Hyphenation of Solid-Phase Extraction with Liquid Chromatography

Apr 29, 2005 - CYANOS: A Data Management System for Natural Product Drug Discovery Efforts Using Cultured Microorganisms. George E. .... Identificatio...
0 downloads 8 Views 245KB Size
Download PDF
Anal. Chem. 2005, 77, 3547-3553

Hyphenation of Solid-Phase Extraction with Liquid Chromatography and Nuclear Magnetic Resonance: Application of HPLC-DAD-SPE-NMR to Identification of Constituents of Kanahia laniflora Cailean Clarkson,† Dan Stærk,† Steen Honore´ Hansen,‡ and Jerzy W. Jaroszewski*,†

Department of Medicinal Chemistry and Department of Analytical Chemistry, The Danish University of Pharmaceutical Sciences, Universitetsparken 2, DK-2100 Copenhagen, Denmark

The introduction of on-line solid-phase extraction (SPE) in HPLC-NMR has dramatically enhanced the sensitivity of this technique by concentration of the analytes in a small-volume NMR flow cell and by increasing the amount of the analyte by multiple peak trapping. In this study, the potential of HPLC-DAD-SPE-NMR hyphenation was demonstrated by structure determination of complex constituents of flower, leaf, root, and stem extracts of an African medicinal plant Kanahia laniflora. The technique was shown to allow acquisition of high-quality homo- and heteronuclear 2D NMR data following analytical-scale HPLC separation of extract constituents. Four flavonol glycosides [kaempferol 3-O-(6-O-r-L-rhamnopyranosyl)-β-D-glucopyranoside; kaempferol 3-O-(2,6-di-Or-L-rhamnopyranosyl)-β-D-glucopyranoside; quercetin 3-O(2,6-di-O-r-L-rhamnopyranosyl)-β-D-glucopyranoside(rutin); and isorhamnetin, 3-O-(6-O-r-L-rhamnopyranosyl)-β-Dglucopyranoside] and three 5r-cardenolides [coroglaucigenin 3-O-6-deoxy-β-D-allopyranoside; coroglaucigenin 3-O(4-O-β-D-glucopyranosyl)-6-deoxy-β-D-glucopyranoside; 3′O-acetyl-3′-epiafroside] were identified, with complete assignments of 1H and 13C resonances based on HSQC and HMBC spectra whenever required. Confirmation of the structures was provided by HPLC-MS data. The HPLC-DAD-SPE-NMR technique therefore speeds up the dereplication of complex mixtures of natural origin significantly, by characterization of individual extract components prior to preparative isolation work. Natural products continue to play an important role in drug discovery.1-3 Since extracts of natural origin usually contain a range of chemically diverse constituents occurring in varying concentrations, it is important to develop improved methods for analysis of these inherently complex mixtures. These methods should provide detailed structural information that enables iden* Corresponding author. E-mail: [email protected]. Fax: +45 3530 6040. † Department of Medicinal Chemistry. ‡ Department of Analytical Chemistry. (1) Bindseil, K. U.; Jakupovic, J.; Wolf, D.; Lavayre, J.; Leboul, J.; Van der Pyl, D. Drug Discovery Today 2001, 6, 840-847. (2) Newman, D.J.; Cragg, G. M.; Snader, K. M. J. Nat. Prod. 2003, 66, 10221037. (3) Lee, K.-H. J. Nat. Prod. 2004, 67, 273-283. 10.1021/ac050212k CCC: $30.25 Published on Web 04/29/2005

© 2005 American Chemical Society

tification of known as well as previously unknown chemical constituents. During the past decade, the development of hyphenated techniques, particularly HPLC-NMR,4-7 has greatly increased the analytical capabilities in natural product research.8-10 The interfacing of liquid chromatography with NMR spectroscopy enables rapid identification of compounds in complex mixtures without preparative-scale isolation. When applied to the area of natural products, the technique is able to provide useful on-line structural information for complete or partial structural elucidation, thereby reducing the amount of time and material required for the investigation. Furthermore, the possibility of introducing artifacts by oxidation, solvolysis, and thermal or photochemical degradation is minimized. Although HPLC-NMR has been applied successfully to the characterization of a wide range of natural products,5,8-10 the S/N ratio values in the NMR spectra are often poor as a result of relatively low sample concentrations in the HPLC eluate. The sensitivity of HPLC-NMR has been improved over the years by miniaturization of the flow cells, higher magnetic field strengths, and cryogenically cooled NMR probes and preamplifiers.5,11-13 However, all these advancements still rely on analyte concentrations delivered by the HPLC column. Moreover, the range of HPLC eluents used is still restricted to those suitable for NMR analysis. The most recent advance in HPLC-NMR is the introduction of an on-line solid-phase extraction (SPE) add-on.14-16 (4) Lindon, J. C.; Nicholson, J. K.; Wilson, I. D. J. Chromatogr., B 2000, 748, 233-258. (5) Albert, K. On-line LC NMR and Related Techniques; John Wiley & Sons: West Sussex, U.K., 2002. (6) Corcoran, O.; Spraul, M. Drug Discovery Today. 2003, 8, 624-631. (7) Cardoza, L. A.; Almeida, V. K.; Carr, A.; Larive, C. K.; Graham, D. W. Trends Anal. Chem. 2003, 22, 766-775. (8) Bobzin, S. C.; Yang, S.; Kasten, T. P. J. Ind. Microb. Biotechnol. 2000, 25, 342-345. (9) Wolfender, J.; Ndjoko, K.; Hostettmann, K. Phytochem. Anal. 2001, 12, 2-22. (10) Wolfender, J.; Ndjoko, K.; Hostettmann, K. J. Chromatogr., A 2003, 1000, 437-455. (11) Spraul, M.; Freund, A. S.; Nast, R. E.; Withers, R. S.; Maas, W. E.; Corcoran, O. Anal. Chem. 2003, 75, 1546-1551. (12) Krucker, M.; Lineau, A.; Putzbach, K.; Grynbaum, M. D.; Schuler, P.; Albert, K. Anal. Chem. 2004, 76, 2623-2628. (13) Olson, D. L.; Norcross, J. A.; O’Neil-Johnson, M.; Molitor, P. F.; Detlefsen, D. J.; Wilson, A. G.; Peck, T. L. Anal. Chem. 2004, 76, 2966-2974. (14) Griffiths, L.; Horton, L. Magn. Reson. Chem. 1998, 36, 104-109. (15) Nyberg, N. T.; Baumann, H.; Kenne, L. Magn. Reson. Chem. 2001, 39, 236240. (16) Nyberg, N. T.; Baumann, H.; Kenne, L. Anal. Chem. 2003, 75, 268-274.

Analytical Chemistry, Vol. 77, No. 11, June 1, 2005 3547

In fully automated HPLC-SPE-NMR,17-20 compounds are detected postcolumn by UV or MS and automatically trapped on SPE cartridges, enabling multiple trapping and subsequent transfer of the analytes to the NMR flow probe using a deuterated solvent. The integrated SPE device thus allows the eluted compounds to be concentrated prior to NMR, significantly improving the S/N ratios and facilitating acquisition of 2D NMR data. The aim of this work is to demonstrate the applicability of this novel technology for the rapid characterization of constituents of plant extracts. Kanahia laniflora (Forssk.) R. Br. (Apocynaceae-Asclepidoideae) is distributed in tropical and subtropical Africa and is used by local populations to treat convulsions associated with malaria, headaches, epilepsy, and stomach pain.21-24 Previous phytochemical studies on the roots of K. laniflora showed the presence of small quantities of cardenolides as well as larger amounts of pregnane glycosides.25,26 Chemical tests on different parts of the plant have also suggested the presence of flavonoids.27 Within the same study, the biological evaluation of K. laniflora revealed an indication of sympathetic stimulation, CNS depression, diuretic activity, peripheral vasodilatation, and slight psychotropic activity and metabolic poisoning.27 In the present study, we have applied high-performance liquid chromatography combined with diode array detection, solid-phase extraction, and NMR spectroscopy (HPLC-DAD-SPE-NMR) for the efficient identification of the main constituents of crude extracts of various parts of K. laniflora. EXPERIMENTAL SECTION Chemicals. Acetonitrile-d3 (99.8 atom % of deuterium) and methanol-d4 (99.8 atom % of deuterium) were obtained from Cambridge Isotope Laboratories. Solvents were analytical or HPLC grade and were used as received. Water was purified by deionization and 0.22-µm membrane filtration (Millipore). Plant Material. Plant material of K. laniflora (Forssk.) R. Br. (Apocynaceae-Asclepiadoideae), including roots, stems, leaves, and flowers, was collected in April 2001 in the Machakos disctrict (K4) in Southern Kenya. A voucher specimen (DFHJJ42) was deposited in Herbarium C (Botanical Museum, University of Copenhagen, Denmark). Sample Preparation. For each plant part, 10 g of powdered material was covered with 200 mL of 96% ethanol and left for 24 h at ambient temperature. The solvent was removed by filtration (17) Exarchou, V.; Godejohann, M.; Van Beek, T. A.; Gerothanassis, I. P.; Vervoort, J. Anal. Chem. 2003, 75, 6288-6294. (18) Simpson, A. J.; Tseng, L.-H., Simpson, M. J.; Spraul, M.; Braumann, U.; Kingery, W. L.; Kellher, B. P.; Hayes, M. H. B. Analyst 2004, 129, 12161222. (19) Lambert, M.; Stærk, D.; Hansen, S. H.; Jaroszewski, J. W. Magn. Reson. Chem. In press. (20) Jaroszewski, J. W. In Magnetic Resonance in Food Science: The Multivariate Challenge; Engelsen, S. B., Belton, P. S., Jacobsen, H. J., Eds.; The Royal Society of Chemistry: Cambridge, U.K., In press. (21) Haerdi, F. Acta Trop. Suppl. 1964, 8, 135. (22) Heine, B.; Ko ¨nig, C. Plant Concepts and Plant Use, Part II, Plants of the So (Uganda); Breitenbach Publishers: Saarbru ¨ cken, 1988; p 80. (23) Beentjie, H. Kenyan Trees Shrubs and Lianas; National Museum of Kenya: Nairobi, 1994; p 494. (24) Neuwinger, H. D. African Traditional Medicine, A Dictionary of Plant Use and Applications; Medpharm Scientific Publishers: Stuttgart, 2000; p 287. (25) Kapur, B. M.; Allgeier, H.; Reichstein, T. Helv. Chim. Acta. 1967, 50, 21472171. (26) Kapur, B. M.; Allgeier, H.; Reichstein, T. Helv. Chim. Acta. 1967, 50, 21712179. (27) Kruger, A. M. C.; Gerritsma-Van der Vijver, L. M. S. A. Tydskr. Nat. Tegnol. 1986, 5, 46-52.

3548

Analytical Chemistry, Vol. 77, No. 11, June 1, 2005

and the extraction procedure repeated with a fresh portion of the solvent. The combined extracts were concentrated below 40 °C and solvent residues removed on a freeze-dryer. The amounts of crude extracts obtained from the roots, stems, leaves, and flowers were 0.6, 0.6, 0.7, and 0.4 g, respectively. The crude extracts were dissolved in H2O/CH3CN (35:65) to a concentration of 50 mg/ mL and the solutions applied to conditioned C18 SPE cartridges (8.5 × 2.0 cm i.d., 5 g of the sorbent, Varian BondElut). The cartridges were eluted with 100 mL of H2O/CH3CN (35:65), and any remaining plant extract was dissolved in H2O/CH3CN (5:95), applied to the same cartridge, and eluted with 100 mL of H2O/ CH3CN (5:95). The collected eluates were combined for each plant part, concentrated under reduced pressure below 40 °C, and freeze-dried. The cartridges were rinsed with 100 mL of CH3CN followed by 100 mL of CHCl3 and conditioned with 100 mL of CH3OH and then with 100 mL of H2O/CH3CN (35:65) before being reused. The SPE-purified extracts were dissolved in CD3OD and analyzed by 600-MHz 1H NMR in 5-mm tubes. Solutions (40 mg/ mL) of the SPE-purified extracts for HPLC and HPLC-SPE-NMR analysis were prepared in H2O/CH3CN (35:65) and centrifuged in a microcentrifuge for 5 min to remove any particulate matter. HPLC Separations. Optimization of HPLC separations were carried out at 40 °C on a 150 × 4.6 mm i.d., C18(2) Phenomenex Luna column (3 µm, 100 Å) using a Shimadzu HPLC system consisting of a SCL-10A system controller, a SIL-10AD autoinjector, a LC-10AT pump, and a SPD-M10A DAD, operated with Shimadzu Class-VP ver. 6.10 software. The extracts were separated using mixtures of water (A) and acetonitrile (B). The eluent flow rate was maintained at 0.8 mL/min, and injections volumes were 25 µL. The optimized linear gradient elution program for root and stem extracts was as follows: 0 min, 18% B; 40 min, 34% B; 60 min, 86% B; 62 min, 18% B, followed by a 10-min conditioning period. The leaf and flower extracts were separated using a linear gradient programmed as follows: 0 min, 18% B; 20 min, 44% B; 25 min, 18% B, and 10 min conditioning. HPLC-DAD-SPE-NMR Experiments. The system consisted of a Bruker LC22 quaternary solvent delivery pump equipped with a degasser (Degases Populaire) and operating at 0.8 mL/min, a column oven (Agilent 1100), an autosampler (Agilent 1100), and a Bruker DAD UV-visible detector. A Knauer K100 Wellchrom makeup pump (flow rate 1.0 mL/min) diluted the postcolumn flow with water before peak trapping using a Prospekt II SPE unit (Spark Holland). The SPE device was coupled to a Bruker Avance 600-MHz spectrometer equipped with a 30-µL inverse 1H{13C} flow probe. HySphere C18 HD SPE cartridges (10 × 2 mm i.d., Spark Holland) were used for trappings based on UV absorption of the eluate at 230 nm. Prior to the trappings, the SPE cartridges were conditioned with 500 µL of CH3CN at 6 mL/min followed by 500 µL of water at 1 mL/min. For each chromatographic peak, a total of eight trappings were performed. The cartridges were dried with a stream of air for 30 min and the analytes eluted with 30 µL of CD3CN into the NMR probe for data acquisition. HPLC separations, SPE peak trapping, and analyte transfer to the NMR spectrometer were controlled by Bruker HyStar ver. 2.3 software, while NMR acquisition and processing were performed using Bruker XWINNMR ver. 3.1 software. For the 1D 1H NMR spectra, 1D nuclear Overhauser effect spectroscopy (NOESY) pulse sequence with presaturation during mixing time (100 ms) and

Figure 1. HPLC profiles of the K. laniflora leaf (A), stem (B), flower (C), and root (D) extracts. The chromatograms were recorded at 230 nm, and the peaks used for HPLC-DAD-SPE-NMR experiments are labeled 1-15.

relaxation delay (2.4 s) was used to suppress residual solvent resonances. Typically, 512 scans were accumulated, recording 64k data points with a sweep width of 30 ppm. The phase-sensitive, gradient-selected homonuclear correlation spectroscopy (COSY) experiments were performed with water suppression enhanced through T1 effects (WET) solvent suppression, collecting 2k × 128 data points with 48 scans per increment in the indirect dimension, and Fourier transformed by applying squared sinebell apodization in both directions. Gradient-selected heteronuclear single quantum coherence (HSQC) spectra with WET solvent suppression and heteronuclear multiple bond correlation (HMBC) spectra were acquired with 128 scans and 1k × 128 or 256 data points. WET solvent suppression sequence included standard 4-fold selective excitation with 20-ms sinusoidal pulses flanked by 1-ms gradient pulses, with GARP 13C decoupling during dephasing and acquisition. The HSQC and HMBC spectra were Fourier transformed using π/2-shifted and unshifted squared sine-bell in both dimensions, respectively. Linear prediction was applied in all 2D experiments to improve resolution in the indirect direction. 1H and 13C chemical shifts were referenced to a residual signal of CH3CN at δ 1.96 and to the signal of CD3CN at δ 1.32 ppm, respectively. HPLC-MS Experiments. The HPLC-MS analysis was performed on an Agilent 1100 LC system consisting of quaternary pump, degasser, autosampler, column oven, and DAD UV detector, interfaced with an Agilent 1100 MSD trap mass spectrometer equipped with an electrospray ionization source and operating in

positive-ion mode; drying temperature, 200 (cardenolides) or 350 °C (remaining compounds); nebulizer pressure, 30 psi; drying gas, 8 L/min; spray voltage, 3500 V; and trap drive, 50-65 (arbitrary units). RESULTS AND DISCUSSION Solid-phase extraction was used for sample preparation in order to remove the most apolar and fatty constituents of the extracts and served as a sample cleanup. The TLC and HPLC fingerprints of the SPE-purified extracts were similar to those generated by liquid-liquid partition between MeOH/H2O (9:1) and light petroleum, with the exception that a greater proportion of the highly hydrophobic constituents were removed by SPE (results not shown). Using UV detection, optimization of HPLC conditions led to satisfactory separation of the main compounds present in the extracts (Figure 1). On initial inspection of the UV traces, the leaf and flower extracts appeared to contain a number of common constituents that varied in concentration (Figure 1A and C). Since the inflorescences contained sepals and bracts, it is proposed that this green tissue is responsible for the similar profiles of the two extracts. The UV traces of the stem and root extracts (Figure 1B and D) showed the presence of a few common constituents with a different polarity compared to those present in the leaf and flower extracts. The major peaks in all four extracts were selected for HPLCDAD-SPE-NMR experiments. The HPLC eluate was monitored by DAD, and absorption thresholds at 230 nm were defined in Analytical Chemistry, Vol. 77, No. 11, June 1, 2005

3549

order to provide start and stop signals for the SPE trappings on C18 cartridges. Postcolumn dilution of the eluate with water (1: 1.25) was applied to increase the trapping efficiency. After eight repeated trappings, the cartridges were dried and eluted with CD3CN into a 30-µL flow probe for the acquisition of 600-MHz NMR data. In the chromatogram of the leaf extract, eight main peaks were selected for trappings (Figure 1A). The 1H NMR spectrum of the major peak (peak 3) showed characteristic signals of a flavonol (3-hydroxyflavone) glycoside with a spin-spin coupling pattern resulting from the presence of oxygen atoms at C4′, C5, and C7. Thus, H6 and H8 formed an AB pattern with Jmeta ) 2.2 Hz, whereas the monohydroxylated ring B exhibited an AA′XX′ pattern with protons A and B centered at δ 8.12 and 6.92. The characteristic singlet at δ 12.26 strongly suggested the presence of a 5-hydroxy group, hydrogen-bonded to the carbonyl group at C4. In addition, the spectrum revealed two free phenol groups at δ 8.26 and 7.88, no methoxy resonances and two anomeric hydrogens at δ 5.02 (d, J ) 8.1 Hz) and 4.51 (d, J ) 1.5 Hz). This indicated the presence of a disaccharide residue, with the remaining task being the elucidation of its structure. The presence of a methyl doublet at δ 1.05 (J ) 5.9 Hz) together with the small coupling constant of the anomeric doublet at δ 4.51 was indicative of a rhamnopyranose moiety, whereas the chemical shift and the coupling constant observed for the other anomeric doublet suggested the presence of a disaccharide unit bound directly to the aglycon moiety with an equatorial glycosidic bond. These data are compatible with 3-(6-O-R-L-rhamnopyranosyl-β-D-glucopyranosyloxy)-4′,5,7-trihydroxyflavone (kaempferol 3-O-rutinoside, 1). This assignment is supported by the chemical shifts of both anomeric hydrogens and of the methyl group, which are sensitive to the position of the intersaccharidic linkage and reflect the substitution of β-glucopyranose with an R-rhamnopyranosyl at O-6.28,29 A HSQC spectrum provided 13C chemical shifts of all protonated carbons of the molecule, and comparison with literature data29,30 confirmed the structure 1. Thus, the HPLC-DADSPE-NMR experiment enabled structure determination of the compound eluted as peak 3, including assignment of all nonquaternary carbon resonances and most of the proton resonances with the aid of HSQC and COSY experiments (Table 1). The 1H NMR and COSY spectra of the compound eluted as peak 1 were similar to that of peak 3 in the aromatic region, but three anomeric protons at δ 5.56 (d, J ) 7.4 Hz), 5.10 (d, J ) 1.5 Hz), and 4.46 (d, J ) 1.5 Hz) as well as two methyl group signals at δ 0.88 (d, J ) 6.2 Hz) and 0.99 (d, J ) 6.2 Hz) were observed (Figure 2 A). The phenol group resonances were also similar to those of 1. This suggested the presence of a kaempferol aglycon with an additional rhamnopyranose moiety. Accordingly, peak 1 had a shorter retention time than peak 3. The chemical shifts of the anomeric protons of the two rhamnopyranose residues, δ 5.10 and 4.46, indicated a 1 f 2 and a 1 f 6 intersaccharidic linkage, respectively. The position of attachment of the rhamnopyranosyl groups also influenced the chemical shift of the C6 methyl protons as expected.29 Full agreement of the observed 1H NMR chemical (28) Lewinsohn, E.; Berman, E.; Mazur, Y.; Gressel, J. Phytochemistry 1986, 25, 2531-2535. (29) Kazuma, K.; Noda, N.; Suzuki, M. Phytochemistry 2003, 62, 229-237. (30) Sang, S.; Lapsley, K.; Jeong, W.-S.; Lachance, P. A.; Ho, C.-T.; Rosen, R. T. J. Agric. Food Chem. 2002, 50, 2459-2463.

3550

Analytical Chemistry, Vol. 77, No. 11, June 1, 2005

shifts to those previously reported29 confirmed the structural assignments of the compound eluted as peak 1 as 3-(2,6-di-O-RL-rhamnopyranosyl-β-D-glucopyranosyloxy)-4′,5,7-trihydroxyflavone (2, Table 1). The 1H NMR spectrum of the leaf extract component eluted as peak 2 was similar to that of 1, except that the AA′XX′ pattern was replaced by a pattern typical of a 1,2,4-trisubstituted benzene: δ 6.93 (d, Jortho ) 8.5 Hz, H5′), 7.66 (dd, Jortho ) 8.5 Hz and Jmeta ) 1.9 Hz, H6′), and 7.76 (d, Jmeta ) 1.9 Hz, H2′), corresponding to a quercetin aglycon. All couplings were verified by a COSY spectrum. The intramolecularly hydrogen-bonded 5-hydroxy group was observed at δ 12.31, in addition to three phenolic hydrogens at δ 8.11, 7.44, and 7.09, further supporting the quercetin structure. The sugar region of the spectrum was identical to that of 1, indicating the presence of the same disaccharide moiety. This was confirmed by a HSQC spectrum that showed the chemical shifts of the protonated carbons, which were in agreement with literature data29 for 3-(6-O-R-L-rhamnopyranosyl-β-D-glucopyranosyloxy)3′,4′,5,7-tetrahydroxyflavone (quercetin 3-O-rutinoside, rutin, 3). The 1H and 13C resonances are reported in Table 1. The 1H, COSY, and HSQC spectra of the compound eluted as peak 4 in the leaf extract chromatogram also showed the presence of a quercetin derivative with the same disaccharide moiety as 1 and 3. However, a three-proton singlet at δ 3.95 demonstrated the presence of a methoxy group (Figure 2B). The C2′ signal (δ 114.7) experienced an upfield shift (-3.7 ppm) compared to that of the corresponding nonmethylated quercetin derivative 3 (Table 1), indicating the methylation site at O3′31-33 and, hence, the presence of a isorhamnetin aglycon. This compound could thus be identified as 3′-methoxy-3-(6-O-R-L-rhamnopyranosyl-β-D-glucopyranosyloxy)-4′,5,7-trihydroxyflavone (isorhamnetin 3-O-rutinoside, 4). HPLC-DAD-SPE-NMR experiments with peaks 1-4 obtained from the flower extract demonstrated the presence of the same compounds as in the leaf extract. Peaks 5-7 in both extracts represented several coeluting, minor constituents. These compounds were flavonoids similar to 1-4, as indicated by the presence of signals corresponding to the hydrogen-bonded 5-hydroxy groups as well as aromatic and sugar resonances. Since flavonol glucosides are known to occur in plants as mixtures of numerous, closely related compounds, and because this type of natural product is unlikely to contribute to the elucidation of the reported pharmacological effects of K. laniflora, peaks 5-7 were not investigated further. The 1H NMR spectrum of the compound eluted as peak 8 contained abundant aliphatic resonances and two singlets at δ 0.86 and 0.88 (Figure 2C), characteristic of the angular methyl groups C18 and C19 in steroids. A pair of diastereotopic geminal protons (δ 4.81 and 4.99, Jgem ) 18.1 Hz) showing allylic coupling to an olefinic hydrogen at δ 5.82 (Jallylic ≈ 1.5 Hz, H22) gave an ABX spectrum characteristic of a γ-lactone moiety in cardenolides. The presence of a singlet at δ 4.48 (acetal group) and a three-proton singlet at δ 2.04 (acetyl group) together with 13C NMR data for (31) Breitmaier, E.; Voelter, W. Carbon-13 NMR Spectroscopy; VCH: Weinheim, 1990; pp 257, 453. (32) Lu, Y.; Sun, Y.; Foo, L. Y.; McNabb, W. C.; Molan, A. L. Phytochemistry 2000, 55, 67-75. (33) Senatore, F.; D’Agostino, M.; Dini, I. J. Agric. Food Chem. 2000, 48, 26592662.

Table 1. Structures of Kanahia laniflora Constituents with 1H and Experiments (CD3CN)

13C

NMR Data Obtained in HPLC-DAD-SPE-NMR

L, leaves; F, flowers; S, stems; R, roots. b Glc, β-D-glucopyranosyl; Rha, R-L-rhamnopyranosyl; All, 6-deoxy-β-D-allopyranosyl. c Obtained from NMR, COSY, and HSQC spectra. d Obtained from HSQC and HMBC spectra.

a 1H

protonated carbons suggested the presence of a general afroside or 3′-epiafroside structure.34 Due to the relative complexity of this compound, COSY and HMBC spectra were acquired for full (34) Cheung, H. T. A.; Chiu, F. C. K.; Watson, T. R. Wells, R. J. J. Chem. Soc., Perkin Trans. 1 1983, 2827-2835.

assignment of the carbon and proton resonances. The signals identified as H2 (δ 3.97) and H3 (δ 3.84), coupled to each other with J2,3 ) 10.0 Hz, indicated that these hydrogens are axial.35,36 The coupling pattern of H3′ (δ 4.75, dd, J ) 12.0 and 4.8 Hz) and (35) Warashina, T.; Noro, T. Phytochemistry 1994, 37, 801-806.

Analytical Chemistry, Vol. 77, No. 11, June 1, 2005

3551

Figure 2. 600-MHz 1H NMR spectra of 2 (A), 4 (B), and 5 (C) from the K. laniflora extracts obtained in the HPLC-DAD-SPE-NMR mode (eight peak trappings). Spectra A and B were recorded with suppression of solvent resonances (H2O and CH3CN/CD2HCN), whereas no solvent peak suppression was applied in (C).

H5′ (δ 3.70, ddq, J ) 12.0, 1.8, and 6.2 Hz) demonstrated that the oxygen attached to C3′ and the 6′-methyl group are both equatorial.35-37 The stereochemistry at C1′ was confirmed by the chemical shift of H1′ reported for nonaromatic solvents,36,37 and corresponds to that of a modified β-D-aldopyranoside. The site of acetylation at O3′ was confirmed by the chemical shift of H3′ and (36) Abdel- Azim, N. S.; Hammouda, F. M.; Hunkler, D.; Rimpler, H. Phytochemistry 1996, 42, 523-529. (37) El-Askary, H.; Ho ¨lzl, J.; Hilal, S.; El-Kashoury, E. Phytochemistry 1993, 34, 1399-1402.

3552 Analytical Chemistry, Vol. 77, No. 11, June 1, 2005

by HMBC data. Thus, peak 8 in the leaf and flower extract represents 3′-O-acetyl-3′-epiafroside34 (5, Table 1). This natural product appears to have been encountered only once prior to this work.34 Separation of the stem and root extracts required different chromatographic conditions compared to the leaf and flower extracts (Figure 1). The chromatograms contained seven and five main peaks, respectively, numbered 9-15 (Figure 1), which were subjected to HPLC-DAD-SPE-NMR analysis. Compounds responsible for peaks 9, 11, and 13-15 were present in both extracts (1H NMR, UV). Peak 10 contained isorhamnetin 3-O-rutinoside (4), already identified as peak 4 in the leaf and flower extract. Peak 9 also contained a flavonol glycoside not investigated in detail. The NMR spectra obtained with peaks 12, 14, and 15 provided insufficient data for structure elucidation. The 1H NMR spectra obtained with peaks 11 (Figure 3A) and 13 both exhibited characteristic features of a 5R-cardenolide with a sugar moiety attached to C3. However, one of the angular methyl groups (C19) was missing and replaced by a hydroxymethyl group, as shown by 1H NMR and HSQC data (Figure 3). The compound eluted as peak 11 showed two anomeric hydrogens at δ 4.62 (d, J ) 8.10 Hz) and 4.37 (d, J ) 7.8 Hz, H1′′), whereas only a single anomeric resonance was observed with peak 13 (δ 4.59, d, J ) 7.96 Hz). Both compounds exhibited a methyl group doublet (δ 1.23 and 1.17 for peaks 11 and 13, respectively; J ) 6.3 Hz) characteristic of a 6-deoxy-β-D-allopyranose moiety common in 5R-cardenolides. The HSQC spectrum obtained with peak 13 provided 13C chemical shifts practically identical to those reported for coroglaucigenin β-D-allopyranoside, i.e., frugoside (6).38 Similarly, HSQC spectrum obtained with peak 11 (Figure 3B) disclosed the presence of a β-D-glucopyranosyl residue attached to C4′ of the 6-deoxyallose moiety, as shown by the chemical shift of C4′ (δ 83.7), changed by +9.5 ppm compared to 6.36 Therefore, the diglycoside eluted as peak 11 was identified as coroglaucigenin 3-O-(4-O-β-D-glucopyranosyl)-6-deoxy-β-D-glucopyranoside 7 (Table 1).39 Although the identification of compounds 1-7 was performed on the basis of detailed interpretation of NMR data alone in order to demonstrate the power of the HPLC-DAD-SPE-NMR technique, molecular masses obtained in separate HPLC-MS experiments fully confirmed the assignments. Thus, ESI spectra obtained in positive-ion mode exhibited MH+ ions and fragments corresponding to the loss of the sugar residues. The MS data are summarized in Table 1. The detection of chromatographic peaks in this work was based on UV absorption. While most organic compounds have some end absorption around 200 nm and can be detected using water/acetonitrile mobile phases, very low-absorbing constituents could give rise to insignificant peak intensities. To make sure that no significant constituents of the K. laniflora extracts have escaped the present analysis, 1H NMR spectra of the crude extracts were acquired in NMR tubes in CD3OD. Inspection of these spectra confirmed the presence of free carbohydrates, which were presumably eluted from the HPLC column with the solvent front. (38) Elgamal, M. H. A.; Hanna, A. G.; Morsy, N. A. M.; Duddeck, H.; Simon, A.; Ga´ti, T.; To´th, G. J. Mol. Struct. 1999, 477, 201-208. (39) Kiuchi, F.; Fukao, Y.; Maruyama, T.; Obata, T.; Tanaka, M.; Sasaki, T.; Mikage, M.; Haque, M. E.; Tsuda, Y. Chem. Pharm. Bull. 1998, 46, 528530.

Figure 3. 600-MHz 1H NMR spectrum (A) and HSQC spectrum (B) obtained in the HPLC-DAD-SPE-NMR mode with peak 11 from the K. laniflora root extract chromatogram (eight peak trappings). Glc, β-D-glucopyranosyl; All, 6-deoxy-β-D-allopyranosyl.

However, no major signals attributable to compounds with flavonoid or cardenolide structures other than 1-7 or other compounds that exhibit resonances outside the carbohydrate region (around δ 3-4) were apparent in the spectra. Therefore, it is believed that the described HPLC-DAD-SPE-NMR analysis allowed identification of all constituents of K. laniflora flowers, leaves, roots, and stems that were present in significant amounts. Although earlier work based on chemical tests indicated the presence of cardenolides and flavonoid glycosides in K. laniflora, this is the first report on the characterization of these constituents in this plant. The failure to obtain satisfactory NMR spectra with peaks 5-7, 9, 12, 14, and 15 is presumably related to the presence of very small amounts of highly absorbing compounds and not to inefficient SPE trapping or SPE cartridge elution, as judged from their polarity reflected by the chromatographic retention times (Figure 1). To date, the majority of HPLC-NMR work in the area of natural products has been done with fractions or extracts that have been enriched for particular classes of compounds. In the present work, the initial preparative SPE step was used only to remove the most highly hydrophobic components, and thus, essentially crude extracts were subjected to the HPLC-DAD-SPE-NMR analysis. The sensitivity gain was illustrated by the ability to acquire HSQC and HMBC NMR data, from which 13C chemical shift data, necessary for structure elucidation of complex natural products, could be obtained. The sensitivity gain is achieved in part by concentrating the analytes present in the chromatographic eluate in a highly sensitive 30-µL NMR flow probe, and in part by accumulation of the analyte by multiple SPE trapping steps from repeated injections. The exchange of the chromatographic solvent to a deuterated solvent prior to the NMR analysis allowed for a more convenient NMR data accumulation. Moreover, multiple chromatographic peaks could be analyzed from the same chromatograms, without peak broadening effects that are commonly observed in stopped-flow methods. The HPLC-DAD-SPE-NMR

technique was used in the present work to achieve full and unambiguous structure elucidation of some of the most complex natural products thus far subjected to HPLC-NMR analysis. CONCLUSIONS The use of HPLC-DAD-SPE-NMR technique was demonstrated to be a major advance for the rapid structural elucidation of a range of complex constituents of a plant extract. Complete structure elucidation and assignment of 1H and 13C NMR signals of complex natural products was shown to be possible from analytical-scale HPLC separation of essentially crude extracts. Although HPLC-SPE-NMR is significantly less sensitive than HPLC-MS, it is superior in its ability to provide detailed structural information, including the possibility of distinguishing between positional isomers and stereoisomers. It is believed that the HPLCDAD-SPE-NMR technique, which is inexpensive and rectifies most of the disadvantages of classical HPLC-NMR methods, will be an increasingly important analytical platform in natural product research and in other areas where rapid structural analysis of complex constituents of mixtures is required. ACKNOWLEDGMENT NMR equipment used in this work was purchased via grants from Apotekerfonden af 1991 (Copenhagen). We thank Ms. E. Lunow Jensen, for running the HPLC-MS experiments on equipment purchased via a grant from the Danish Technical Research Council (grant 56-01-0016). A Postdoctoral fellowship to C.C. from the Drug Research Academy, The Danish University of Pharmaceutical Sciences, is gratefully acknowledged.

Received for review February 2, 2005. Accepted March 30, 2005. AC050212K

Analytical Chemistry, Vol. 77, No. 11, June 1, 2005

3553