Optimisation of supercritical fluid extraction of indole alkaloids from Catharanthus roseus using...

52 A. VERMA ET AL.

Copyright © 2007 John Wiley & Sons, Ltd. Phytochem. Anal. 19: 52–63 (2008)DOI: 10.1002.pca

Phytochemical AnalysisPhytochem. Anal. 19: 52–63 (2008)Published online 24 July 2007 in Wiley InterScience (www.interscience.wiley.com) DOI: 10.1002/pca.1015

Optimisation of Supercritical Fluid Extraction of IndoleAlkaloids from Catharanthus roseus using ExperimentalDesign Methodology—Comparison with other ExtractionTechniques

ARVIND VERMA, KARI HARTONEN* and MARJA-LIISA RIEKKOLALaboratory of Analytical Chemistry, Department of Chemistry, University of Helsinki, PO Box 55, FIN-00014 University of Helsinki, Finland

Received 31 August 2006; Revised 4 June 2007; Accepted 4 June 2007

Abstract: Response surface modelling, using MODDE 6 software for Design of Experiments and Optimisation, was applied tooptimise supercritical fluid extraction (SFE) conditions for the extraction of indole alkaloids from the dried leaves of Catharanthusroseus. The effects of pressure (200–400 bar), temperature (40–80°C), modifier concentration (2.2–6.6 vol%) and dynamic extrac-tion time (20–60 min) on the yield of alkaloids were evaluated. The extracts were analysed by high-performance liquid chromato-graphy and the analytes were identified using ion trap-electrospray ionisation–mass spectrometry. The method was linear foralkaloid concentration in the range 0.18–31 μg/mL. The limits of detection and quantification for catharanthine, vindoline,vinblastine and vincristine were 0.2, 0.15, 0.1 and 0.08 μg/mL and 2.7, 2.0, 1.3 and 1.1 μg/g, respectively. The dry weight contentof major alkaloids in the plants were compared using different extraction methods, i.e. SFE, Soxhlet extraction, solid–liquid extrac-tion with sonication and hot water extraction at various temperatures. The extraction techniques were also compared in terms ofreproducibility, selectivity and analyte recoveries. Relative standard deviations for the major alkaloids varied from 4.1 to 17.5% indifferent extraction methods. The best recoveries (100%) for catharanthine were obtained by SFE at 250 bar and 80°C using 6.6vol% methanol as modifier for 40 min, for vindoline by Soxhlet extraction using dichloromethane in a reflux for 16 h, and for 3′,4′-anhydrovinblastine by solid–liquid extraction using a solution of 0.5 M sulphuric acid and methanol (3:1 v/v) in an ultrasonic bathfor 3 h. Copyright © 2007 John Wiley & Sons, Ltd.

Keywords: Supercritical fluid extraction; indole alkaloids; Catharanthus roseus; experimental design; liquid chromatography; massspectrometry.

INTRODUCTION

Catharanthus roseus (L.) G. Don is a well-known medi-cinal plant belonging to the family Apocynaceae. It isregarded as a rich source of pharmaceutically importantterpenoid indole alkaloids. Vindoline, catharanthineand 3′,4′-anhydrovinblastine are its major alkaloids(Naaranlahti et al., 1991; Sottomayor et al., 1998) aswell as precursors in the biosynthetic pathways ofvinblastine and vincristine. The two latter alkaloidsare well-known anti-cancer drugs used in the treat-ment of acute leukaemia and Hodgkin’s disease (Noble,1990). 3′,4′-Anhydrovinblastine has been claimed foruse as anti-neoplastic agent and has shown excellentpreliminary results in reducing tumours relating tohuman lung, cervical and colon cancers as well as tonon-Hodgkin’s lymphoma. This compound is being

developed as a new lead drug with improved therapeuticproperties, demonstrating a significantly higher maxi-mum tolerated dose and less toxicity than its parentand related compounds (Schmidt et al., 2003).

Several extraction methods for vinblastine, vincri-stine and 3′,4′-anhydrovinblastine have been developedin the past. Owing to the high economic value of thesedrugs, some of these methods have been patented andused for commercial production by research laboratoriesand pharmaceutical companies. Previously appliedextraction methods involve solid–liquid extraction,using ultrasonic bath with dilute acid or alcohol assolvents, followed by pH control and re-extraction withan organic solvent (Renaudin, 1984; Naaranlahti et al.,1987; Miura et al., 1987). These are long and tediousprocedures employing a large quantity of toxic organicsolvents, which is expensive, hazardous to use andgenerates waste that is costly to dispose of. Therefore,it would be desirable to develop new environmentallyfriendly methods that can be scaled up for commercialproduction.

Supercritical fluid extraction (SFE) technology has beenconsidered a good option for the extraction of naturalproducts, particularly for food and pharmaceutical

PhytochemicalAnalysis

* Correspondence to: K. Hartonen, Laboratory of Analytical Chemistry,Department of Chemistry, University of Helsinki, PO Box 55, FIN-00014 University of Helsinki, Finland.E-mail: [email protected]/grant sponsor: University of Helsinki; Contract/grantnumber: project 2105040.

SUPERCRITICAL FLUID EXTRACTION OF INDOLE ALKALOIDS FROM CATHARANTHUS ROSEUS 53


ingredients (Jarvis and Morgan, 1997; Hamburgeret al., 2004). Supercritical carbon dioxide is the mostcommon fluid used because of its physiological com-patibility, non-toxicity, inflammability, inexpensivenessand availability. In addition, high selectivity, conven-ient critical parameters (Tc = 31.1°C, Pc = 73.8 bar)and environmental friendliness are other significantadvantages of using carbon dioxide as supercriticalfluid.

SFE shows numerous advantages when comparedwith traditional Soxhlet and solid-liquid extraction ofsolid matrices such as plant material. Advantagesinclude (i) faster and more efficient extractions, (ii) noresidual solvent in the final product, which lowers theoperating costs due to the reduction in post-processingand clean-up steps, and (iii) sensitive and relativelynon-volatile compounds can be separated underthermally mild conditions without decomposition(Song et al., 1992). The non-polarity of carbon dioxide isoften a drawback, but this limitation can be overcomethrough the addition of a small amount of a more polarsolvent as modifier (Ollanketo et al., 2001), which under-goes dipole–dipole interaction and hydrogen bondingwith polar functional groups of the compounds ofinterest (Hamburger et al., 2004). This results insignificant increases in the solubilities of polar com-pounds and in their extraction efficiencies.

Several applications employing supercritical carbondioxide with polar modifiers for the extraction ofmoderately polar to polar natural products have beenpublished, including alkaloids (Heaton et al., 1993).Many alkaloids including thebaine, codeine and mor-phine from poppy (Janicot et al., 1990), coronaridineand voacangine from Tabernaemontana catherinensis(Pereira et al., 2004), cocaine from coca leaves (Brachetet al., 2000), purine alkaloids from Maté (Ilex paragua-riensis) (Saldaña et al., 1999), colchicine and colchico-side from Colchicum autumnale (Ellington et al., 2003)and vindoline and vinblastine from Catharanthusroseus (Song et al., 1992, Choi et al., 2002) have beensuccessfully extracted using SFE.

In this work, we tested the applicability of the SFEprocess for the selective extraction of indole alkaloidsfrom freeze-dried leaves of C. roseus using carbon dioxideas supercritical solvent. The compounds were analysedusing HPLC-UV and identified by HPLC-ESI/MS. Aquantitative determination of dry weight content (μg/g)and a comparative account of the relative recoveries (%)of the major alkaloids catharanthine, vindoline and3′,4′-anhydrovinblastine is presented using differentextraction methods, i.e. SFE, Soxhlet extraction, solid–liquid extraction with sonication and hot water extrac-tion at various temperatures. Hot water extraction couldbe an environmentally friendly method for the extrac-tion of plant secondary metabolites such as glycosidesand flavonoids. Kim et al. (2004) used a hot waterextraction method to extract secologanin, a monoter-

penoid glycoside, which is also the precursor in thebiosynthesis of Catharanthus alkaloids. Hot water wasfound to be a more efficient technique than conven-tional organic acid extraction to extract iridoid glyco-sides such as catapol and aucubin from Veronicalonifolia leaves (Suomi et al., 2000). Bergeron et al.(2005) extracted flavonoid glycosides and amino acidswith accelerated solvent extraction (ASE) at 85°C usingwater and hot water extraction techniques. In orderto compare the extraction efficiencies of Catharanthusalkaloids we used hot water extraction at differenttemperatures, as an alternative method.

The aims of the present study were (i) to assess thesuitability of supercritical carbon dioxide for the ex-traction of Catharanthus alkaloids, (ii) to optimise thevariables that affect the extraction, such as pressure,temperature, time of extraction and the amount ofmodifier, using factorial design experiments and (iii)to compare the results with those obtained by otherextraction methods.

EXPERIMENTAL

Reagents and chemicals

Methanol (VWR International AB, Stockholm, Sweden),acetonitrile (far UV) and dichloromethane (Lab-Scan,Analytical Sciences, Ireland) were of HPLC grade.Methyl tert-butyl ether was from Rathburn Chemicals(Walkerburg, Scotland). Ammonium acetate was sup-plied by Merck (Darmstadt, Germany) and ammonia(25%), hydrochloric acid (min. 32%) and sulphuricacid (95–97%) from Riedel-de Haën (Seelze, Germany).The standards vindoline and catharanthine were pur-chased from LKT Laboratories Inc. (St Paul, MN, USA).Vinblastine and vincristine were from Alexis Biochemi-cals Corporation (San Diego, CA, USA). The internalstandard, ajmalicine hydrochloride was from Roth(Karlsruhe, Germany). Water was obtained from theMilli-Q plus purification system (Millipore, Molsheim,France).

Plant material

A new variety of Catharanthus roseus, Petrus (Grant ofCommunity Plant Variety Rights, decision no. EU39561998), developed at the University of Helsinki, wasused for the extraction of alkaloids. The plants weregrown in the greenhouse from February to June underthe following conditions: temperature 24°C, humidity60%, light intensity 4000 lx with a photoperiod of16 h/day. After a growing period of 6 weeks, a total of27 seedlings were selected and transferred into pots.Leaves were collected from adult plants, freeze-driedand crushed to provide a homogenised sample.

54 A. VERMA ET AL.


SFE instrumentation

Supercritical fluid extractions were performed withan automated ISCO SFX™ 3560 (Lincoln, NE, USA)instrument using 6 mL extraction vessels. Theextraction vessels were filled with 100 mg of driedplant samples mixed with anhydrous sodium sulphate(J.T. Baker, Deventer, Holland). The extracted analyteswere collected into 10 mL of methanol. The internalstandard was added to the collection vials immediatelyafter extraction. The collection temperature was +5°C.The carbon dioxide for extraction (99.9992%, SFEgrade) and for cooling (99.7%), was purchased fromOy AGA Ab (Espoo, Finland). The SFE instrumentwas equipped with a 260 mL syringe pump for theaddition of carbon dioxide at a flow-rate of 1.5 mL/minand a manually controlled Jasco (Tokyo, Japan)PU-980 HPLC pump for addition of the modifier(methanol) at flow-rates of 0.04–0.1 mL/min (2.6–6.6%). The restrictor temperature was set at 60°C in allextractions.

HPLC analysis

Each extract was evaporated to dryness under nitrogenand the residue was dissolved in 1.0 mL of methanol.Samples were filtered through a non-sterile 13 mmMillipore Millex® syringe filter unit (0.45 μm membrane)and analysed with an Agilent (Santa Clara, CA, USA)1050 HPLC system equipped with a UV detector set at214 nm. Chromatographic separations were carried outon a Phenomenex (Torrance, CA, USA) Gemini C18

column (150 × 2.0 mm i.d.) packed with 5 μm particleshaving pore size 100 Å. The mobile phase contained10 mM ammonium acetate buffer (pH 5.0), acetonitrileand methanol in proportions of 65:20:15 changing to30:40:30 during 30 min linear gradient at a flow rateof 0.3 mL/min. The total analysis time was 36 min butall major alkaloids were eluted within 21 min. Theinjections (5 μL) were made by an autosampler withan injection needle. The data were collected andanalysed using a Hewlett-Packard computing system(Agilent ChemStation for LC, Rev. A.09.01).

HPLC-MS analysis

The major alkaloids in the plant extract were identifiedwith a Hewlett-Packard 1100 HPLC system coupledto an Esquire-LC ion-trap mass spectrometer (BrukerDaltonics, Bremen, Germany) using an electrosprayionisation (ESI) interface in the positive ionisationmode. The chromatographic conditions were the sameas described above for HPLC. The source temperaturewas set at 365°C and the source voltage was constantat 3.5 kV. Nitrogen gas was used as sheath and

nebuliser gas at 9.0 L/min and 40.0 psi. The ion trapwas scanned from 100 to 900 Da.

Software for the design of SFE experiments

Response surface modelling, using MODDE 6 softwarefor Design of Experiments and Optimisation (UMETRICSAB, SE-907 19 Umeå, Sweden) was applied tooptimise the SFE conditions. A full factorial designwith four factors, temperature, pressure, modifierflow-rate and dynamic extraction time was created attwo levels, comprising all the possible combinations ofthe factor levels. Initially 19 experiments (N) weredesigned by the software for four factors (p) at twolevels (N = 2p + 3 centre points). All optimisationexperiments were randomly performed without replica-tion. The measured response was calculated as thedry weight content (μg/g) from the peak areas ofcatharanthine and vindoline.

Traditional extraction methods

Solid–liquid extraction. Plant samples (100 mg) wereextracted with 10 mL solution of 0.5 M sulphuric acidand methanol (3:1 v/v, pH 1.4) in an ultrasonic bathfor 3 h. The extracts were filtered through Schleicher &Schuell (Micro Science, Dassel, Germany) filter papers(90 mm) on a Büchner funnel, and the residue wasre-extracted with another 10 mL of the same solventfor a further period of 1 h. The internal standard wasadded to the combined supernatant, which wasfiltered, made alkaline (pH 9.5) with 2 mL of 25%ammonia and then extracted with 2 × 10 mL of methyltert-butyl ether. Ether fraction was separated andevaporated to dryness. The dry residue was dissolvedin 1 mL of methanol, filtered (0.45 μm membrane), anda 5 μL aliquot was injected into the HPLC.

Soxhlet extraction. Soxhlet extraction was performed byextracting 100 mg of the plant sample with 70 mL ofmethanol (boiling point 65.5°C) and dichloromethane(boiling point 40.5°C) separately, in a reflux for 16 h.The internal standard was added immediately afterextraction. The extracts were evaporated to drynessand the residues were dissolved in 1 mL of methanol,filtered (0.45 μm membrane), and 5 μL aliquots wereinjected into the HPLC.

Hot water extraction at 50, 70 and 90°C. Plant samples(100 mg) were extracted with 10 mL of water in hotwater bath with constant shaking for 3 h at 50, 70 and90°C separately. The extracts were filtered throughWhatman no. 40 filter paper on a Büchner funnel andthe residue was re-extracted with another 10 mL ofwater for 1 h. The internal standard was added to the



Table 1 Dry weight content (μg/g) and RSD (%) of major alkaloids in different extraction methods

Dry weight content of major alkaloids in μg/g (RSD %)

Method of extraction CTR VDL AVLB

SFE 198.8 (6.5) 208.2 (7.1) 77.8 (7.2)Sox M 329.5 (4.1)Sox D 177.9 (11.6) 353.8 (11.4) 29.4 (14.0)SLE 132.8 (15.9) 287.6 (13.3) 156.8 (14.5)HWE 50 101.2 (6.4) 159.1 (7.5) 12.4 (6.3)HWE 70 95.4 (6.3) 153.4 (10.8) 12.9 (13.0)HWE 90 5.9 (11.5) 8.2 (11.9) 4.5 (17.5)

For SFE n = 6; for other methods n = 3. CTR = catharanthine; VDL = vindoline; AVLB = 3′,4′-anhydrovinblastine; SFE =supercritical fluid extraction; Sox M = Soxhlet extraction using methanol; Sox D = Soxhlet extraction using dichloromethane; SLE =solid–liquid extraction; HWE = hot water extraction at 50, 70 and 90°C.

combined supernatant, which was filtered, made alka-line (pH 9.5) with 2 mL of 25% ammonia and thenextracted with methyl tert-butyl ether (2 × 10 mL).The ether fraction was separated and evaporated todryness. The dry residue was dissolved in 1 mL ofmethanol, filtered (0.45 μm membrane), and a 5 μLaliquot was injected into the HPLC.

RESULTS AND DISCUSSION

HPLC method development

The C. roseus plant extract is a complex mixture ofseveral alkaloids with a wide range of polarities (Fig. 1)and different pKa values. The selectivity is highlyaffected by mobile phase pH because of the substantialdifferences between the pKa values of the alkaloids. Thechromatographic profile and the total analysis time canbe altered by making slight changes in the organiccontents and the pH of the mobile phase (Theodoridiset al., 1997). Several HPLC methods have been devel-oped in the past for the analysis of indole alkaloidsof C. roseus using solvent systems having buffersincluding ammonium acetate (Naaranlahti et al., 1987;Chu et al., 1997), triethylamine (Kohl et al., 1983) andsodium or ammonium hydrogen phosphate (Renaudin,1984; Miura et al., 1987). Although phosphate bufferis more commonly used for the analysis of basicalkaloids, ammonium acetate has an advantage ofbeing compatible with HPLC-MS owing to its volatility.The addition of ammonium acetate to the mobile phasealters the ionic strength, stabilises the pH and resultsin improved peak shape and selectivity (Theodoridiset al., 1997).

The described method was developed using 10 mM

ammonium acetate (pH 5.0), methanol and acetonitrileas mobile phase, which resulted in a good chromato-graphic separation of standard alkaloids within 20 minby gradient elution [Fig. 2(A)]. A higher (20 mM) or lower

(5 mM) concentration of ammonium acetate did notresult in improved peak shape or better resolution.Hence, the buffer concentration (within the studiedrange) does not affect the chromatographic profile.

Catharanthine, vindoline, vinblastine, vincristine,leurosine and 3′,4′-anhydrovinblastine were separatedby varying the pH of the mobile phase. Vinblastinealways co-eluted with an unknown compound at pHs4.3, 5.0 and 5.6 [Fig. 2(B)], making it difficult to calcu-late its dry weight precisely. On increasing the pH ofthe mobile phase to 6.8, the separation of vinblastineand vincristine was achieved but leurocine co-elutedwith another compound (Fig. 3). The repeatability ofretention times at pH 5.0 (within the buffer range) wasbetter than that at pH 6.8 (outside the buffer range).The relative standard deviation (RSD) varied from 0.1to 0.4% at pH 5.0 and from 4.7 to 5.3% at pH 6.8.Increasing the pH of the mobile phase also resulted inan increase in the retention of the major alkaloids onthe column, and the total elution time increased from21 min (pH 5.0) to 31 min (pH 6.8). However,catharanthine, vindoline and 3′,4′-anhydrovinblastinewere well separated at pH 5.0 and their dry weight con-tents were determined quantitatively (Table 1). Serpen-tine was not observed in the plant extracts, whereaspeaks of tabersonine and vindolinine were observed inthe HPLC chromatograms [Figs 2(B) and 3]. Sincestandards for these alkaloids were not available, theirdry weight contents could not be determined.

In most of the earlier HPLC methods, detectionwavelengths of 298, 280 or 254 nm have been used(Naaranlahti et al., 1987; Song et al., 1992; Volkov andGrodnitskaya, 1994, Chu et al., 1997). Uniyal et al.(2001) observed that catharanthine, vindoline, vinbla-stine and vincristine showed greater absorption at220 nm as compared with that at 298, 280 or 254 nm.We used the detection wavelength of 214 nm becausevindoline and other dimeric alkaloids absorb slightlybetter at this wavelength than at 220 nm (Verma A,Laakso I; unpublished results). A slight drift in the

56 A. VERMA ET AL.


Figure 1 Chemical structures of Catharanthus alkaloids. The numbers 1–7 represent the peak numbers in Figs 2 and 3.



baseline was observed at 214 nm; however, it did notaffect the integration of the peaks.

Quantitation

Quantitation of the alkaloids was based on the internalstandard method. Ajmalicine was used as an internalstandard to quantify the major alkaloids in the leaves.The radioimmunoassay for ajmalicine and serpentineby Arens et al. (1978) showed that ajmalicine and ser-pentine are primarily found in the Catharanthus rootsystem, whilst the green parts of the plant are free ofajmalicine and contain only small amounts of serpen-

tine. The HPLC method developed by Naaranlahti et al.(1987) for the analysis of Catharanthus alkaloids alsoshowed that ajmalicine is present only in root samples.In more recent works by Verpoorte et al. (2002)and Van der Heijden et al. (2004), the presence ofajmalicine in the green parts of the plant was dis-cussed. In the present work, the leaf extracts obtainedby SFE, solid–liquid extraction, Soxhlet and hot waterextraction methods were analysed without the additionof the internal standard ajmalicine (blanks). Sincewe could not detect the presence of ajmalicine in theblanks, this compound was chosen as the internalstandard for the quantification of major leaf alkaloids.The HPLC-MS spectra of the leaf sample did not show

Figure 2 HPLC chromatograms of (A) standard compounds and (B) leaf extract of C. roseus obtained by SFE (mobile phase con-sisted of 10 mM ammonium acetate buffer at pH = 5.0). Key to peak identity: 1 = catharanthine (Rt = 7.6 min); 2 = ajmalicine(internal standard; Rt = 12.4 min); 5 = vindoline (Rt = 19.1 min); 6 = leurocine (Rt = 20.1); 7 = 3′,4′-anhydrovinblastine (Rt = 21.0);and 8 = unknown compound (Rt = 16.8 min) co-eluting with vinblastine (Rt = 17.0); x = tabersonine/vindolinine (Rt = 2.8/3.3). (Foranalytical protocols see the Experimental section.)

58 A. VERMA ET AL.


the presence of the molecular ion (M + H)+ of ajmalicineat m/z 353.

In all of the extraction methods the internal standardwas added after the extraction in order to avoid itspossible thermal degradation owing to high tempera-tures and long residence times of extraction. Ten-pointcalibration graphs (n = 3) were created for each of thefour standards, namely catharanthine, vindoline, vinbla-stine and vincristine, for the range 0.18–31 μg/mL byplotting the peak area ratios of analyte and internalstandard vs the amounts of analytes. Linear regressionanalysis was used to calculate the calibration curveparameters. Good linearity was observed (r2 ≥ 0.999 in

all instances) for the above range. The limits of detec-tion (LOD) were calculated from the lowest concentra-tion of calibration standards for each compound basedon three times the noise level. For catharanthine, vindo-line, vinblastine and vincristine, the LOD were foundto be 0.2, 0.15, 0.1 and 0.08 μg/mL, respectively. Thenoise level in the sample run was only slightly higher(four times) than that in the standard run. The limits ofquantification (LOQ) were calculated to be 2.7, 2.0, 1.3and 1.1 μg/g, respectively. Since the standard for 3′,4′-anhydrovinblastine was not available, it was quantifiedin the plant sample using the calibration parameters ofvinblastine.

Figure 3 HPLC chromatogram of leaf extract of C. roseus obtained by solid–liquid extraction (mobile phase consisted of 10 mM

ammonium acetate buffer at pH = 6.8). Key to peak identity: 1 = catharanthine (Rt = 19.9 min); 2 = ajmalicine (internal standard; Rt

= 24.35 min); 3 = vincristine (Rt = 23.0 min); 4 = vinblastine (Rt = 25.7); 5 = vindoline (Rt = 21.8); and 7 = 3′,4′-anhydrovinblastine(Rt = 30.7); x = tabersonine/vindolinine (Rt = 7.3/9.5). (For analytical protocols see the Experimental section.)

Figure 4 Comparison of relative recoveries of the three major alkaloids of C. roseus (the best recovery obtained is represented as100%). For SFE n = 6, and for all other extraction methods n = 3. SFE = supercritical fluid extraction; Sox M = Soxhlet extractionusing methanol; Sox D = Soxhlet extraction using dichloromethane; SLE = solid–liquid extraction using ultrasonic bath; and HWE= hot water extraction at 50, 70 and 90°C. CTR = catharanthine; VDL = vindoline; and AVLB = 3′,4′-anhydrovinblastine.



HPLC-MS analysis

Soft ionisation techniques such as atmospheric pressurechemical ionisation (APCI) or electrospray ionisation(ESI) have often been used for the characterisationof plant secondary metabolites because the MS soobtained are dominated by respective protonated mole-cules in high yields (Chu et al., 1997). The chromato-graphic profile of C. roseus is very complex due tothe presence of a large number of different alkaloids,some of which co-elute with each other. Therefore,the major peaks in the leaf extracts were identifiedusing HPLC-ESI/MS). For monomeric alkaloids, suchas catharanthine and vindoline, only monoprotonatedions [M + H]+ were observed, whilst for dimeric alka-loids, such as vinblastine, vincristine, leurocine and3′,4′-anhydrovinblastine, both mono- and diprotonated[M + 2H]2+ ions were detected at their correspondingm/z values. Both tabersonine and vindolinine wereidentified by their protonated molecular ions (M + H)+

at an m/z value of 337.1.

Optimisation of SFE conditions

Response surface regression was used to obtain anoptimum set of conditions. With 19 initial experiments(Table 2, Expt. N1–N19) no clear maximum or mini-mum response could be found within the design factorranges used. Therefore, a complement design for all thefour factors was made to allow the estimation of quad-ratic effects, evaluate how the factors influenced theresponse and to optimise or find a region of operability.This further proposed two runs per factor plus an addi-tional centre point (Table 2, Expt. C20–C28). In thisway, non-linear responses could be revealed and aclear optimum was found for each parameter. The runswere carried out in random order, and the responsesurface plots were drawn for each parameter. Cathar-anthine and vindoline were extracted in all 19 initialexperiments (N1–N19) as well in the complementaryexperiments (C20–C28), whereas the peaks for thedimeric alkaloids vinblastine, vincristine, leurocine and3′,4′-anhydrovinblastine were present in only some of

Table 2 Factor levels and design matrix (24 factorial design) for SFE

Experiment Experiment Run P T Modifier flow TimeDry wt content (μg/g)

No. name order (bar) (°C) (mL/min) (min) CTR VDL

1 N1 1 200 40 0.04 20 39.34 194.352 N2 8 450 40 0.04 20 114.27 235.293 N3 16 200 80 0.04 20 141.98 211.144 N4 4 450 80 0.04 20 124.75 258.555 N5 13 200 40 0.1 20 124.3 176.986 N6 16 450 40 0.1 20 33.76 170.937 N7 10 200 80 0.1 20 146.39 244.148 N8 15 450 80 0.1 20 56.89 176.139 N9 14 200 40 0.04 60 132.28 284.42

10 N10 19 450 40 0.04 60 88.73 257.3511 N11 9 200 80 0.04 60 133.35 231.8112 N12 11 450 80 0.04 60 101.53 168.8713 N13 18 200 40 0.1 60 61.45 195.9114 N14 5 450 40 0.1 60 95.63 201.6215 N15 7 200 80 0.1 60 131.85 189.1816 N16 3 450 80 0.1 60 72.44 201.0317 N17 12 325 60 0.07 40 132.15 220.6718 N18 17 325 60 0.07 40 135.73 205.8319 N19 2 325 60 0.07 40 99.8 194.1420 C20 23 325 40 0.04 20 59.76 205.8321 C21 28 325 80 0.1 60 150.43 288.8722 C22 21 200 60 0.04 20 48.06 194.1823 C23 20 450 60 0.1 60 65.87 201.6624 C24 27 200 40 0.07 20 60.33 167.5825 C25 24 450 80 0.07 60 68.2 158.4826 C26 25 200 40 0.04 40 180.1 259.1227 C27 22 450 80 0.1 40 88.62 186.2128 C28 26 325 60 0.07 40 128.47 203.37

Nineteen initial runs (N1–N19) for four factors (p) at two levels (N = 2p + 3 centre points) and nine complimentary runs (C20–C28;two runs for each factor + one centre point) were designed by the software. Response is shown as dry weight content (μg/g) forcatharanthine (CTR) and vindoline (VDL).

60 A. VERMA ET AL.


and vindoline for all four parameters. The optimum setof conditions was determined from Figs 5–7. Otherplots showing similar response surfaces (trends) are notpresented. The effect of pressure and temperature onthe dry weight content of catharanthine and vindolineclearly showed that high temperature (80°C) favours theextraction recoveries of both alkaloids. The recoveryof catharanthine was highest at the lowest pressure

Figure 5 Response surface plots for (A) catharanthine and (B) vindoline showing their dry weight content (μg/g) as a function oftemperature (°C) and pressure (bar); the other two factors (dynamic extraction time and modifier flow) were kept constant at theirmiddle values.

Figure 6 Response surface plots for catharanthine showing its dry weight content (μg/g), (A) as a function of temperature (°C)and dynamic extraction time (min), the other two factors (modifier flow and pressure) being kept constant at theirmiddle values; (B) as a function of pressure (bar) and dynamic extraction time (min), the other two factors (modifier flow andtemperature) being kept constant at their middle values.

the initial and complementary runs. The response (dryweight content) could only be determined precisely forcatharanthine and vindoline.

The response surface plots were created for indivi-dual compounds as three-dimensional plots by pre-senting the response as a function of two factors andkeeping the other two constant at their centre values.Six separate plots were made for both catharanthine



(200 bar), whereas for vindoline the pressure of 300 barwas more favourable [Fig. 5(A, B)]. Song et al. (1992)also obtained the highest recovery for vindoline at300 bar. Since the same matrix contained both thealkaloids, therefore, the optimum pressure (250 bar)and temperature (80°C) were selected for carrying outreplicate extractions (n = 6). Figure 6(A, B) show theinteraction of temperature and pressure with thedynamic extraction time on the dry weight content ofcatharanthine. It can be seen that a maximum yield ofcatharanthine was obtained when the extractions werecarried out for 40 min, both higher and lower extrac-tion times showed a decrease in the dry weight con-tent. For vindoline, a higher extraction time (60 min)was favourable when the interaction of pressure withdynamic extraction time was evaluated, whereas, alower extraction time (20 min) gave higher yields whenthe interaction of temperature was evaluated. Thiscould be due to possible degradation at higher tem-peratures with long extraction times. Therefore, theoptimum time of 40 min was selected for the replicateextractions.

Figure 7(A) shows the interaction of pressure andmodifier flow on the yield of catharanthine. Higher dryweight contents were obtained with a modifier flow of0.1 mL/min (6.6 vol%). Similar results were obtainedfor vindoline. The interaction of temperature and modi-fier flow gave a higher yield of vindoline at the lowest(0.04 mL/min) as well as at the highest (0.1 mL/min)modifier flow rates [Fig. 7(B)]. Since the flow rate of0.1 mL/min was favourable for catharanthine as well,it was chosen as the optimum modifier flow.

Comparison with traditional extraction methods

The supercritical fluid extraction of Catharanthusalkaloids at optimised conditions was compared withtraditional solid–liquid extraction, Soxhlet extractionand hot water extraction at various temperatures. Sixreplicate experiments were carried out with SFE, whereasthree replicates were performed with all other extrac-tion methods. The relative recoveries of the three majoralkaloids using different extraction methods are shownin Fig. 4. The best recoveries (100%) for catharanthinewere obtained with SFE at optimised conditions, forvindoline with Soxhlet extraction using dichloromethaneunder reflux for 16 h, and that for 3′,4′-anhydrovinblastinewith traditional solid–liquid extraction using a solutionof 0.5 M sulphuric acid and methanol (3:1 v/v, pH 1.4)in an ultrasonic bath for 3 h.

Dichloromethane has been used for the extraction ofseveral types of alkaloids using different extractionmethods. El Jaber-Vazdekis et al. (2006) claimed thehigher extraction efficiency of dichloromethane and itsreplacement with chloroform. However, the use of chlo-roform in laboratories incurs major health, securityand regulation problems. Although dichloromethanehas also been suspected of being a toxicant to humanorgans, it is still widely used as a process solvent in themanufacturing of drugs and pharmaceuticals. It is amedium polar solvent generally suitable for the extrac-tion of compounds with different polarities. Dichloro-methane can provoke quaternisation of certain types ofalkaloids (such as strychnine) producing crystallineprecipitates at room temperature. The quaternisation

Figure 7 (A) Response surface plot for catharanthine showing its dry weight content (μg/g) as a function of pressure (bar) andmodifier flow (mL/min), the other two factors (temperature and dynamic extraction time) being kept constant at their middle val-ues. (B) Response surface plot for vindoline showing its dry weight content (μg/g) as a function of temperature (°C) and modifierflow (mL/min), the other two factors (pressure and dynamic extraction time) being kept constant at their middle values.

62 A. VERMA ET AL.


of tropane alkaloids has been reported by Vincze andGefen (1978) during MS analysis when high tempera-tures were used for sample volatilisation. We did notobserve this phenomenon during the Soxhlet extrac-tion of Catharanthus alkaloids using dichloromethane.Furthermore, if quaternisation had taken place, verypolar compounds would have formed eluting muchearlier than other major alkaloids, thus producing newpeaks. No extra peaks were observed in the HPLCchromatogram of the dicholoromethane extract.

The high recoveries of catharanthine (89.7%) andvindoline (100%) with Soxhlet extraction using dichlo-romethane as solvent could be due to the stability ofthese monomeric alkaloids at lower temperature (40°C,boiling point of dichloromethane). When methanol(boiling point 65°C) was used as solvent in the Soxhletextraction catharanthine was completely decomposedafter 16 h of extraction owing to its thermolabile nature.On the other hand, a ca. 48% recovery of cathar-anthine using hot water extraction for 3 h at 70°Cwas obtained, showing that this compound can beextracted at higher temperatures using shorter extrac-tion times. Hot water extraction at 90°C resulted inthe nearly complete decomposition of catharanthineand 3′,4′-anhydrovinblastine. The recovery of vindoline(63.1–65.3%) remained unaffected by the increase intemperature during hot water extraction; vindolineseemed to be the most stable Catharanthus alkaloid.

Both hydrochloric acid and sulphuric acid were com-pared for the extraction of major alkaloids. Extractionwith sulphuric acid gave better results in comparisonto hydrochloric acid. Hallard (2000) also showed thatthe relative extraction efficiency of vindolinine wasbetter with sulphuric acid (50%) in comparison withhydrochloric acid (44%). The important result of thiswork was the extraction of 3′,4′-anhydrovinblastine(38.4%) by SFE, although, its best recovery (100%) wasobtained by the traditional solid–liquid extraction usinga solution of 0.5 M sulphuric acid and methanol (3:1v/v, pH 1.4) in an ultrasonic bath for 3 h. This couldbe due to its improved solubility and stability at lowpH, since the alkaloids are basic compounds and formsalts easily in aqueous acidic medium, which enhancestheir solubility. In addition, the proton in aqueousacidic medium is probably more reactive and maybreak the sample matrix to release the analytes moreeasily in solid–liquid extraction as compared with theinert supercritical carbon dioxide in SFE. The alkaloidsin free base form were easily separated by liquid–liquidextraction using organic solvents. We used methyltert-butyl ether because it is a more environmentallyfriendly and non-toxic solvent in comparison withchloroform and dichloromethane and its extractionefficiency was similar to that of dichloromethane.

High temperatures and longer extraction times alsoaffected the recovery of 3′,4′-anhydrovinblastine. Solid–liquid extraction at room temperature for 3 h resulted

in 100% relative recovery, whereas hot water extractionat 50–90°C for 3 h gave only 9–2% recoveries. Soxhletextraction with dichloromethane and with methanolfor 16 h gave 18 and 0% recoveries, respectively. InSFE, the extractions were carried out at 80°C with adynamic extraction time of 40 min, which resulted in38% recovery. The above results clearly shows that3′,4′-anhydrovinblastine is a thermolabile compoundand degrades easily if extracted at higher temperatureswith longer extraction (or residence) times.

The RSD for the major alkaloids varied from 4.1 to17.5% with different extraction methods (Table 1). TheRSD values for replicate extractions in SFE were from6.5 to 7.2%, whereas in solid–liquid extraction and hotwater extraction the RSD values varied from 6.3 to17.5%. The higher RSD percentages obtained withtraditional extraction methods could be explained bythe long and tedious sample preparation proceduresemployed, which included filtration, treatment withorganic solvents, liquid–liquid extraction, separationand evaporation steps, whereas in SFE the extractiontook place in closed vessels and the collected extractwas simply evaporated.

In general, the extracts from SFE [Fig. 2(B)] werefound to be much cleaner in comparison with thoseobtained by solid-liquid extractions (Fig. 3) or Soxhletextractions. The results showed that supercritical fluidextraction is a valuable alternative technique to tradi-tional extraction methods of Catharanthus alkaloidsfrom dried leaves. To the best of our knowledge, theextraction of 3′,4′-anhydrovinblastine and leurocinehas not been reported earlier using SFE. The effect offour SFE factors on the dry weight content of cathar-anthine and vindoline was measured by applying a fullfactorial design and response surfaces were plotted toevaluate their interactions. Clearly the pressure andtemperature were dominant factors in the extraction ofCatharanthus alkaloids.

AcknowledgementsThe authors wish to thank Matti Jussila for technicalassistance in the use of ion-trap mass spectrometer(Bruker Daltonics). Financial support for this workwas provided by the University of Helsinki (project2105040).

REFERENCES

Arens H, Stöckigt J, Weiler EW, Zenk MH. 1978. Radioimmunoassaysfor the determination of the indole alkaloids ajmalicine and serpen-tine in plants. Planta Med 34: 37–46.

Bergeron C, Gafner S, Clausen E, Carrier DJ. 2005. Comparison of thechemical composition of extracts from Scutellaria lateriflora usingaccelerated solvent extraction and supercritical fluid extractionversus standard hot water or 70% ethanol extraction. J Agric Food

Chem 53: 3076–3080.



Brachet A, Christen P, Gauvrit J-Y, Longeray R, Lantéri P, VeutheyJ-L. 2000. Experimental design in supercritical fluid extraction ofcocaine from coca leaves. J Biochem Biophys Meth 43: 353–366.

Choi YH, Yoo K-P, Kim J, 2002. Supercritical fluid extraction andliquid chromatography–electrospray mass analysis of vinblastinefrom Catharanthus roseus. Chem Pharm Bull 50 (9): 1294–1296.

Chu I-H, Bodnar JA, Bowman RN, White EL. 1997. Determinationof vincristine and vinblastine in Catharanthus roseus plants byhigh performance liquid chromatography/electrospray ionizationmass spectrometry. J Liq Chromatogr Relat Technol 20(8): 1159–1174.

El Jaber-Vazdekis N, Gutierrez-Nicolás F, Ravelo ÁG, Zárate R. 2006.Studies on tropane alkaloid extraction by volatile organic solvents:dichloromethane vs. chloroform. Phytochem Anal 17: 107–113.

Ellington E, Bastida J, Viladomat F, Codina C. 2003. Supercriticalcarbon dioxide extraction of Colchicum autumnale L. Phytochem

Anal 14: 164–169.Hallard DAC. 2000. Transgenic plant cells for the production of indole

alkaloids. PhD thesis, Leiden University.Hamburger M, Baumann D, Adler S. 2004. Supercritical carbon diox-

ide extraction of selected medicinal plants—effects of high pressureand added ethanol on yield of extracted substances. Phytochem

Anal 15: 46–54.Heaton DM, Bartle KD, Rayner CM, Clifford AA. 1993. Application of

supercritical fluid extraction and supercritical fluid chromato-graphy to the production of taxanes as anticancer drugs. J High

Resol Chromatogr 16: 666–670.Janicot JL, Caude M, Rosset R, Veuthey J. L. 1990. Extraction of

major alkaloids from poppy straw with near-critical mixtures ofcarbon dioxide and polar modifiers. J Chromatogr A 505: 247–256.

Jarvis PA and Morgan ED. 1997. Isolation of plant products bysupercritical-fluid extraction. Phytochem Anal 8: 217–222.

Kim HK, Choi YH, Luijendijk, Vera Rocha RA, Verpoorte R. 2004. Com-parison of extraction methods for secologanin from Symphoricarpos

albus using 1H-NMR. Phytochem Anal 15: 257–261.Kohl W, Witte B, Hoefle G. 1983. Quantitative and qualitative HPLC

analysis of indole alkaloids from Catharanthus roseus cell cultures.Planta Med 47: 177–182.

Miura Y, Hirata K, Kurano N. 1987. Isolation of vinblastine in callusculture with differentiated roots of Catharanthus roseus. Agric Biol

Chem 51: 611–614.Naaranlahti T, Nordström M, Huhtikangas A, Lounasmaa M. 1987.

Determination of Catharanthus alkaloids by reversed-phase high-performance liquid chromatography. J Chromatogr 410: 488–493.

Naaranlahti T, Auriola S, Lapinjoki S. 1991. Growth-related dimerisa-tion of vindoline and catharanthine in Catharanthus roseus andeffect of wounding on the process. Phytochemistry 30: 1451–1453.

Noble RL. 1990. The discovery of the vinca alkaloids—chemotherapeuticagents against cancer. Biochem. Cell Biol 68: 1344–1351.

Ollanketo M, Hartonen K, Riekkola M-L, Holm Y, Hiltunen R. 2001.Supercritical carbon dioxide extraction of lycopene in tomatoskins. Eur Food Res Technol. 212: 561–565.

Pereira GC, Marques MOM, Barreto AS, Siani AC, Fernandes EC,Meireles MAA. 2004. Extraction of indole alkaloids from Tabernae-

montana catharinensis using supercritical CO2 + ethanol: an evalu-ation of the process variables and the raw material origin.J Supercrit Fluids 30: 51–61.

Renaudin J-P. 1984. Reversed-phased high-performance liquid chromato-graphic characteristics of indole alkaloids from cell suspensioncultures of Catharanthus roseus. J Chromatogr 291: 165–174.

Saldaña DAM, Mohamed RS, Baer MG, Mazzafera P. 1999. Extractionof purine alkaloids from Maté (Ilex paraguariensis using supercri-tical CO2. J Agric Food Chem 47: 3804–3808.

Schmidt B, Kutney J, Mayer L. 2003. Anhydrovinblastine for the treat-ment of cancer. US Patent 20030162803.

Song K-M, Park S-W, Hong W-H, Lee H, Kwak S-S and Liu J-R. 1992.Isolation of vindoline by supercritical fluid extraction. Biotechnol

Prog 8: 583–586.Sottomayor M, Lopez-Serrano M, DiCosmo F, Barcelo AR. 1998. Purifi-

cation and characterization of α-3´,4´-anhydrovinblastine synthase(peroxidase-like) from Catharanthus roseus (L.) G. Don. FEBS Lett

428: 299–303.Suomi J, Sirén H, Hartonen K, Riekkola M-L. 2000. Extraction of

iridoid glycosides and their determination by micellar electrokineticcapillary chromatography. J Chromatogr A 868: 73–83.

Theodoridis G, Papadoyannis I, Hermans-Lokkerbol A, Verpoorte R.1997. A study of the behaviour of some new column materials inthe chromatographic analysis of Catharanthus alkaloids. Chromato-

graphia 45: 52–58.Uniyal GC, Shashi B, Mathus AK, Kulkarni RN. 2001. Symmetry C18

column: a better choice for the analysis of indole alkaloids ofCatharanthus roseus. Phytochem Anal 12: 206–210.

Van der Heijden R, Jacobs DI, Snoeijer W, Hallard D, Verpoorte R.2004. The Catharanthus alkaloids: pharmacognosy andbiotechnology. Curr Med Chem 11: 607–628.

Verpoorte R, Contin A, Memelink J. 2002. Biotechnology for theproduction of plant secondary metabolites. Phytochem Rev 1: 13–25.

Vincze A, Gefen L. 1978. Solvent-cause quaternization as a possiblesource of error in the mass spectral quantitation of tertiaryamines. I. Methylene chloride quaternization. Isr J Chem. 17: 236–238.

Volkov SK, Grodnitskaya EI. 1994. Application of high-performanceliquid chromatography to the determination of vinblastine inCatharanthus roseus. J Chromatogr B 660: 405–408.

Optimisation of supercritical fluid extraction of indole alkaloids from Catharanthus roseus using...

Documents

Transcript of Optimisation of supercritical fluid extraction of indole alkaloids from Catharanthus roseus using...