Calculating Relative Correction Factors for Quantitative Analysis with HILIC-HPLC-ELSD Method: Eight Fructooligosaccharides of Morinda Officinalis as a Case Study

Objective Because the response of evaporating light scattering detector (ELSD) being in a nonlinear mode, there is no consensus on the method of calculating its relative correction factors (RCF), which limits the application of the quantitative analysis for multi-components by a single marker (QAMS) with LC-ELSD. Methods Using eight fructooligosaccharides of Morinda officinalis as a case study, the nystose (GF3) as a single standard was adopted to develop a QAMS method to simultaneously determine the other seven fructooligosaccharides with HILIC-HPLC-ELSD method. Six calculation methods of RCF were investigated to select the most reasonable method. The relative error of content between the QAMS and the external standard method (ESM) obtained from 30 batches of samples was used as an indicator to evaluate the six methods. Finally, a chemometrics analysis was performed to find the differential components among MO and its three processing products. Results It was first reported that only one calculation method was scientific for calculating RCF for the LC-ELSD method. The RCFs of GF3 to the other seven fructooligosaccharides (GF1–GF8) were obtained as 0.86, 0.91, 0.93, 1.05, 1.15, 1.12, and 1.18, respectively. The QAMS of eight fructooligosaccharides of Morinda officinalis was validated with good linearity (R2 > 0.9998) and accepted the accuracy of 95–105% (RSD < 1.81%) based on nystose. Finally, Morinda officinalis and its three processed products were distinguished and could be differed based on the content of the eight fructooligosaccharides. Conclusion The scientific calculation method of RCF would be of great significance for developing the QAMS method in Pharmacopoeia when performing the LC-ELSD method.


Introduction
Quantitative analysis of multi-components by a single marker (QAMS) is a method for obtaining accurate content of many components using only a single reference standard combined with a relative correction factor (RCF) and relative retention time (RRT). It is widely used for the simultaneous determination of multiple components in Chinese medicinal materials and preparations, improving and enhancing the quality standards of Chinese medicines and ensuring safe and effective clinical use, while greatly cutting the cost of additional analytes for quantitative methods. RCF is the ratio of the correction factor (f ) of the reference to the component to be measured. e f-value reflects the proportionality between the quantity (mass or concentration) of the constituent and the response value of the detector over a specific linear range. Detectors typically used for quantitative analysis are ultraviolet absorption detectors (UV) and evaporative light scattering detectors (ELSD). In the UV detector, the f-value is the slope of the standard curve, as in the case of saponins, isoflavonoids, and glycosides in red ginseng and Astragali Radix [1,2]. Since the response of the ELSD is in a nonlinear mode, there is no consensus on the calculation method of its RCF [3]. It is necessary to select the optimal RCF calculation method and establish the accurate QAMS method for the ELSD system, which is of great significance for the improvement and formulation of pharmacopoeia.
Morinda officinalis Radix (MO) is the dried root of Morinda officinalis How, sweet and pungent in taste, slightly warm. It has the efficacy of tonifying the liver and kidney, strengthening the muscles and bones, and dispelling wind and dampness [4]. In the Chinese Pharmacopoeia (ChP.2020), MO was processed into three kinds of product, steam-processed MO (StMO, St), salt-steamed MO (SMO, S), and licorice-boiled MO (LMO, L), for enhancing its effects. MO contains various chemical components such as polysaccharides, oligosaccharides, anthraquinones, iridoid glycosides, and organic acids [5][6][7]. Studies have shown that fructooligosaccharides (GFns), with anti-depressant, antiaging, anti-osteoporosis, and other activities, are one of the main types of active ingredients in MO [8][9][10][11][12]. In the current standards, such as ChP.2020 and Hong Kong Chinese Materia Medica Standard (Volume X), the content of nystose (GF3) is used as an indicator for the quality control of raw and its three processed products, with a lower limit of 2.0% and 2.3%, respectively [13]. It is worth noting that in addition to GF3, the MO also contains many other fructooligosaccharides, such as sucrose (GF1), 1-Kestose (GF2), 1F-fructofuranosylnystose (GF4), 1,1,1,1-kestohexose (GF5), fructoheptasaccharide (GF6), fructo-oligosaccharide DP8/GF7 (GF7), and fructo-oligosaccharide DP9/GF8 (GF8) with high content and physiological activity [14,15]. erefore, for the improvement of the quality standard of MO and its processed products, the quality should be comprehensively evaluated by the content of GFns rather than that of GF3.
Quality control studies on the GFns components of MO have been performed in the literature and the qualitative/ quantitative detection methods were established on UHPLC-ELSD [16], HPLC-CAD [17], NIR [18], and LC-MS [19], among which UHPLC-ELSD was widely used. In addition, the hydrophilic interaction chromatography (HILIC) system is a primary method for separating GFns [20]. For example, Y. Yu determined GF1 to GF11 in MO, StMO, SMO, and L. Yang determined GF1 to GF4 in different parts of MO, in both of which HILIC-HPLC-ELSD was used [20,21]. However, the content of GFns in MO was all calculated by the external standard methods in the existing literatures, in which many expensive standard substances were required. erefore, it is urgent to establish a QAMS method based on the HILIC-HPLC-ELSD system to evaluate the GFns of MO and its processed products.
Based on the previous studies, the calculation method of RCF in the HILIC-HPLC-ELSD system was investigated with MO as an example. e QAMS methods with six RCF calculation methods for the eight oligosaccharides in ELSD were established and compared with the external standard method to select the optimal RCF. Subsequently, the eight fructooligosaccharides from 30 samples were analyzed by chemometrics to reveal their variations during steaming-and boiling-process.

Preparation of Sample
Solutions. All samples were powdered using a pulverizer and passed through a 24-mesh sieve. e powdered sample (0.25 g) was dissolved in 25 mL of 60% ethanol (v/v) in a conical flask with a stopper, weighed, and sonicated (250 W, 53 kHz) for 10 minutes, and allowed to cool. e mixture was weighed again and replenished the lost weight with the same solvent. e supernatant was passed through a 0.45 μm nylon66 membrane and the successive filtrate was collected as the sample solution for further analyses.

HILIC-HPLC-ELSD Analysis.
e quantitative assay was performed on an Agilent 1260 series HPLC system equipped with an ELSD (Agilent Technologies, Palo Alto CA, USA), which was controlled by Agilent ChemStation software (B.04.03-SP1). e separation of analytes was conducted on a Waters XBridge HILIC column (4.6 × 150 mm, 3.5 μm) with a flow rate of 1.0 mL/min at 30°C. e mobile phases were acetonitrile (A) and water (B) with a gradient elution of 88% A at 0-1 min, 88-78% A at 1-10 min, 78-65% A at 10-20 min, 65-88% A at 20-20.1 min, and 88% A at 20.1-35 min. e injection volume was 5 μL. e drift tube temperature of ELSD was 50°C and the nitrogen cumulative flow rate was 1.0 mL/min. Table 2, six methods were employed to calculate the RCFs in ELSD. F k was the slope of the linear equation of other seven analytes, C k was the true concentration of other seven analytes in standard solution, C k−detected was the concentration of other seven analytes calculated by calibration curves of nystose (GF3) in standard solution, and A k was the peak area of other seven analytes in standard solution. F s was the slope of the linear equation of nystose (GF3), C s was the concentration of GF3 in standard solution, and A s was the peak area of GF3 in standard solution.

Chemometrics Analysis.
e Hierarchical cluster analysis (HCA) was carried out by calculating Squared Euclidean distance with Origin software (2021). Principal component analysis (PCA) and orthogonal partial least-squares discrimination analysis (OPLS-DA) were performed with SIMCA software v.14.1 Umetrics, Umea, Sweden, and components with VIP values >1.0 in OPLS-DA were defined as potential chemical markers and applied for further analysis.

Optimization of Sample Preparation.
Taking the theoretical plates and tailing factors of GF3 and the extraction efficiency of the eight components as indicators, sample preparation was optimized systematically concerning extraction solvents (ethanol (20%, 40%, 60%, and 80%, v/v), methanol (20%, 40%, 60%, and 80%, v/v), and water), extraction methods, extraction solvent volumes, extraction time, and extraction frequency. e ultimate choice was 60% ethanol (v/v) as the extraction solvent, 25 mL as the extraction solvent volume, and 10 min as the ultrasonic extraction time.

Optimization of Chromatographic Conditions.
As the GFns component was strongly hydrophilic and had no UV absorption, the HILIC-HPLC-ELSD system was performed in this research, among which HPLC-ELSD has been used to determine GF3 in both ChP.2020 and Hong Kong Chinese Materia Medica standard (Volume X). e separation of GFns has investigated in three columns, XBridge HILIC (silyl group), ACHROM XAmide (amide group), and ZIC HILIC column (amphoteric group). e XBridge HILIC column was selected ultimately for its smoother baseline and the shortest retention time of eight GFns in approximately 20 minutes ( Figure S2). For the improvement of the efficiency, the column particle size (2.7, 3.0, 5.0 μm), the column temperature (25, 30, 35, 40°C), the flow rate (0.9, 1.0, 1.1 mL/ min), and the ratio of acetonitrile in the mobile phase (±1%) were investigated this research. In addition, four ELSD parameters, including the gain value, evaporator temperature, nebulizer temperature, and gas flow rate, were also optimized.

Calibration Curves, Limits of Detection and Limit of
Quantification. e structures of eight GFns were shown in Figure 1. e calibration curve, regression coefficients, linear range, limits of detection (LOD), and limits of quantification (LOQ) were shown in Table 3 and Figure 2(a). All calibration curves showed good linearity (R 2 > 0.9998) within the test ranges. In addition, the LOD and LOQ of each standard were in the ranges of 1.29-3.41 μg/mL and 4.32-11.37 μg/mL, respectively, which were 10 times better than the reported results [14], indicating the high sensitivity of the HILIC-HPLC-ELSD system in this study.

Calculation of RCFs with Six Methods.
As summarized in Section 2.5, there were six main calculation methods of the RCFs in the HPLC-ELSD system. For establishing a more accurate QAMS method, the RCFs of the analytes were calculated by the six methods. As the quantitative marker of MO in ChP.2020, GF3 had the advantages of moderate retention time, stability, and inexpensive, which was selected as the single marker.
e results showed there were significant differences in the RCFs among different methods (Table 4 and Figure 2(b)). en, further analysis would be carried out to compare them.

Precision, Repeatability, Stability Testing, And Recovery
Test.
e content determination results of the QAMS methods with six RCFs calculation methods should be compared with those of the external standard method to select the optimal RCFs calculation method. Methodological validation of the method was required prior to the content determination. To evaluate the precision of the instrument, six successive injections of the same sample solution on one instrument were performed. e results (Table 5) showed the relative standard deviation (RSD) of the peak areas was less than 1.91%. e repeatability was tested with nine test solutions covering three different concentration levels (0.125 g, 0.250 g, and 0.375 g).
e RSDs of the peak areas for each analyte were less than 3.2%. e stability was analyzed by storing the sample and standard solutions at room temperature for 0, 6, 12, 24, and 36 hours. e RSDs of the sample and standard solution were less than 2.42% and 2.23%, respectively, indicating both sample and standard solutions were stable over 36 hours.
e recovery experiments were performed by adding three different volumes (5 mL, 10 mL, and 15 mL) of GF3 solution (500 μg/mL) to the flasks with the powder of MO samples (125 mg), and triplicate experiments were performed at each level. e recovery was calculated according to the formula: % recovery � measured amount − origin content spiked amount × 100%. (1) As shown in Table 6, the recoveries of spiked GF3 ranged from 95 to 105% (RSD < 1.81%), which indicated the good accuracy of the method.

Selection of the Optimal RCFs by the Comparison of QAMS
and External Standard Method. As previously mentioned, the optimal RCFs were selected by comparing the content determination results of the QAMS methods with six RCFs calculation methods and those of the external standard method. So, the QAMS and the external standard method were used to determine the content of eight GFns in 30 batches of MO (raw) and its three processed products. e HPLC chromatograms were shown in Figure 3.
en the relative error (RE) between the QAMS and the external standard method was used as the index to evaluate the accuracy of the six QAMS methods. e results (Figure 4)

e Evaluation of System Suitability on RCFs.
e system suitability test of RCFs was investigated on two Agilent 1260 ELSD instruments from different vendors, and the results showed that the RCFs of eight GFns exhibited good repeatability (RSD < 1.8%) ( Table 7).

Chemometric Analysis on the Content of the Samples.
Subsequently, the content of 30 batches of raw and processed MO samples calculated by the optimal RCFs was used for chemometric analysis.
In the results of the content ( Figure S3), it was obviously found that the total oligosaccharides content of one of the three parallel samples L_02-3 (36.22%) was significantly lower than that of the other two parallel samples L_02-1 and L_02-2 (49.86% and 47.33%) which probably came from the preparation process. To find the variation of the oligosaccharides between the raw and processed products, the sample L_02-3 was eliminated in the further analysis, and so was the sample S_02-3.
Twenty-eight batches of samples were eventually used for further analysis, including three batches of raw and 25 batches of processed products. Figure 5 showed the distribution of the content of GFns in 28 batches of samples. As shown in Figure 5(a) and Figure 5(b), the content of the individual and the total oligosaccharides differed slightly between the raw and processed products. It was also found that the contents of GF3 in the four decoction pieces were about 5% (Table S1 and Figure 5), while the total content of the eight oligosaccharides were about 45%, which indicated that it was unreasonable to use GF3 only to evaluate the quality of raw and processed MO.
Since there were differences among the three batches of the raw, the content ratio of the processed products to the corresponding raw samples was used for analysis. HCA could group samples with the same characteristics and determine the variation degree of samples with the same characteristics in the group, which could reveal the differences between the raw and processed products. e HCA analysis, an unsupervised pattern recognition method based on Euclidean distance, differentiated the samples into two major groups in which S was in Group One and St, L, and the  Figure 1: e structures of eight GFns.  raw MO were in Group Two. PCA was utilized to investigate the chemical differences between raw and processed products. In the PCA score plot, samples were clustered into two major groups: S in one group, while St, L, and the raw MO in the other group. e PCA results were the same as those of HCA. e PCA loading scatters plot showed the correlation between the variables in the PC1 and PC2 coordinate systems and the association between the variables and the samples. From the PCA loading scatter plot, the classification was influenced by all nine GFns components ( Figure S4). Among the nine GFns components, the     GF1A  GF2A  GF4A  GF5A  GF6A  GF7A  GF8A  TotalA  GF1B  GF2B  GF4B  GF5B  GF6B  GF7B  GF8B  TotalB  GF1C  GF2C  GF4C  GF5C  GF6C  GF7C  GF8C  TotalC  GF1D  GF2D  GF4D  GF5D  GF6D  GF7D  GF8D  TotalD  GF1E  GF2E  GF4E  GF5E  GF6E  GF7E  GF8E  TotalE  GF1F  GF2F  GF4F  GF5F  GF6F  GF7F GF8F TotalF  Figure 4: e relative error between the content calculated by the QAMS and the external standard method A, B, C, D, E and F were six RCF calculation methods, respectively.  variation trends of GF1 and GF2, GF7 and GF8 were consistent, while the variation trends of GF1 and GF2 were opposite to those of GF7 and GF8. Next, OPLS-DA was performed, which could find the greatest contributing constituents to the differences between the raw and processed products. As shown in Figure 6, these four groups (raw, St, S, and L) clustered separately in the OPLS-DA score plot, indicating a significant chemical variation between the raw and processed products. In the OPLS-DA model, the parameters R 2 and Q 2 indicated the explanatory and predictive ability of the model, respectively. Both the two parameters were above 0.5 in this model (R 2 � 0.74 and Q 2 � 0.56, respectively), suggesting that the explanatory and predictive capacity of the established model was satisfactory. e permutation test was used to ascertain whether the model was over-fitted, and the intercept of Q 2 on the y-axis in this model was less than 0.05, meaning that the model was not over-fitted.
e VIP values in the OPLS-DA model indicated the influence intensity and explanatory power of each metabolite effect on the sample classification, which could be used for dissecting the potential markers. In this work, the components with VIP values >1.0 were selected as differential compounds between the raw and processed products. e results showed that the magnitude of VIP values of differential compounds for raw-St, raw-S, and raw-L were GF2 > GF8 > GF7, GF2 > GF1 > GF8 > GF7, and GF2 > GF8 > GF1 > GF3, respectively ( Figure S5). Subsequently, we further analyzed the content ratio of GF1, GF2, GF3, GF7, and GF8 for the processed products to the raw (Figure 7). Compared to the raw MO, the content of GF8 in L decreased slightly, and the content of GF1, GF2, and GF3 increased slightly. Nevertheless, compared to the raw, the contents of GF7 and GF8 in S and St were significantly decreased, and the contents of GF2 were significantly increased. e difference in the content of oligosaccharides between the processed products and the raw may come from the different processing methods, where St and S were steaming while L was boiling.
ere were two possible reasons for the insignificant changes of GFns content in L:  Journal of Analytical Methods in Chemistry (1) GFns with a high degree of polymerization may be hydrolyzed at high temperatures. e higher the temperature was, the stronger the hydrolysis was. In the steamingprocess (S and St), when the temperature was greater than 100°C, the hydrolysis reaction of GFns was strong and the content changed greatly. However, in the boiling-process (L), when the temperature was about 100°C, the hydrolysis reaction of GFns was weak.
(2) e procedure of the preparation of L consisted of two steps, first, decocting licorice and removing the residue for the decoction, and then boiling the raw MO with the decoction. e increase of the GF3 content in L may be due to the increase in its solubility caused by the addition of licorice [27]. Although both St and S were obtained by steaming-process, the change in the content of S was more significant than that of St, which may be due to the addition of salt.

Conclusions
In this study, the calculation methods of RCFs in the HILIC-HPLC-ELSD system were first investigated by six methods, using eight GFns in MO as an example. It was found that only one method was scientific. It was much important for developing the QAMS method for the Pharmacopoeia method when performing the LC-ELSD method. Based on the selected calculation method of RCF, seven fructooligosaccharides of Morinda officinalis were determined and validated with nystose (GF3) as a single standard. It was found that the contents of GF3 in the raw and processed MO were about 5%, while the total content of the eight oligosaccharides was about 45%, indicating that it was important to evaluate the quality using the eight GFns rather than GF3. Finally, the decrease of GF7 and GF8 and the increase of GF1 and GF2 during the steaming-process (S and St) and boilingprocess(L) were revealed by the chemometrics.
It could be concluded that during the processing of MO, among the eight fructooligosaccharides, the content of GF7 and GF8 decreased while the content of GF1 and GF2 increased. It might be that during heat-processing, GF7 and GF8 with a high degree of polymerization degraded into GF1 and GF2 with a low degree of polymerization. e content variations of GFns before and after processing might be related to the processing method (steaming or boiling) and the addition of the excipients (salt or licorice). For L, the content variations of GFns before and after processing were less than those of S and St, which were steamed and saltsteamed. It might be that the heating temperature during boiling was slightly lower than that during steaming and the addition of licorice improved the solubility of GFns. For S and St, probably due to the addition of salt, the content variations of GFns in St before and after processing were slightly less than that in S.

Data Availability
e data used to support the findings of this study are available from the corresponding authors upon request.

Conflicts of Interest
e author(s) declare that there are no conflicts of interest regarding the publication of this paper. Table S1. Comparison of the content determined by QAMS (Method A) and ESM in raw and processed products (%, g/ g) (n = 2). Figure Figure S3. Ten batches of Raw (02) and its processed products, two of which had abnormal values, i. e. S_02-3 and L_02-3. Figure S4. HCA results (A), PCA score plots (B), and loadings scatter plots (C) for Raw, St, S, and L. Figure S5 Journal of Analytical Methods in Chemistry 9