basis of XXZOL. L-phenylalanine, sinapine, forsythoside I, pinoresinol-4-O-β-D-glucopyranoside, naringin, and neohesperidin were the potential bitter material bases of XXZOL.

Conclusions
In this work, 15 high content compounds in Xiao' er Xiaoji Zhike oral liquid were identified according to the high-resolution mass data and the MS/MS fragments. The e-tongue collected the sour, bitter, sweet, pungent, and salty samples, which were 1.0 mg/mL solutions of the reference components with above tastes, and DFA model to recognize the taste characteristics of 23 liquids samples was established to predict the tastes of 15 main components in XXZOL, and the accurate recognition rate was 73.33%, the false rate was 6.67%, and the unrecognized rate was 20%. The concordance rate of their authentic tastes with the tastes of their respective originated Chinese medicinal pieces was 78.57%, and 11 compounds were chosen because of consistent tastes with the main tastes of their respective originated Chinese medicinal pieces. Trigonelline, malic acid, citric acid, and caffeic acid were the potential sour material bases of XXZOL. Mannitol was the potential sweet material basis of XXZOL. L-phenylalanine, sinapine, forsythoside I, pinoresinol-4-O-β-D-glucopyranoside, naringin, and neohesperidin were the potential bitter material bases of XXZOL. Starting with the taste prediction and confirmation of the main components in XXZOL, this paper provided a method for determining potential quality markers based on the concordant tastes of compounds with their respective originated Chinese medicinal pieces, which provided a scientific basis for the quality evaluation of XXZOL.