Computer-Assisted Optimization of a Gradient HPLC Method for the Separation of Flavonoids

25 Computer-Assisted Optimization of a Gradient HPLC Method for the Separation of Flavonoids Tomasz Baczek, Grazyna Lewandowska, Roman Kaliszan, Department of Biopharmaceutics and Pharmacodynamics, Medical University of Gdansk, Poland, Miroslawa Krauze-Baranowska and Wojciech Cisowski, Department of Pharmacognosy, Medical University of Gdansk, Poland A high-performance liquid chromatography procedure based on the linear gradient elution technique was used to separate flavonoids present in methanolic extracts from Taxus baccata var. elegantissima and Metasequoia glyptostroboides. Optimization of chromatographic separations was obtained by means of a computer. Chromatographic separation parameters were computer optimized and these predicted conditions were then successfully applied to actual separations. Predictive errors were calculated and a satisfactory correlation between the predicted and experimental retention data was found. Computer-assisted optimization with the use of linear gradients is able to provide good chromatographic separations of flavonoids in an easy, inexpensive and reproducible manner. Introduction Flavonoids have recently received increased interest as food and beverage components and additives. The reason is their reported preventative activity in circulatory diseases and post-climacteric osteoporosis (1, 2). Hence, there is a need for a convenient, widely applicable and inexpensive method of their determination. Flavonoids are a class of compounds that are widespread in plants all over the world. Reversed-phase high-performance liquid chromatography (HPLC) is often used to analyse flavonoids either quantitatively or qualitatively (3 6). However, their separation is still a challenge for analysts because it is difficult to find satisfactory chromatographic conditions (i.e., good resolution with reasonable analysis time). Methanolic extracts from Taxus baccata var. elegantissima and Metasequoia glyptostroboides contain various flavonoid aglycones and glycosides, including dimer forms (biflavonoids) (7 9). Gradient elution is generally recommended for the analysis of flavonoid mixtures. The support of a specialist computer program can facilitate the determination of optimal chromatographic conditions for the separation of analytes of interest. The purpose of this study was, therefore, to optimize the HPLC separation of flavonoids present in methanolic extracts from Taxus baccata var. elegantissima and Metasequoja glyptostroboides. A general HPLC optimization procedure proposed by Snyder (10) was employed. Predictions of gradient conditions for the analysed group of compounds were performed as was an evaluation of simulation reliability. Experimental Sample preparation: Methanolic extracts from leaves of Taxus baccata var. elegantissima and Metasequoja glyptostroboides were prepared as follows. Dried and pulverized plant material (5 g) was extracted in Soxhlet apparatus with 2-naphthyl ether to remove ballast substances (lipophilic compounds, chlorophyll, resins etc.). After drying and removal of ether, plant material was extracted in sequence with chloroform and methanol. From the chloroformic and methanolic extracts solvent was distilled under pressure and the dry residue was dissolved in methanol. In this form it was subjected to chromatographic analysis. The following compounds were dissolved in methanol to prepare 0.1% solutions as reference analytes: quercetin- 7-o-glucoside, luteolin-7-o-glucoside, kaemferol-7-o-glucoside, tricetin-3 -oglucoside, apigenin-7-o-glucoside, quercetin-3-rhamnoside, myricetin, quercetin, kaemferol-3-rhainoside, luteolin, kaemferol, apigenin, cupressuflavone,

26 LC GC Europe On-line Supplement Column length Column diameter Table 1: Chromatographic parameters used for computer predictions. 15 cm 0.46 cm Particle size 5.0 µm Pore diameter 10 nm Plate number 13 000 Temperature Mobile phase Flow-rate Dead time, t 0 Dwell volume, V D 30 C Methanol/phosphate buffer ph 3.0 (v/v) 1.0 ml/min 2.12 min 1.93 ml Retention time, t R (min) Analyte Gradient time Gradient time t G 20 min t G 60 min n. i. 1 6.11 7.15 n. i. 2 7.36 9.79 n. i. 3 7.55 10.32 n. i. 4 8.24 12.00 n. i. 5 8.96 14.13 n. i. 6 9.55 16.29 n. i. 7 10.40 18.40 Quercetin-7-O-glucoside 13.15 26.72 Flavonoid glucoside 1 13.47 27.63 Flavonoid glucoside 2 13.92 29.01 Kaemferol-7-O-glucoside 14.32 29.95 n. i. 8 14.53 30.61 Flavonoid glucoside 3 14.88 31.71 Myricetin 15.71 32.13 Quercetin 16.03 33.39 Amentoflavone 19.07 45.09 Bilobetin 19.87 46.29 4 -O-methylamentoflavone 20.67 49.92 7-O-methylamentoflavone 20.91 50.77 Ginkgetin 21.39 52.03 Sciadopitysin 22.69 56.11 Table 2 Retention data obtained for methanolic extract from Taxus baccata var. elegantissima in two initial experiments; linear gradient 5-100% B, t G = 20 and 60 min. amentoflavone, bilobetin 4 -omethylamentoflavone, 7-omethylarnentoflavone, ginkgetin, 2,3- dihydrosciadopitysin, sciadopitysin. Chromatographic conditions: Chromatographic measurements were made on a liquid chromatograph (Merck- Hitachi, Frankfurt-Tokyo, Germany-Japan), consisting of a pump (L-7100), diode-array detector (L-7455), autosampler (L-7200), thermostat (L-7350), membrane degasser (L-7612) and interface (D-7000). The column used was an Inertsil ODS-3, 15.0 0.46 cm, packed with octadecyl-bonded silica, particle size 5 µm (GL Sciences Inc., Shinjuku-ku, Tokyo, Japan). The injected sample volume was 20 µl. All chromatographic investigations were performed at 30 C with the flow-rate 1 ml/min. Methanol was obtained from Przedsiebiorstwo Chemiczne Odczynniki Sp z.o.o. Lublin, Poland. Water was prepared with a Milli-Q Water Purification System (Millipore Corp., Bedford, Massachusetts, USA). Phosphate buffer (20 mm, ph 3.0) was prepared by dissolving the appropriate quantity of sodium dihydrogen phosphate dihydrate (Merck KGaA, Darmstadt, Germany) in pure water, and adjusting the ph with hydrochloric acid 37% (Fluka Chemie AG, Buchs, Switzerland). The buffer ph was measured at 21 C before addition of the organic modifier. The measurements were performed with an HI 9017 ph meter (Hanna Instruments, Bedfordshire, UK). HPLC procedure and computer simulations: Chromatographic analysis of the extracts was performed according to the general optimization scheme recommended in reference 10. First, two initial gradient experiments were performed (for t G at 20 and 60 min, and linear gradient 5 100% B, where B is the organic modifier [methanol]). Retention data for all the identified and most of the unidentified analytes from these initial experiments were input data for computer predictions. Computer simulations were performed with DryLab for Windows, version 2.0 (LC Resources, Walnut Creek, California, USA). Attempts were made to choose the best chromatographic conditions (optimal gradient time and

27 No. Analyte t Rcalc t Rexpt t Rcalc -t Rexpt δϕ δδϕ 1 n. i. 1 6.65 6.67 0.02 0.0005 0.0005 2 n. i. 2 8.53 8.53 0.00 0.0000 0.0005 3 n. i. 3 8.86 8.88 0.02 0.0005 0.0005 4 n. i. 4 9.96 10.00 0.04 0.0011 0.0008 5 n. i. 5 11.24 11.25 0.01 0.0003 0.0005 6 n. i. 6 12.40 12.43 0.03 0.0008 0.0000 7 n. i. 7 13.76 13.79 0.03 0.0008 0.0003 8 Quercetin-7-O-glucoside 18.60 18.64 0.04 0.0011 0.0003 9 Flavonoid glucoside 1 19.14 19.17 0.03 0.0008 0.0000 10 Flavonoid glucoside 2 19.94 19.97 0.03 0.0008 0.0005 11 Kaemferol-7-O-glucoside 20.56 20.61 0.05 0.0014 0.0000 12 n. i. 8 20.94 20.99 0.05 0.0014 0.0005 13 Flavonoid glucoside 3 21.57 21.60 0.03 0.0008 0.0005 14 Myricetin 22.40 22.45 0.05 0.0014 0.0003 15 Quercetin 23.05 23.09 0.04 0.0011 0.0014 16 Amentoflavone 29.14 29.23 0.09 0.0024 0.0041 17 Bilobetin 30.18 30.42 0.24 0.0065 0.0038 18 4 -O-methyloamentoflavone 31.95 32.04 0.10 0.0027 0.0014 19 7-O-methyloamentoflavone 32.41 32.56 0.15 0.0041 0.0008 20 Ginkgetin 33.19 33.31 0.12 0.0033 0.0003 21 Sciadopitysin 35.52 35.56 0.13 0.0035 Mean: 0.0017 0.0009 Table 3 A comparison of predicted and experimental retention data for a separation of methanolic extract Taxus baccata var. elegantissima; linear gradient 5-100% B, t G 35 min. No. Analyte t Rcalc t Rexpt t Rcalc -t Rexpt δϕ δδϕ 1 Quercetin-7-O-glucoside 15.03 15.07 0.04 0.0015 0.0000 2 Flavonoid glucoside 1 15.43 15.47 0.04 0.0015 0.0008 3 Flavonoid glucoside 2 15.99 16.05 0.06 0.0023 0.0000 4 Kaemferol-7-O-glucoside 16.47 16.53 0.06 0.0023 0.0004 5 Flavonoid glucoside 3 17.18 17.23 0.05 0.0019 0.0004 6 Myricetin 18.04 18.08 0.04 0.0015 0.0000 7 Quercetin 18.47 18.51 0.04 0.0015 0.0015 8 Amentoflavone 22.48 22.56 0.08 0.0030 0.0008 9 Bilobetin 23.38 23.44 0.06 0.0023 0.0004 10 4 -O-methyloamentoflavone 24.49 24.56 0.07 0.0027 0.0015 11 7-O-methyloamentoflavone 24.80 24.91 0.11 0.0042 0.0008 12 Ginkgetin 25.38 25.47 0.09 0.0034 0.0008 13 Sciadopitysin 27.02 27.09 0.07 0.0027 Mean: 0.0024 0.0006 Table 4 A comparison of predicted and experimental retention data for a separation of methanolic extract Taxus baccata var. elegantissima; linear gradient 5-100% B, t G 25 min.

28 LC GC Europe On-line Supplement 6.651 8.530 8.860 9.965 0 10 20 30 2.13 1.52 1.9 1.76 2.05 6.67 8.53 8.88 10.00 11.236 12.400 12.43 11.25 11.73 13.758 13.79 18.598 19.144 19.940 20.563 20.940 21.567 22.401 23.052 29.141 30.184 31.950 32.408 0 5 10 15 20 25 30 35 19.17 17.89 18.29 18.64 19.97 20.61 20.99 21.60 22.45 23.09 29.23 30.42 30.72 32.05 32.56 33.31 Figure 1: HPLC separation of methanolic extract from Taxus baccata var. elegantissima obtained with a linear gradient 5 100% B, tg = 35 min: (a) Predicted chromatogram, 15.990 33.195 35.523 35.65 36.29 gradient range). Under the conditions chosen experiments were executed and the results achieved were compared with those obtained by computer simulations (predicted results). Predictive errors were evaluated according to the procedure reported in reference 11. Errors in predicted values of retention time, t R are expressed in terms of an equivalent change (error),, in the volume fraction of the organic modifier,, for the predicted separation. Therefore, the predicted value of t R (k) for a mobile-phase composition should correspond to the correct value of t R (k) for the mobilephase composition. So, in the first part of the calculation, the experimental and predicted retention factors are compared on the basis of the following equation (11): = t R ( /t G ) [1] where is the error in predicted retention factor expressed as an equivalent change in ; t R is the difference between predicted and experimental retention time; is the volume fraction of the strong solvent in the mobile phase; is the change in w during the gradient and t G is gradient time. 15.033 18.040 17.179 22.485 23.383 24.488 24.799 15.429 16.472 18.466 25.383 27.024 0 10 20 30 1.52 1.73 2.11 6.37 7.84 8.08 8.93 9.84 11.25 11.60 12.67 15.07 14.80 14.61 17.23 18.08 15.47 16.80 18.51 10.59 16.53 22.56 23.44 24.56 24.91 25.47 26.37 27.36 16.05 0 5 10 15 20 25 30 Figure 2: HPLC separation of methanolic extract from Taxus baccata var. elegantissima obtained with a linear gradient 5 100% B, tg = 25 min: (a) Predicted chromatogram,

29 The second part of the calculation uses the relationship between differences. Errors in t R for the adjacent peaks i and j can be expressed as (11): = ( ) j ( ) i [2] where is the error in predicted resolution (R s ) because of the errors in the predicted retention expressed as the equivalent change of ; ( ) j and ( ) i are the values of for adjacent bands i and j. The maximum average value of is 0.001, which corresponds to an average error of 0.2 in R s (acceptable for complex samples if one assumes an average value of column plate number, N = 10 000) (11). Results and Discussion Analysis of Taxus baccata var. elegantissima extract: In the Taxus baccata var. elegantissima extract, 13 flavonoids were identified. Among them, six were flavonoid dimers. Chromatographic process parameters (Table 1) and retention times (Table 2) obtained experimentally from two initial experiments (input data) were entered into the DryLab program. Retention time, t R (min) Analyte Gradient time Gradient time t G 20 min t G 60 min n. i. 1 8.93 14.24 n. i. 2 9.60 15.57 n. i. 3 10.80 18.80 n. i. 4 11.31 20.61 n. i. 5 12.56 25.25 Luteolin-7-O-glucoside 13.68 28.05 n. i. 6 14.03 29.17 Tricetin-3 -O-glukoside 14.43 30.35 Apigenin-7-O-glucoside 14.56 30.61 Quercetin-3-rhamnoside 14.91 31.55 n. i. 7 15.52 33.47 Kaemferol-3-α-rhamnoside 15.89 34.53 n. i. 8 16.37 35.65 Luteolin 16.80 36.85 Kaemferol 17.57 39.15 Apigenin 17.76 39.60 n. i. 9 18.80 44.40 Amentoflavone 19.07 45.04 Bilobetin 19.76 47.04 4 -methylcupressuflavone 19.89 47.76 4 -O-methylamentoflavone 20.43 49.23 n. i. 10 20.56 49.55 7-O-methylamentoflavone 20.93 50.75 Ginkgetin 21.36 51.97 2,3-dihydrosciadopitysin 22.19 54.77 Sciadopitysin 22.69 56.05 Table 6 Retention data obtained for methanolic extract from Metasequoia glyptostroboides in two initial experiments; linear gradient 5-100% B, t G equal 20 and 60 min. No. Analyte t Rcalc t Rexpt t Rcalc -t Rexpt δϕ δδϕ 1 Quercetin-7-O-glucoside 6.27 6.77 0.50 0.0396 0.0063 2 Flavonoid glucoside 1 6.67 7.09 0.42 0.0332 0.0063 3 Flavonoid glucoside 2 7.23 7.57 0.34 0.0269 0.0744 4 Kaemferol-7-O-glucoside 6.72 8.00 1.28 0.1013 0.0847 5 Flavonoid glucoside 3 8.40 8.61 0.21 0.0166 0.0008 6 Myricetin 9.35 9.57 0.22 0.0174 0.0024 7 Quercetin 9.73 9.92 0.19 0.0150 0.0111 8 Amentoflavone 13.55 13.60 0.05 0.0040 0.0016 9 Bilobetin 14.44 14.51 0.07 0.0055 0.0000 10 4 -O-methyloamentoflavone 15.50 15.57 0.07 0.0055 0.0127 11 7-O-methyloamentoflavone 15.80 16.03 0.23 0.0182 0.0135 12 Ginkgetin 16.40 16.80 0.40 0.0317 0.0301 13 Sciadopitysin 20.07 20.85 0.78 0.0618 Mean: 0.0290 0.0203 Table 5 A comparison of predicted and experimental retention data for a separation of methanolic extract Taxus baccata var. elegantissima; linear gradient 38-85% B, t G 12 min.

30 LC GC Europe On-line Supplement 6.273 6.673 7.715 7.230 8.403 9.352 9.733 13.554 14.440 15.497 15.798 0 10 20 1.55 1.76 2.05 2.13 2.91 3.20 3.49 3.89 4.21 4.48 4.85 6.13 7.09 6.53 6.77 7.57 6.00 8.97 8.61 9.57 9.92 13.60 14.51 0 5 10 15 20 25 16.404 15.57 16.03 16.80 Figure 3: HPLC separation of methanolic extract from Taxus baccata var. elegantissima obtained with a linear gradient 38 85% B, tg = 12 min: (a) Predicted chromatogram, 11.916 12.965 15.220 16.365 19.248 0 10 20 30 40 2.03 2.11 11.89 12.96 15.23 16.35 18.80 19.28 20.03 21.252 21.35 23.630 21.982 22.774 22.979 22.77 22.99 22.00 29.123 24.891 25.610 26.435 27.258 28.794 31.910 32.378 33.729 34.128 35.445 36.153 35.375 36.984 20.067 20.85 38.774 39.682 0 5 10 15 20 25 30 35 40 45 23.65 24.93 25.63 26.48 27.31 28.85 29.17 31.97 32.43 33.81 34.24 35.47 35.23 36.27 37.12 38.93 39.84 Satisfactory resolution for 5 100% B linear gradient were found at the gradient time t G = 35 min. The comparison of simulated and experimental chromatograms is shown in Figure 1. Calculated predictive errors ( and ) are collected in Table 3. It can be seen that the predictions are very accurate (average = 0.0017 and average = 0.0009). Good resolution for the known flavonoids was achieved using 5 100% B linear gradient at gradient time t G = 25 min. The predicted and experimental retention data are in excellent agreement (Table 4, Figure 2). Average predictive errors were: = 0.0024 and =0.0006. Because the first known compound eluted at a retention time, t R = 15.07 min, attempts were made to accelerate the whole separation. Using the software it was found that the best separation for analysed flavonoids would be achieved with a linear gradient of 38 85% B at t G = 20 min (Figure 3(a)). By increasing the initial concentration of methanol (from 5 to 38%), the retention time for the first fiavonoid was shortened to 6.77 min (Figure 3(b)). However, the accuracy of the whole prediction under such conditions appeared to decrease in comparison with the former analysis (Table 5). This could be the result of a big difference in the initial composition of the mobile phase between that gradient (an initial concentration of methanol: 38%) and the gradients used for obtaining the input data (an initial concentration of methanol: 5%). Analysis of Metasequoja glyptostroboides extract: Sixteen flavonoids were identified, among them eight biflavonoids. Similar to the extract from Taxus baccata var. elegantissima, the retention data from two gradient experiments provided input data to create computer predictions (Table 6). General experimental conditions were the same as with Taxus baccata var. elegantissima (Table 1). It was found that for the linear gradient 5 100% B a satisfactory separation results at a gradient time t G = 40 min (Figure 4(a)). Experimental results are shown in Figure 4(b). A comparison of experimental and Figure 4: HPLC separation of methanolic extract from Metasequoia glytostroboides obtained with a linear gradient 5 100% B, tg = 40 min: (a) Predicted chromatogram,

31 predicted retention data is illustrated in Table 7. It can be seen that average predictive errors are very small ( = 0.0013 and =0.0004). An attempt to shorten the analysis time was undertaken by simulation of a 30 90% B linear gradient at the gradient time t G = 30 min (Figure 5(a)). The use of a higher concentration of methanol in the initial composition of the mobile phase decreased retention times of all the compounds and effectively shortened the analysis time (Figure 5(b)). However, the predictive errors appeared to increase in comparison with those observed in the analysis performed with 5 100% B gradient (Table 8). Conclusions Optimization of the chromatographic separations of flavonoids from Tarus No. Analyte t Rcalc t Rexpt t Rcalc -t Rexpt δϕ δδϕ 1 n. i. 1 11.92 11.89 0.03 0.0007 0.0007 2 n. i. 2 12.96 12.96 0.00 0.0000 0.0002 3 n. i. 3 15.22 15.23 0.01 0.0002 0.0002 4 n. i. 4 16.37 16.35 0.02 0.0005 0.0002 5 n. i. 5 19.25 19.28 0.03 0.0007 0.0007 6 Luteolin-7-O-glucoside 21.25 21.25 0.00 0.0000 0.0005 7 n. i. 6 21.98 22.00 0.02 0.0005 0.0005 8 Tricetin-3 -O-glukoside 22.77 22.77 0.00 0.0000 0.0002 9 Apigenin-7-O-glucoside 22.98 22.99 0.01 0.0002 0.0002 10 Quercetin-3-rhamnoside 23.63 23.65 0.02 0.0005 0.0005 11 n. i. 7 24.89 24.93 0.04 0.0010 0.0005 12 Kaemferol-3-α-rhamnoside 25.61 25.63 0.02 0.0005 0.0005 13 n. i. 8 26.44 26.48 0.04 0.0010 0.0002 14 Luteolin 27.26 27.31 0.05 0.0012 0.0002 15 Kaemferol 28.79 28.85 0.06 0.0014 0.0002 16 Apigenin 29.12 29.17 0.05 0.0012 0.0002 17 n. i. 9 31.91 31.97 0.06 0.0014 0.0002 18 Amentoflavone 32.38 32.43 0.05 0.0012 0.0007 19 Bilobetin 33.73 33.81 0.08 0.0019 0.0007 20 4 -methylcupressuflavone 34.13 34.24 0.11 0.0026 0.0005 21 4 -O-methylamentoflavone 35.14 35.23 0.09 0.0021 0.0007 22 n. i. 10 35.38 35.47 0.09 0.0021 0.0012 23 7-O-methylamentoflavone 36.15 36.27 0.12 0.0028 0.0005 24 Ginkgetin 36.98 37.12 0.14 0.0033 0.0005 25 2,3-dihydrosciadopitysin 38.77 38.93 0.16 0.0038 0.0000 26 Sciadopitysin 39.68 39.84 0.16 0.0038 Mean: 0.0013 0.0004 Table 7 A comparison of predicted and experimental retention data for a separation of methanolic extract from Metasequoia glyptostroboides in two initial experiments; linear gradient 5-100% B, t G 40 min.

32 LC GC Europe On-line Supplement 3.231 4.103 5.790 6.611 9.290 11.563 14.280 12.388 13.295 13.529 15.744 16.581 17.526 18.483 0 10 20 30 20.666 20.288 24.151 24.687 26.282 26.796 27.985 28.250 29.185 30.162 32.327 33.377 baccata var. elegantissima and Metasequoia glyptostroboides can be performed with the use of a computer program. Optimizing the chromatographic separation of analytes in a complex mixture can be realized without the need for traditional trial-and-error methods. The use of linear gradients provides good chromatographic separations of flavonoids in accordance with simulations predicted by means of a computer program. The results obtained confirm the unique potential of the optimized gradient approach to separate components of biological samples as has independently been observed by other authors (12, 13). 1.47 1.76 3.44 4.27 5.71 6.48 9.15 10.03 11.41 13.33 13.09 12.27 14.13 15.97 15.63 16.48 17.49 18.43 0 5 10 15 20 25 30 35 20.27 20.64 24.19 24.72 28.05 26.32 26.85 28.32 29.25 30.24 32.45 33.49 Figure 5: HPLC separation of methanolic extract from Metasequoia glytostroboides obtained with a linear gradient 30 90% B, tg = 30 min: (a) Predicted chromatogram, References (1) H.-J. Wang, and P.A. Murphy, J. Agric. Food Chem., 42 (1994) 1666 1673. (2) D.E. Pratt and P.M. Birac, J. Food Sci., 44 (1979) 1720 1722. (3) H. Schulz and G. Albroscheit, J. Chromatogr., 442 (1988) 353 361. (4) O. Sticher, Planta Med., 59 (1993) 2 11. (5) P. Pietta et al., J. Chromatogr., 553 (1991) 223 231. (6) M. Keinänen and R. Julkunen-Tutto, J. Chromatogr. A, 793 (1998) 370 377. (7) P. Pieffa, P. Mauri and A. Rava, J. Chromatogr., 437 (1988) 453 456. (8) A. Hasler, O. Sticher and B. Meier, J. Chroniatogr., 605 (1992) 41 48. (9) O. Sticher, Planta Med., 59 (1993) 2 11. (10) L.R. Snyder, J.J. Kirkland and J.L. Glajch, Practical HPLC Method Development, John Wiley & Sons, New York, 1997. (11) J.W. Dolan et al., J. Chromatogr. A, 857 (1999) 41 68. (12) P.L. Zhu et al., J. Chromatogr. A, 756 (1996) 63 72. (13) J.W. Dolan et al., J. Chromatogr. A, 803 (1998) 33 50.

33 No. Analit t Rcalc t Rexpt t Rcalc -t Rexpt δϕ δδϕ 1 n. i. 1 3.23 3.44 0.21 0.0050 0.0010 2 n. i. 2 4.10 4.27 0.17 0.0040 0.0021 3 n. i. 3 5.79 5.71 0.08 0.0019 0.0012 4 n. i. 4 6.61 6.48 0.13 0.0031 0.0002 5 n. i. 5 9.29 9.15 0.14 0.0033 0.0002 6 Luteolin-7-O-glucoside 11.56 11.41 0.15 0.0036 0.0007 7 n. i. 6 12.39 12.27 0.12 0.0028 0.0019 8 Tricetin-3 -O-glukoside 13.29 13.09 0.20 0.0048 0.0000 9 Apigenin-7-O-rhamnoside 13.53 13.33 0.20 0.0048 0.0012 10 Quercetin-3-rhamnoside 14.28 14.13 0.15 0.0036 0.0010 11 n. i. 7 15.74 15.63 0.11 0.0026 0.0002 12 Kaemferol-3-α-rhamnoside 16.58 16.48 0.10 0.0024 0.0014 13 n. i. 8 17.53 17.49 0.04 0.0010 0.0002 14 Luteolin 18.48 18.43 0.05 0.0012 0.0007 15 Kaemferol 20.29 20.27 0.02 0.0005 0.0002 16 Apigenin 20.67 20.64 0.03 0.0007 0.0002 17 n. i. 9 24.15 24.19 0.04 0.0010 0.0002 18 Amentoflavone 24.69 24.72 0.03 0.0007 0.0002 19 Bilobetin 26.28 26.32 0.04 0.0010 0.0005 20 4 -O-methylcupressuflavone 26.80 26.85 0.05 0.0012 0.0005 21 4 -O-methylamentoflavone 27.98 28.05 0.07 0.0017 0.0000 22 n. i. 10 28.25 28.32 0.07 0.0017 0.0002 23 7-O-methylamentoflavone 29.18 29.25 0.07 0.0017 0.0002 24 Ginkgetin 30.16 30.24 0.08 0.0019 0.0010 25 2,3-dihdrosciadopitysin 32.33 32.45 0.12 0.0028 0.0002 26 Sciadopitysin 33.38 33.49 0.11 0.0026 Mean: 0.0024 0.0006 Table 8 A comparison of predicted and experimental retention data for a separation of methanolic extract from Metasequoia glyptostroboides; linear gradient 30-90% B, t G 30 min.