Anal. Chem. 2005, 77, 2966-2970
Qualitative and Quantitative Analyses of Ginkgo Terpene Trilactones by Liquid Chromatography/ Sonic Spray Ionization Ion Trap Mass Spectrometry Erqin Chen,*,† Chen Ding,‡ and Robert C. Lindsay‡
Wisconsin Center for Space Automation and Robotics and Department of Food Science, University of Wisconsin-Madison, 545 Science Drive, Madison, Wisconsin 53711
A liquid chromatography/sonic spray ionization mass spectrometry method (LC/SSI-MS) was developed for qualitative and quantitative analyses of ginkgo terpene trilactones. Five ginkgo terpene trilactones were successfully protonated for qualitative and quantitative analyses under the study conditions. The typical ion adducts were identified as (M + H)+, (M + NH4)+, and (M + Na)+. The limits of detection were achieved between 2.5 and 10 ng with RSD of 0.173-4.82% and a linear range of 10-80 ng with R2 ) 0.991-0.999. This method was used to identify and quantify ginkgo terpene trilactones in extractions of ginkgo biloba leaves obtained from three different extraction methods. This is the first completely validated LC/MS method for quantification of ginkgo terpene trilactones. The factors that contributed to reduce the errors of identification and quantification of ginkgo terpene trilactones are systematically reported, and the advantages and disadvantages of LC/MS method in quantitative analysis are also discussed. Ginkgo biloba is known as the oldest gymnosperms tree and the extract of ginkgo leaves has been used for the treatment of respiratory and circulatory disorders for over 500 years.1 Currently, the most widely used ginkgo product, EGb 761, has been standardized with 6.0% terpene trilactones and 24.0% flavonol glycosides. The products of ginkgo leaf extract were one of the topselling herbal supplements in 2002 with the market value of $33 million.2 The reported health benefits of EGb 761 include the improvement of cerebrovascular and peripheral circuitry functions, regulation of platelet activating factor (PAF),3 and antiinflammatory effect.4 In addition, many studies indicated that EGb 761 could alleviate the cerebral symptoms in learning, memory, and behavior.5 * To whom correspondence should be addressed. Phone: (608) 262-4588; fax: (608) 262-9458; e-mail:
[email protected]. † Wisconsin Center for Space Automation and Robotics. ‡ Department of Food Science (1) van Beek, T. A.; Bombardelli, E.; Morazzoni, P.; Peterlongo, F. Fitoterapia 1998, 69, 195-244. (2) Blumenthal, M. HerbalGram 2003, 71. (3) Nunez, D.; Chignard, M.; Norel, X.; Braquet, P.; Benveniste, J. Thromb. Haemostasis 1985, 54, 135-135. (4) Borchers, A. T.; Keen, C. L.; Stern, J. S.; Gershwin, M. E. Am. J. Clin. Nutr. 2000, 72, 339-347.
2966 Analytical Chemistry, Vol. 77, No. 9, May 1, 2005
Although many publications have constantly reported the therapeutic effects of ginkgo products on the basis of clinical research for decades, the molecular mechanisms for the effect are ambiguous. Hence, lately, there is an increasing interest in investigation of the therapeutic mechanisms of ginkgo extract.6 The scientific community has been suspecting that ginkgo terpene trilactons in ginkgo extracts could contribute to the therapeutic effects. Therefore, a qualitative and quantitative analytical method with high sensitivity, accuracy, and reproducibility will benefit the future research in exploring the therapeutic mechanisms of ginkgo terpene trilactones. Ginkgo terpene trilactones (TTLs) are a group of bioactive compounds exclusively found in ginkgo trees. These include ginkgolides (G-A, G-B, G-C, G-J) and bilobalide (BB) (Figure 1). Analysis of TTLs was technically difficult because of their poor UV absorbance and difficulties in extraction from the tissue matrix. In 1991, van Beek et al. reported a validated RP-HPLC method,7 which for the first time significantly improved the detection limit by using an RI detector, and propelled the RP-HPLC method to a practical technique for quantifying TTLs. To date, HPLC coupled with refractive index detector (RI), evaporative light scattering detection (ELSD), or mass spectrometry (MS) has significantly enhanced detection sensitivity of TTLs quantification. However, in contrast with those detection techniques, the extraction technique of TTLs remains problematic. Despite many publications discussing the extraction, separation, and the quantification of TTLs during the past 20 years, time-consuming extraction procedures, poor reproducibility, and serious matrix interference were the major obstacles for the quantitative analysis of TTLs in ginkgo leaves.8 Therefore, development of an effective extraction method has become a high priority to improve the quantification of TTLs. A new extraction method based on the ionization character of ginkgolide B9 was reported by our group,10 which had high selectivity to TTLs and significantly reduced the interference from (5) Gertz, H. J.; Kiefer, M. Curr. Pharm. Des. 2004, 10, 261-264. (6) Stromgaard, K.; Nakanishi, K. Angew. Chem., Int. Ed. 2004, 43, 16401658. (7) van Beek, T. A.; Scheeren, H. A.; Rantio, T.; Melger, W. C.; Lelyveld, G. P. J. Chromatogr. 1991, 543, 375-387. (8) van Beek, T. A. J. Chromatogr., A 2002, 967, 21-55. (9) Zekri, O.; Boudeville, P.; Genay, P.; Perly, B.; Braquet, P.; Jouenne, P.; Burgot, J. L. Anal. Chem. 1996, 68, 2598-2604. (10) Ding, C.; Chen, E. Q.; Zhou, W. J.; Lindsay, R. C. Anal. Chem. 2004, 76, 4332-4336. 10.1021/ac048510p CCC: $30.25
© 2005 American Chemical Society Published on Web 04/07/2005
Figure 1. Chemical structures of ginkgo terpene trilactones.
impurities. More importantly, the extraction procedure markedly reduced extraction and cleanup time. This will be helpful to pharmaceutical product development and quality control. To confirm the accuracy of this extraction procedure, in this report, we validated the extraction procedure with two other extraction procedures: aqueous/methanol extraction with column cleanup (reference method)7 and enzymatic treatment. TTLs obtained from all three extraction methods were quantified by LC/SSI-MS methods, which could not only improve the sensitivity, but also remove the interference from the impurities that might coelute with TTLs. MATERIALS AND METHODS Materials. Leaves were collected from ginkgo trees on Walnut Street and the arboretum of the University of Wisconsin-Madison (Madison, WI) in August and September 2002 (Table 1, samples 3-7). Fresh ginkgo leaves were washed with distilled water and petioles were removed. After air-drying at room temperature (21 °C), dry ginkgo leaves were pulverized to a fine powder (300420 µm), which was then stored in a desiccator until used. Reference sample (Table 1, sample 1 and 2) was prepared with a mixture of leaf samples from different trees. Authentic ginkgolide A and B, bilobalide, benzyl alcohol, and HPLC grade solvents were purchased from Sigma-Aldrich Chemical Co., St. Louis, MO. Ginkgolide C was kindly provided by Dr. Jian Zhang (Tsinghua University, China). Other reagents were obtained from Fisher Scientific, Chicago, IL. The cellulase, hemicellulase, and pectinase were obtained from ICN Biomedicals Inc., Aurora, OH, Sigma-Aldrich Chemical Co., and EMD Biosciences, Inc., La Jolla, CA, respectively. Double-distilled water (ddH2O) was used in all experiments. High-Performance Liquid Chromatography (HPLC). Chromatographic analysis was performed on a Hitachi liquid chromatograph system (D-7000 model, Hitachi Instrument Inc., San Jose, CA) coupled with an RI detector. The chromatographs were monitored with Hitachi data collection software. Separation was achieved with a C18 analytical column (3 µm Microsorb 3 Spherical, ODS2, 100 × 4.6 mm, Waters Co., Milford, MA) and a guard column (5 µm, ODS2, Waters Co.). All separations in LC/ MS were carried out at 37 °C at a flow rate of 0.5 mL/min. Mobile phase, methanol/water, was used in a gradient program in which methanol concentration was increased from 20 to 40% from 0 to 20 min and degreased to 20% from 20 to 28 min. Mass Spectrometry. Hitachi M-8000 LC/3DQ ion trap mass spectrometer system (Hitachi Instrument Inc.) was used to identify and quantify extracted TTLs. Sonic spray ionization (SSI) source was used for ionization with temperatures of 145, 180, and 120 °C for the cover plate, plate aperture 1, and plate aperture 2,
respectively. The discharge voltages were 0.8 kV, 50, 40, and 450 V for SSI chamber, drift, focus, and detector, respectively. The nitrogen pressure was maintained at 365 kPa. All LC/MS analyses were carried out in the positive standard MS mode from m/z of 200 to 700. Samples used for HPLC analysis were diluted 40 times during LC/MS injection with final concentration of 50% (v/v) methanol and 0.002% (v/v) acetic acid. Extraction Method 1 (Reference Method). The extraction procedure reported by van Beek et al.7 was used as a reference method to monitor accuracy in this study. Extraction Method 2. Leaf sample (0.5 g) and 6 mL of boiling ddH2O in a propylene centrifuge tube (35 mL, Nalgene, Fisher Scientific) were heated in a water bath (100 °C) for 5 min. The supernatant was obtained after centrifuging at 2000g for 15 min. This process was repeated once. In the third extraction, the ddH2O was replaced with a boiling Na2HPO4 solution (0.1% w/w, pH 8.0). The residue was extracted for 15 min and centrifuged to obtain the supernatant. This process was also repeated once. All supernatants were pooled, adjused to pH 4.5, and made up to 25 mL, from which 10-mL aliquots were sequentially partitioned three times with ethyl acetate (10, 5, and 5 mL) by vortexing for 2 min. The organic phases were quantitatively collected and dried at 45 °C with N2 gas in Turbovap evaporator (Zymark Co., Hopkinton, MA). The residue was redissolved in 400 µL of methanol and submitted to HPLC and LC/MS analyses. Benzyl alcohol (final concentration of 0.25-0.5 mg/mL) was added to the samples as an internal standard. Extraction Method 3. Leaf sample (0.5 g) was incubated with three carbohydrases, namely, cellulase, hemicellulase, and pectinase (1%, w/w), in 12 mL of potassium phthalate buffer (0.025 M, pH 5.5) in a water bath at 35 °C for 36 h and then was centrifuged at 2000g for 15 min to separate the supernatant and residue, which was washed twice with 5 mL of Na2HPO4 (0.1%, w/w, pH 8). The supernatants were combined, and pH was adjusted to 4.5. Then, the resulting supernatants were partitioned with ethyl acetate and analyzed by HPLC and LC/MS as method 2. Quantification. In HPLC/RI analysis, both external and internal standards were used for quantification of BB, G-C, G-A, and G-B. The content of G-J was calculated with a mean response factor obtained from G-C as external standards because of commercial unavailability of authetic G-J. Benzyl alcohol was used as an internal standard. In LC/SSI-MS analysis, the peak areas of total ion chromatograms from scan mode were used for quantification. External standard method was used for calculation of BB, G-A, G-B, and G-C and G-J was calculated with standard curve of G-C. Analytical Chemistry, Vol. 77, No. 9, May 1, 2005
2967
a
1 and 2 were two reference samples and 3-7 were different ginkgo leaf samples. b Numbered as in Materials and Methods secrion. c RSD: Refers to HPLC/MS method.
2.80 1.04 1.79 2.89 2.54 3.52 0.173 0.503 0.954 0.757 1.08 0.227 0.204 0.201 0.192 0.206 0.175 0.0752 0.109 0.120 0.0660 0.182 3.18 1.31 1.67 3.26 3.14 2.56 1.04 1.31 2.78 0.950 1.51 0.267 0.282 0.226 0.283 0.274 0.287 0.0741 0.212 0.148 0.0680 0.342 1 2 3 1 2 3 2 2 2 2 2 1 1 1 2 2 2 3 4 5 6 7
1.15 1.17 1.50 1.01 1.10 1.45 0.716 0.807 1.23 0.749 1.94
1.16 1.15 1.17 1.08 1.11 1.11 0.727 0.830 1.23 0.748 1.95
4.82 4.37 3.82 1.68 2.35 2.35 3.88 2.66 4.81 3.20 3.01
0.273 0.279 0.343 0.269 0.279 0.401 0.0814 0.305 0.131 0.0744 0.189
0.254 0.287 0.235 0.278 0.273 0.287 0.0855 0.307 0.129 0.0823 0.190
3.15 1.88 1.16 2.56 2.82 3.24 1.63 1.02 1.70 1.79 3.83
0.275 0.283 0.224 0.270 0.269 0.286 0.0785 0.214 0.143 0.0650 0.354
0.325 0.336 0.335 0.330 0.342 0.308 0.0765 0.267 0.133 0.0990 0.259
0.344 0.349 0.324 0.345 0.326 0.331 0.0789 0.257 0.147 0.108 0.315
2.11 1.83 1.51 1.83 2.65 3.25 1.61 3.51 3.08 0.961 1.47
0.208 0.217 0.205 0.194 0.198 0.189 0.0749 0.116 0.131 0.0610 0.182
RSD (%) G-B
MS RI RSD (%)
G-A
MS RI RSD (%)
G-C
MS RI RSD (%)
G-J
MS RI RSDc (%)
BB
MS RI extraction methodb samplea
terpene trilactones content (mg/g of DW)
Table 1. Contents of TTLs (mg/g) from Different Extraction Methods and Quantified by RP-HPLC/RI and LC/SSI-MS Techniques
2968 Analytical Chemistry, Vol. 77, No. 9, May 1, 2005
Figure 2. The effects of hydrogen ion on BB ionization by SSI.
Figure 3. Effect of SSI chamber voltage on the ionization of BB, G-C, G-A, and G-B.
RESULTS AND DISCUSSION Ionization and Characterization of Adducts of TTLs. Ionization of TTLs by SSI was significantly impacted by hydrogen ion coexisting in samples and SSI chamber voltage. BB was one of the most difficult to be protonated, especially in the absence of acetic acid, while G-A was one of the most easily obtainable adducts at almost any conditions for qualification purpose. The effect of coexisting hydrogen ion in test sample on BB ionization was shown in Figure 2. Peak areas of BB standard at concentration of 20 ng in the present of acetic acid (0.002%, v/v) were about the same as that of 200 ng in the absence of acetic acid. BB could be undetectable at a concentration range of 20-100 ng when the qualitative analysis was carried out in the absence of acetic acid. The ionization of other TTLs was relatively easy compared to BB, though coexisting hydrogen ions could also improve the protonation. Therefore, 0.001-0.003% (v/v) of acetic acid was always recommended for both qualitative and quantitative analyses. All results reported in this paper were measured with the presence of 0.002% (v/v) acetic acid in the sample. In terms of the effects of SSI parameters on the ionization of the target compounds, cover plate temperature at 145-150 °C and plate aperture 1 temperature at 170-180 °C were suitable for ionization of all TTLs. The ionization was significantly impacted by SSI chamber voltage. The optimized voltage for TTLs ionization was 0.8 kV. When the SSI chamber voltage was higher or lower than 0.8 kV, the signal could be reduced significantly. For instance, when SSI chamber voltage was at 7.5 kV, the peak areas of a 60-ng standard TTL were reduced by about 86.2, 62.2, 57.5, and 62.9% for BB, G-C, G-A, and G-B, respectively (Figure 3). Once again, BB was the most affected by the change of SSI chamber voltage; BB signals were absent at 0.5 kV or over 1.0 kV. Special attention should be paid
Figure 4. Total ion chromatograms and standard mass spectra of BB, G-J, G-C, G-A, and G-B obtained by LC/SSI-MS. RT: retention time.
Figure 5. Standard curves of G-A, G-B, G-C, and BB by HPLC/ SSI-MS.
to identification of BB in low-content biological samples in medical investigations. Otherwise, BB could be incorrectly reported as absent when it is actually present in the extracts. The ion adducts of all TTLs obtained by SSI ionization were (M + H)+, (M + NH4)+, and (M + Na)+ (Figure 4). The allotment of individual adduct showed arbitrarily. (M + NH4)+ (m/z 349) was the dominant adduct of BB; however, to other TTLs, either (M + NH4)+ or (M + Na)+ could be the dominant adduct even on identical ionization conditions. Therefore, it is important to use the peak areas of total ion for quantitative analyses of TTLs. Quantitative Analysis of TTLs by LC/SSI-MS. The limit of detection (LOD) of ginkgo terpene trilactones is 2.5-10.0 ng under the study conditions. G-A confers the lowest LOD of 2.5 ng while BB had the highest at 10 ng. Typically, the linear range is 10-80 ng (Figure 5). For BB, a linear range of 10-150 ng could be possibly achieved by carefully manipulating all conditions. For other TTLs, concentrations over 80 ng could result in a serious error. This linear range was very narrow and inconvenient for quantitative analysis. According to the previous report, the reproducibility of quantifying TTLs by MS was not ideal with RSD over 17%.11 In this study, RSD of 0.173-4.82% was achieved when a simultaneous standard curve was performed. The linear range and dilution were well-controlled. However, deviation would (11) Li, X. F.; Ma, M. S.; Scherban, K.; Tam, Y. K. Analyst 2002, 127, 641-646.
noticeably increase if any conditions described above were not satisfied, especially when using a nonsimultaneous standard curve under the same experimental condition. Calculations for Quantification. It was a common technique to use an internal standard for quantification in HPLC/RI when authentic compounds were available. For instance, studies in our laboratory indicated that any TTL standards could be used for quantification of G-J, given the result that agreed with that obtained from using benzyl alcohol as an internal standard in reference method. In the LC/MS method, however, quantification of G-J was difficult without its authentic standard. In previous reports, G-J concentration was calculated by using G-B.12,13 In this study, we found that using G-C as an internal standard to calculate the content of G-J was more accurate. Results for Extraction Method Validation. Results for the HPLC/SSI-MS method used to quantify TTLs from three extraction methods are reported in Table 1. Compared to the results obtained from the HPLC/RI method, the amounts of BB and G-J detected by RI and MS in enzyme treatment method were not in agreement. The amounts quantified by RI were significantly higher than that by MS. This indicated that the impurities of ginkgo leaf extracts from enzyme treatment were coeluted and overlapped. All other TTLs, quantified by both RI and MS, were consistent in three extraction methods. CONCLUSION The validated results indicated that the sequential aqueous extraction method developed in our laboratory is suitable for extraction of TTLs. The extraction method coupled with RPHPLC/RI method is a reliable routine method for extraction and quantification of TTLs. Although LC/MS technique is able to deliver a higher sensitivity for quantification, the method is not recommended for routine use in ginkgo phytopharmaceutical product development. The main reason is the sufficient sensitivity of HPLC/RI method for quantification of TTLs in the natural (12) Jensen, A. G.; Ndjoko, K.; Wolfender, J. L.; Hostettmann, K.; Camponovo, F.; Soldati, F. Phytochem. Anal. 2002, 13, 31-38. (13) Mauri, P.; Migliazza, B.; Pietta, P. J. Mass Spectrom. 1999, 34, 13611367.
Analytical Chemistry, Vol. 77, No. 9, May 1, 2005
2969
accumulation. The contents of TTLs in dry leaves are 0.1-1.5 mg/ g, which gives concentrations of 0.1-0.6 mg/mL in the injected sample. The limit of detection of HPLC/RI reached 0.005-0.01 mg/mL. The second reason is the high cost associated with LC/ MS assay. Last, the linear range of LC/MS is so narrow that it is liable to result in a dilution error and it is hard to obtain good reproducibility. This LC/MS method is practical for TTLs analysis
2970
Analytical Chemistry, Vol. 77, No. 9, May 1, 2005
in low-content biological samples from clinical trials and other medical investigations, especially in the exploration of therapeutic mechanisms. Received for review October 7, 2004. Accepted February 17, 2005. AC048510P
