Chemical reference standards (CRS) of natural products (CRSNPs) play an important role in quality control of traditional Chinese medicines (TCM), e.g. identification of medicinal plants or animals, assay of marker compounds in TCM, etc. In the content assignment of CRS, mass balance method is generally applied, i.e. the total amount of target compound, organic impurities, water, residual solvents plus inorganic impurities is 100 percent. As the majority of CRSNPs are obtained from medicinal plants or animals by extraction and further purification, there are high risks that some structural analogues or isomers may co-exist with the target main compound(s). Thus, organic impurity determination is a critical aspect which may influence accuracy of the assigned content of CRSNPs.

The total amount of organic impurities of CRSNPs is generally determined by HPLC methods described in pharmacopoeia or literature. In the descriptions of these methods, exact eluent composition and other experimental conditions (e.g. detection wavelength, column temperature) are provided, meanwhile only general information is given in pharmacopoeia about the stationary phase. However, there have already been over 800 brands of commercial C₁₈ columns on the market and their selectivities may be influenced by many factors, such as the presence of free silanol groups, end-capping or embedded polar groups. Moreover, new technologies such as hybrid particle technology and polymer-coating make these properties even more diverse. Thus, column selection may be a critical risk factor which can cause inaccurate result of organic impurities in establishment of CRSNPs. Determination of emodin and chrysophanol in Rhei Radix et Rhizoma of Liuwei Anxiao Capsule is one typical example to demonstrate the influence of column selection ^[1]. Though all the columns used belong to the C₁₈ class, some of them (e.g. Zorbax Eclipse XDB-C₁₈, Symmetry C₁₈) gave incorrect results due to the co-elution of emodin and the interfering acid product of Liuwei Anxiao Capsule. However, up to date, few attention has been paid to the risk caused by column selection in CRS establishment. Thus, it’s very necessary to establish a scientific and efficient approach to aid rational column selection from enormous brands in purity determination of CRSNPs.

Characterization of RP-LC columns was of high interest in the last decade. Several research groups published articles about the validated characterization systems for chromatographic columns, e.g. Hoogmartens ^[2-5], Dolan/Snyder ^[6-11]and Euerby ^[12] system. Reference ^[13] gave detailed descriptions about the above systems.

Hoogmartens system is based on four parameters: k′_amylbenzene, rk′_{benzylamine/phenol, pH2.7}, k′_{2, 2′-dipyridyl} and rk′_{triphenylene/O-terphenyl}, determined with three reproducible chromatographic conditions ^[2-5]. The parameters of more than 100 types of columns (mostly RP-LC C₁₈) are available on the website: http://pharm.kuleuven.be/farmanalyse/ccs/ccs-app, including free software to compare the selectivity difference between the reference one (ref, e.g. a reported suitable column) and any other one in the database (i) or between two columns of interest based on the “F-value” (see Equation (1)):

(1)

in which k′_amylbenzene, rk′_{benzylamine/phenol, pH 2.7}, k′_{2, 2′-dipyridyl} and rk′_{triphenylene/O-terphenyl} stand for the capacity factor of amylbenzene, the relative capacity factor of benzylamine/phenol at pH 2.7, the capacity factor of 2, 2′-dipyridyl and the relative capacity factor of triphenylene/O-terphenyl, respectively. If F≤2, the columns to be compared are predicted to obtain similar separations. If F≥6, they are expected to get significantly different separations.

Dolan/Snyder system is the most comprehensive scientific one which uses five column parameters to characterize column selectivity: H, hydrophobicity; S, steric resistance; A, hydrogen-bond acidity; B, hydrogen-bond basicity; and C, cation-exchange capacity, which includes C_{pH 2.8} and C_{pH 7.0}. Eighteen test solutes (originally 67 ones) with different properties (neutral/acidic/basic, size, shape, polarity, hydrogen bonding, etc.) are used to measure H, S, A, B and C (C_{pH 2.8}, C_{pH 7.0}) with pH 2.8 and 7.0 mobile phases, respectively. The parameters of 664 columns including C₄, C₈, cyano, phenyl ones (PQRI database) have been available in USP website (http://www.usp.org/app/USPNF/columnsDB.html). Selectivity difference among different columns is based on F_s (selectivity function) calculated as follows:

(2)

in which H₂/H₁, S₂/S₁, A₂/A₁, B₂/B₁ and C₂/C₁ stand for the H, S, A, B and C values of columns 1 and 2 to be compared. Criteria for similar or orthogonal columns are F_s≤2 (similar) and F_s≥6 (orthogonal), respectively. The F_s value can be influenced by the presence of acid/basic solute(s) and the pH value of mobile phase. If no acid or basic solute(s) is present, B and C parameters can be ignored. If only acid one(s) is available, C can be disregarded.

Euerby system (ACD (advanced chemistry development, Inc) column selector) involves six parameters (rk′_{benzylamine/phenol, pH 2.7}, rk′_{benzylamine/phenol, pH 7.6}, rk′_{caffeine/phenol}, rk′_{amylbenzene/butylbenzene}, k′_amylbenzene, rk′_{triphenylene/o-terphenyl}) based on the methods described by Tanaka and incorporates PCA (principal component analysis), and its characterization results showed good correlation with Hoogmartens system ^[13]. Besides PQRI database, USP website provides another approach for selecting columns of equivalent selectivity: use SRM870 (a mixture of uracil, toluene, ethyl benzene, quinizarin and amitriptyline) to obtain Hy (capacity factor of ethyl benzene), CTF (tailing factor of quinizarin), CFA (capacity factor of amitriptyline), TFA (tailing factor of amitriptyline) and BD (bonding density) parameters of columns to be examined. The parameters of 117 columns have been available in USP website (http://www.usp.org/app/USPNF/columnsDB.html).

In this paper, column characterization databases were originally attempted in organic impurities determination of Chinese national N-trans-p-coumaroyltyramine and (-)-epicatechin-3-O-gallate CRSNPs to establish possible rational column selection approach. As Hoogmartens system shows good correlation with Euerby one and is much easier to evaluate the parameters of columns unavailable in the database, therefore in our work Hoogmartens database was preferred, meanwhile PQRI and USP databases were also tried.

1 Experimental

1.1 Reagents and samples

Acetonitrile and methanol were HPLC-grade, purchased from Fisher Scientific (Pittsburg, USA). Analytical grade phosphoric acid and formic acid were supplied by Sinopharm Chemical Reagent Beijing Co., Ltd (Beijing, China). Chinese national (-)-epicatechin-3-O-gallate and N-trans-p-coumaroyltyramine CRSNPs (batch 1) were obtained from Shanghai Standard Biotechnology Co. Ltd (Shanghai, China) and Guizhou Niusida Company (Guizhou, China), respectively. A milli-Q water purification system (Millipore, Milford, USA) was used to further purify glass-distilled water.

1.2 Instrumentation

The e2695 Waters HPLC system consisted of a quaternary pump, a vacuum degasser, a column heater and a photo diode array (PDA) detector, controlled by Empower^® software (version 3) (Waters, Milford, USA).

1.3 Chromatographic conditions

1.3.1 Determination of organic impurities of N-trans-p-coumaroyltyramine CRSNPs

C₁₈ columns and gradient elution, using methanol as mobile phase A and 0.1% (v/v) formic acid as mobile phase B at a flow rate of 1.0 mL/min, were applied: 0-20 min, 30%A→55%A; 20-30 min, 55%A; 30-35 min, 55%A→30%A; 35-50 min, 30%A. Samples were prepared in methanol at a concentration of about 1.0 g/L. The injection volume was 20 μL. Ultraviolet detection was performed from 190 to 400 nm (extraction: 300 nm) ^[14].

1.3.2 Determination of organic impurities of (-)-epicatechin-3-O-gallate CRSNPs

Mobile phases A and B contained acetonitrile-0.2% (v/v) phosphoric acid in the ratios of 1:9 (v/v) and 8:2 (v/v) respectively. The flow rate was 1.0 mL/min and the gradient was: 0-10 min, 100% A; 10-30 min, 100%A→90%A; 30-50 min, 90%A→100%A; 50-60 min, 100%A. Samples were prepared in methanol at a concentration of 1.0 g/L. The injection volume was 20 μL. Ultraviolet detection was performed from 190 to 400 nm (extraction: 280 nm). C₁₈ column should be used in the determination, maintained at 40 ℃ ^[15]. The extracted ultraviolet spectra and peak areas of the eluted organic impurities were applied to identify their relative elution order in different C₁₈ columns.

2 Results and discussion

2.1 Column selection approach

To minimize the risk to obtain different values caused by use of various brands of columns in purity determination of CRSNPs, two columns with orthogonal selectivity (F≥6) would be selected from Hoogmartens database. In organic impurity determination of N-trans-p-coumaroyl tyramine and (-)-epicatechin-3-O-gallate CRSNPs, Zorbax 300 SB-C₁₈(Agilent) and XTerra RP 18 (Waters) were applied with F=8.489＞6. They were both 250 mm×4.6 mm, 5μm.

Then their separation results were compared in term of number, elution order and total amount of organic impurities. If there is no significant difference between their results, then the purity result is cross-validated. Otherwise, a third column has to be selected from the database to confirm the above results. Its selectivity should be similar to the above one which gave better separation (F≤2). In the chromatographic purity determination of both N-trans-p-coumaroyltyramine and (-)-epicatechin-3-O-gallate CRSNPs, better separation results were obtained on Zorbax 300 SB-C₁₈(Fig. 1). Thus, based on Hoogmartens database, Zorbax Eclipse XDB-C₁₈ (250 mm×4.6 mm, 5μm, Agilent) and Capcell Pak C₁₈ MG (250 mm×4.6 mm, 5μm) (Shiseido, Japan) were selected, respectively. Their corresponding selectivity differences compared with Zorbax 300 SB-C₁₈(F) were 0.375 and 0.559.

Fig. 1

Fig. 1 Chromatograms of N-trans-p-coumaroyltyramine CRSNPs obtained on different brands of C18columns a. Zorbax 300SB-C18(Serial No. (S/N): USHH004485); b. XTerra RP 18 (S/N: AKAD09948); c. Zorbax Eclipse XDB-C18(S/N: USNH038818). Their sizes were all 250 mm×4.6 mm, 5μm.

In column selection, PQRI and USP databases were also tried and it was found that their column characterization results were different from those of Hoogmartens’, as indicated in Table 1. Using the Zorbax 300 SB-C₁₈ column as the reference column, the F_s values of the applied columns based on PQRI database were much higher, which may due to more parameters were involved in its column characterization system, while the F values of the columns obtained with USP database were much closer and its characterization result of XTerra RP 18 was significantly different from those of other two databases. Based on their practical performances of the columns in purity determinations of N-trans-p-coumaroyltyramine and (-)-epicatechin-3-O-gallate CRSNPs (see Figs. 1 and 2), Hoogmartens database is preferred in this paper.

表 1 (Table 1)

Table 1 Comparison of column characterization results based on Hoogmartens, PQRI and USP databases using Zorbax 300 SB-C18 column as the reference column Test Column Hoogmartens database PQRI database (Dolan/Snyder) USP database Zorbax Eclipse XDB-C18 F=0.375 Fs=8.36 F=2.63 Capcell Pak C18 MG F=0.559 Fs=6.13 F=1.31 XTerra RP 18 F=8.489 Fs=15.96 F=1.72 Criteria for orthogonal columns Fs/F≥6 / Criteria for similar columns Fs/F≤2 The smaller the F value is, more similar are the columns.

Table 1 Comparison of column characterization results based on Hoogmartens, PQRI and USP databases using Zorbax 300 SB-C18 column as the reference column

Fig. 2

Fig. 2 Chromatograms of (-)-epicatechin-3-O-gallate CRSNPs obtained on different brands of C18 columns a. Zorbax 300SB-C18 (S/N: USHH004485); b. XTerra RP 18 (S/N: 02313409213650); c. Capcell Pak MG C18 (S/N: AKAD09948). Their sizes were all 250 mm×4.6 mm, 5μm.

For detailed parameters of the above columns and related column characterization systems, please check the corresponding databases: http://pharm.kuleuven.be/pharmchem/column classification (Hoogmartens database), http://www.usp.org/app/USPNF/columnsDB.html (PQRI and USP databases).

2.2 Establishment of the chromatographic purities of N-trans-p-coumaroyltyramine and (-)-epicatechin-3-O-gallate CRSNPs by the column selection approach combined with orthogonal and similar selectivity

As shown in Fig. 1, there were significant differences between Zorbax 300 SB-C₁₈ and XTerra RP 18 (F=8.489＞6) in terms of number and total amount of impurities during purity determination of N-trans-p-coumaroyltyramine CRSNPs. The major impurity (0.8% by peak area normalization) with a relative retention time (RRT) of 1.1 was separated well from the N-trans-p-coumaroyltyramine peak with Zorbax 300 SB-C₁₈ column. While on XTerra RP 18 column, the major impurity co-eluted with the N-trans-p-coumaroyltyramine peak. That explained well why the purity result of N-trans-p-coumaroyltyramine CRSNPs obtained with XTerra RP 18 (99.43%) was 0.8% higher than the one with Zorbax 300 SB-C₁₈ (98.64%). If in the purity determination only XTerra RP 18 column was applied, then inaccurate results would be obtained and later accuracy of the assigned value of the established N-trans-p-coumaroyltyramine CRSNPs would be effected. The combination of orthogonal columns (e.g. Zorbax 300 SB-C₁₈ and XTerra RP 18) could scientifically and efficiently reduce the above risk caused by unsuitable column selection. To further verify the accuracy of the above purity result obtained with the Zorbax 300 SB-C₁₈ column, the Zorbax Eclipse XDB-C₁₈ column with similar properties (F=0.375＜2) was applied. As indicated in Fig. 1, Zorbax Eclipse XDB-C₁₈ indeed gave a similar separation, in accordance with its column characterization prediction, and its purity result was close to that of Zorbax 300 SB-C₁₈ (98.83% and 98.64%, respectively). Thus, the purity result of N-trans-p-coumaroyltyramine CRSNPs got well cross-validated.

Similarly, in the chromatographic purity determination of (-)-epicatechin-3-O-gallate CRSNPs, firstly Zorbax 300 SB-C₁₈ column and XTerra RP 18 column (F=8.489 ＞ 6) were chosen, then Capcell Pak C₁₈ MG column (F=0.559＜2 compared with Zorbax 300 SB-C₁₈ column) were applied. As demonstrated in Fig. 2, XTerra RP 18 column gave a different elution order of the impurities meanwhile Zorbax 300 SB-C₁₈ column and Capcell Pak C₁₈ MG column showed similar separations, in consistent with the prediction results based on Hoogmartens column characterization database. Although a different selectivity was obtained on XTerra RP 18 column, its purity result of (-)-epicatechin-3-O-gallate CRSNPs was similar to those obtained on Zorbax 300 SB-C₁₈ column and Capcell Pak C₁₈ MG column as their normalized peak area percentages were 98.99%, 98.99% and 99.04%, respectively. When calibrated to a 2% dilution of the test solution, the purity results were 98.95%, 98.98% and 99.01%, respectively. Seen the negligible differences, the chromatographic purity of (-)-epicatechin-3-O-gallate CRSNPs was cross-validated in this way.

The column selection approach proposed in our paper has been proved to be effective in the purity determination of N-trans-p-coumaroyloctopamine CRSNPs as well. And it would be applied to more CRSNPs, especially those basic or weak acid ones which may be more sensitive to the differences among various brands of columns.

3 Conclusions

The choice of orthogonal columns (F≥6) based on Hoogmartens database has demonstrated to be an efficient and successful tool for the purity determination of (-)-epicatechin-3-O-gallate, N-trans-p-coumaroyltyramine and N-trans-p-coumaroyloctopamine CRSNPs. This way, potential differences in purity determination of CRSNPs due to different selectivities obtained on different brands of columns may be revealed. Thus, the possible risk to select an unsuitable column in the purity determination of CRSNPs could be greatly reduced. A third column, whose selectivity should be similar to the one showing the better separation in the previous step (F≤2), was introduced in our approach to further verify and cross-validate the accuracy of the assigned content.

参考文献

[1]	Wang M J, Liu Y D, Hu X R, et al. Chinese Journal of Chromatography, 2016, 34 (11): 1077. Doi: 10.3724/SP.J.1123.2016.08008
[2]	Ivanyi T, Vander Heyden Y, Visky D, et al. J Chromatogr A, 2002, 954 : 99. Doi: 10.1016/S0021-9673(02)00166-8
[3]	Visky D, Vander Heyden Y, Ivanyi T, et al. J Chromatogr A, 2003, 1012 : 11. Doi: 10.1016/S0021-9673(03)01142-7
[4]	Dehouck P, Visky D, Vander Heyden Y, et al. J Chromatogr A, 2004, 1025 : 189. Doi: 10.1016/j.chroma.2003.10.089
[5]	Visky D, Haghedooren E, Dehouck P, et al. J Chromatogr A, 2006, 1101 : 103. Doi: 10.1016/j.chroma.2005.09.069
[6]	Gilroy J J, Dolan J W, Carr P W, et al. J Chroamtogr A, 2004, 1026 : 77. Doi: 10.1016/j.chroma.2003.11.043
[7]	Gilroy J J, Dolan J W, Snyder L R. J Chroamtogr A, 2003, 1000 : 757. Doi: 10.1016/S0021-9673(03)00512-0
[8]	Wilson N S, Gilroy J J, Dolan J W, et al. J Chroamtogr A, 2004, 1026 : 91. Doi: 10.1016/j.chroma.2003.11.041
[9]	Wilson N S, Nelson M D, Dolan J W, et al. J Chromatogr A, 2002, 961 : 171. Doi: 10.1016/S0021-9673(02)00659-3
[10]	Snyder L R, Maule A, Heebsh A, et al. J Chromatogr A, 2004, 1057 : 49. Doi: 10.1016/j.chroma.2004.09.063
[11]	Snyder L R, Dolan J W, Carr P W. J Chromatogr A, 2004, 1060 : 77. Doi: 10.1016/S0021-9673(04)01480-3
[12]	Euerby M R, Petersson P. LC-GC Europe, 2000, 13 : 665.
[13]	Nemeth T, Haghedooren E, Noszal B, et al. J Chemom, 2008, 22 (3/4): 178.
[14]	Sun J, Gu Y F, Li M M, et al. China J Chin Materia Medica, 2014, 39 (12): 2301.
[15]	Pharmacopoeia Commission of People's Republic of China, Pharmacopoeia of People's Republic of China[M]. Beijing: Chemical Industry Press, 2015 : 724