This article was downloaded by: [University of Hong Kong Libraries]On: 02 October 2013, At: 08:45Publisher: Taylor & FrancisInforma Ltd Registered in England and Wales Registered Number: 1072954 Registered office: MortimerHouse, 37-41 Mortimer Street, London W1T 3JH, UK

Food Additives & ContaminantsPublication details, including instructions for authors and subscription information:http://www.tandfonline.com/loi/tfac19

Determination of pyrrolizidine alkaloids in honeyfrom selected sites by solid phase extraction andHPLC‐MSColin Crews a , James R. Startin a & Philip A. Clarke aa Fisheries and Food CSL Food Science Laboratory, Ministry of Agriculture, NorwichResearch Park, Colney, Norwich, NR4 7UQ, UKPublished online: 10 Jan 2009.

To cite this article: Colin Crews , James R. Startin & Philip A. Clarke (1997) Determination of pyrrolizidine alkaloids inhoney from selected sites by solid phase extraction and HPLC‐MS, Food Additives & Contaminants, 14:5, 419-428, DOI:10.1080/02652039709374547

To link to this article: http://dx.doi.org/10.1080/02652039709374547

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) containedin the publications on our platform. However, Taylor & Francis, our agents, and our licensors make norepresentations or warranties whatsoever as to the accuracy, completeness, or suitability for any purposeof the Content. Any opinions and views expressed in this publication are the opinions and views of theauthors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should notbe relied upon and should be independently verified with primary sources of information. Taylor and Francisshall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, andother liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relationto or arising out of the use of the Content.

This article may be used for research, teaching, and private study purposes. Any substantial or systematicreproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in anyform to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions

Food Additives and Contaminants, 1997, Vol. 14, No. 5, 419-428

Determination of pyrrolizidine alkaloids in honey fromselected sites by solid phase extraction and HPLC-MS

Colin Crews, James R. Startin and Philip A. ClarkeMinistry of Agriculture, Fisheries and Food CSL Food ScienceLaboratory, Norwich Research Park, Colney, Norwich NR4 7UQ,UK

(Received 31 October 1996; revised 10 January 1997; accepted7 February 1997)

A method was developed for the determination in honeyof the Ragwort (Senecio jacobaea) derived pyrrolizi-dine alkaloids jacoline, jacozine, jacobine, seneciphyl-line and senecionine, combining solid-phase extractionwith high performance liquid chromatography andatmospheric pressure chemical ionization mass spectro-metric detection. The method allowed determination ofindividual alkaloids and offered a considerable improve-ment in terms of speed, sensitivity and specificity overprevious approaches, but was not suitable for determi-nation of jaconine, a minor alkaloid in Ragwort.Instrument calibrations were linear over the range0.005 to 10.0 µg/ml, equivalent to approximately0.001 to 2.0 mg/kg in honey with the extraction methodused and allowing for observed recoveries. Detectionlimits in honey were 0.002mg/kg. Recoveries for mostof the alkaloids were between 57 and 70%. Thealkaloids have been determined in a number of samplesof honey selected after pollen identification and count-ing. The alkaloids were not detectable in samplescontaining two grains or less of Ragwort pollen pergram of honey. Samples collected in late July andAugust contained Ragwort pollen at 15-21 grains/gand total alkaloid concentrations of 0.011-0.056 µg/kg. Similar contributions to the total were made byjacozine, seneciophylline and senecionine, with jacobinemaking a larger and jacoline a smaller contribution.Two samples of honey containing Ragwort pollen at 24and 16 grains/g had total alkaloid concentrations of0.42 and 1.48 mg/kg respectively (not corrected forrecovery). The alkaloid profile in these samples wasdominated by seneciphylline and senecionine which

†To whom correspondence should be addressed.

together comprised 90-95% of the total. Alkaloidswere not detected in retail honeys.

Keywords: high performance liquid chromatogra-phy-mass spectrometry, honey, pyrrolizidine alka-loids, Senecio jacobaea

Introduction

Pyrrolizidine alkaloids (PAs) are toxic compoundswhich occur in a large number of plant species. Inthe UK Senecio jacobaea, a common agriculturalweed known popularly as Ragwort (Tansy Ragwortin the USA), is a major plant source of PAs. Ashepatotoxins the Ragwort PAs particularly affectlivestock, causing liver damage and frequently deathfrom veno-occlusive disease, and there is also evi-dence that certain PAs are carcinogenic, mutagenicand/or teratogenic (Roitmann 1983). PAs can be ofimportance to human health if medicines or foodscontaminated with PAs are ingested (Huxtable et al.1977, Ridker et al. 1985, Ridker and McDermott1989, Molyneux and James 1990).

Senecio jacobaea contains six major PAs (Bradburyand Culvenor 1954), all of which are cyclic diesters ofthe pyrrole retronecine (Figure 1). In some annualspecies of Senecio (S. vulgaris and S. vernalis) the PAshave been shown to be produced in the root tissue asthe N-oxides and translocated to the nectaries of themature plant where they concentrate (Hartmann andZimmer 1986, Hartmann et al. 1989) so that there ispotential for the transfer of PAs to honey.

Ragwort is a biennial or occasionally perennial plantwith a strong natural tendency to increase (Cooperand Johnson 1984) and this tendency could be en-couraged by modern agricultural practices, such asthe restoration of wild flower habitats, which maytherefore lead to an increase in the risk of honeybecoming contaminated with PAs. In the UK, theWeeds Act 1959 (Anon. 1959) empowers Agriculture

© 1997 Crown Copyright

Dow

nloa

ded

by [

Uni

vers

ity o

f H

ong

Kon

g L

ibra

ries

] at

08:

45 0

2 O

ctob

er 2

013

420 C. Crews et al.

seneciomne seneciphylline

Figure 1. The major pyrrolizidine alkaloids of Seneciojacobaea.

Ministers to take action against occupiers of landfrom which certain weeds, including Ragwort, arespreading.

The honey bee is a major pollinator of Ragwortwhich dominates the late-summer flowering periodwhen there are few competitors for pollinating in-spects. Concern has been raised over the possibility offoraging bees making honey from the nectar ofRagwort which may contain PAs.

Poisoning by ingestion of honey containing toxinstransferred from flowers has been reported for Rho-dodendron (Sutliipinar et al. 1992) and Aconitum(Saisho et al. 1994). Evidence of the transfer of PAtoxins from plants to honey has also been described inthe literature; Deinzer et al. (1977) reported levels ofPAs derived from S. jacobaea in honey in the range

0-3-3-9 mg/kg and Culvenor et al. (1981) found PAsfrom Echium species in honey at levels in the range0-25-1 mg/kg.

This small survey was undertaken to determine thelevels of PAs in UK honey produced by bees likely tohave access to Ragwort. The objective was to measurethe concentration of PAs in a small number ofsamples of honey produced by bees known to havehad access to Ragwort and collected from differenthives and on different dates.

Although the PAs tend to occur as N-oxides, reduc-tion to free base prior to solvent extraction has beenthe accepted method for the analysis of aqueoussamples. Established methods for the analysis ofPAs in plant material rely on methanol extraction,reduction with zinc and acid and clean-up by liquid-liquid partition. This has disadvantages in that allstages are lengthy and the liquid-liquid partitionleads to the formation of emulsions. Similar methodshave been applied to the analysis of PAs in honey.Deinzer et al. (1977) and Culvenor et a/. (1981) bothextracted honey with chloroform and then deter-mined the alkaloids either spectrophotometrically orby gas chromatography, respectively. The extractionsrequired centrifugation to separate emulsions. Dein-zer did not measure recoveries while Culvenor re-ported 60-70% recovery of a single PA of 'averagedifficulty of extraction'.

Hartmann and Toppel (1987) have described the useof a solid-phase extraction (SPE) technique to isolatePAs from aqueous extracts of plant material, and theuse of a rapid oxide-reduction system using sodiumdithionite. We have applied these techniques to theextraction of PAs from honey.

Gas chromatography can be used for separation andmeasurement of PAs but derivatization is usuallyrequired and use of a flame ionization detector givesinsufficient sensitivity and specificity. Separation byhigh performance liquid chromatography (HPLC)has been reported (Ramsdell and Buhler 1981, Ked-zierski and Buhler 1986, Brown et al. 1994), butconventional HPLC detectors lack the requisite sen-sitivity and specificity. Combined HPLC-MS in ther-mospray mode has been used to analyse PAs ofSenecio (Parker et al. 1990), but this lacks thereliability and sensitivity required. High performanceliquid cnromatography-mass spectrometry (HPLC-MS) using the atmospheric pressure chemical ioniza-tion (APCI) technique offers great potential for the

Dow

nloa

ded

by [

Uni

vers

ity o

f H

ong

Kon

g L

ibra

ries

] at

08:

45 0

2 O

ctob

er 2

013

Pyrrolizidine alkaloids in honey 421

analysis of a great variety of compound classes whichare otherwise problematic.

APCI involves the formation of a fine spray ofdroplets and solvent vapour at atmospheric pressureby pneumatic nebulization from a capillary in aheated probe. Ions are formed by a high voltagecorona discharge and extracted from the atmosphericpressure source region and injected into the massanalysing quadrupole in a high vacuum region byappropriate voltages applied to conical lenses. APCIis a soft ionization technique producing ions withlittle excess internal energy so that little fragmenta-tion occurs.

This paper thus documents a method of analysisbased on SPE followed by HPLC-MS, and givesresults from its application to analysis of samples ofhoney. Since preliminary test analyses indicated thatPAs were detectable in some samples but at concen-trations far below 10(ig/ml (taking this as equivalentto lOmg/kg) the full potential sensitivity of the newmethod has been employed.

(a) Samples of honey (ca 50 g) were collected fromhives in three different locations in proximity toareas where Ragwort growth was expected. Thehives were sampled at intervals from pre-flower-ing in early July until flowering had declined inmid-August. Further details of these samples areincluded in table 1.

(b) Two examples of honey having a characteristicappearance and smell attributed to heavy Rag-wort contamination, and thus rendered unpalat-able, were provided directly by a local beekeeper.Further details of these samples are included inTable 2.

(c) Samples of eight retail honeys were purchased.Further details of these samples are included inTable 3.

In addition, a locally purchased retail example ofhoney (produced in Australia) which was found tobe free of S. jacobaea PAs was used to prepare blankand spiked samples.

Experimental

Samples

Three separate series of samples were obtained andwere stored frozen for up to 6 weeks prior to analysis.

Standards

Purified PA standards isolated from S. jacobaea weredonated by Dr K R Gaul, CSIRO, Australia. Mono-crotaline, retrorsine, and retrorsine N-oxide (isati-dine) were obtained from the Aldrich Chemical Co.Ltd. These standards were used as received andwithout assessment of purity.

Table 1. Details of honey samples from specific hive sites and concentrations of alkaloids found.

Laboratorycode

PA1PA4PA6PA2PA3PA5PA9

PAHPA12PA13PA7PA8PA10

Hivesite

AAABBBBBBBCCC

Date

03 07 9413 07 9420 07 9405 07 9413 07 9420 07 9430 07 9405 08 9410 08 9417 08 9427 07 9430 07 9403 08 94

Moisture(%)

——23202226242122———

Ragwort

Grains/g

00020

15202021000

pollen

% o ftotal

0000105-58-96-33-5000

Jacoline

———

< 0002< 0002< 0002< 0002< 0-002< 0002

———

Alkaloid concentration (mg/kg)

Jacozine————

< 0002< 0002< 0009

0006< 0002

0003———

Jacobine

———

< 0002< 0002

0025001900050010———

Seneci-phylline

————

< 0-002< 0002

0009000700030004———

Seneci-onine

———

< 0-002< 0002

0013000800030005———

Total

———

<001<001

0056004000110-022———

—, not analysed. Alkaloid concentrations are means of duplicate determinations which were in good agreement.

Dow

nloa

ded

by [

Uni

vers

ity o

f H

ong

Kon

g L

ibra

ries

] at

08:

45 0

2 O

ctob

er 2

013

422 C. Crews et al.

Table 2. Details of honey samples believed to contain Ragwort nectar and concentrations of alkaloids found.

Ragwortpollen Alkaloid concentration (mg/kg)

Laboratorycode

Date of Grains/ % ofproduction g total Jacoline Jacozine Jacobine Seneciphylline Senecionine Total

JH1990JH1994

19901994

24 4-416 2

Table 3. Details of retail honey samples

Laboratorycode

Ragwort

Grains/g

: pollen

%oftotal

00060003

00200012

recieved and concentrations

Jacoline Jacozine

0031 0-260025 007

of alkaloids found.

Alkaloid concentration (mg/kg)

Jacobine Seneciphylline

1-160-31

Senecionine

1-480-42

Total

RH1RH2RH3RH4RH5RH6RH7RH8

indicates

00020000

not analysed.

Q0 — — — —0 — — — —1-1 < 0-002 < 0-002 < 0-002 < 00020 — — — —0 — — — —0 — — — —0 < 0-002 < 0-002 < 0-002 < 0002

< 0002

< 0002

< 0002

< 0002

Methods

Pollen analysis. Identification and counting of pollenin honey was performed on 2g subsamples bymicroscopic examination following dilution of thehoney and centrifugation.

Moisture content. Moisture contents of the hivesamples were determined from refractive indexmeasurements.

Extraction and clean-up of honey. Honey (10 g) wasdiluted with 5ml 005M sulphuric acid and stirredfor 1 h with sodium dithionite (ca 30 mg) to reducethe alkaloid N-oxides. The solution was madestrongly basic by the addition of concentratedammonia solution (5 ml) and applied to anExtrelut™ 20 column (Merck). After standing for20min the PAs were eluted with 100 ml dichlor-omethane/methanol (95:5). The extract wasevaporated to dryness under vacuum and the residueredissolved in 1 ml of methanol containing 0-25 ug ofmonocrotaline (a PA not found in S. jacobaea) asan injection standard. The solutions were filteredthrough a 0-45 um disposable syringe filter.

Extractions were performed in batches of six consist-ing of a blank honey, a blank honey spiked with astandard mixture to contain each PA at 0-1 mg/kg,and four honey samples.

HPLC-MS of honey extracts. HPLC was performedusinga Spectra-Physics 8800 HPLC pump and a150x4-6 mm 100 A PLRP-S column (PolymerLabs). The mobile phase consisted of 01Mammonium hydroxide and acetonitrile with agradient of 10% to 90% acetonitrile over 30min ata flow rate of 1 ml/min and a temperature of 65°C.Injections (50 ul) were made via a Gilson 231 XLautosampler. Mass spectrometric detection used aVG Platform instrument and Masslynx datasystem(version 2) operated in APCI mode and with selectedion monitoring of the protonated molecular ion([M + H]+) of each analyte using a 0-1 s dwell timeand 002s interscan delay. Scanned spectra wereobtained for a single sample by scanning over themass range 60-600 daltons using a scan time of 1 sand interscan delay of 0-1 s. Reference spectra ob-tained on this instrument are reproduced in Figure2. Typical selected ion monitoring chromatogramsare shown in Figure 3.

Dow

nloa

ded

by [

Uni

vers

ity o

f H

ong

Kon

g L

ibra

ries

] at

08:

45 0

2 O

ctob

er 2

013

Pyrrolizidine alkaloids in honey 423

(0

(ii)

200

Figure 2. APCI mass spectra of pyrrolizidine alkaloidstandards, (i) Jacoline, (ii)jacozine, (iii)jacobine, (iv)jaconine, (v) seneciphylline, (vi) senecionine.

Peak detection, baseline fitting and area measurementwere performed using the data system facilities whichincluded smoothing of the raw data. All baselineassignments were inspected and corrected manuallywhere necessary. Quantification was based on theratio of peak areas to those of the monocrotalineinjection standard.

Analysis of flowers. The PA content of Ragwortflowers collected in an area near to the beehives wasdetermined qualitatively using established extractiontechniques (Bicchi et al. 1985) and GC-MSfollowing derivatization with methane boronic acidand bis-trimethylsilyltrifluoroacetamide. GC-MSwas carried out using a non-polar capillary column,with electron impact ionization and in full-scanmode.

PAstandards

10CH

blankhoney

honeyPA9

a jacolinebISc jacozinedjacobinee jaconinef seneciphyllineg senecionine

25.00 35.00

Figure 3. Examples of chromatograms for determinationof pyrrolizidine alkaloids by APCI HPLC-MS. Tracesare summations from selected ion monitoring at m/z 334,336, 350, 352, 370 and 388. (i) Standards, (ii) extract ofhoney in which PAs were not detectable, (Hi) extract ofhoney PA9. The peak in (ii) at \8-5min was of unknownidentity, was not observed consistently but did not inter-fere with known PAs.

Results and discussion

Method validation

N-oxide reduction. Three samples of blank honeyspiked with 50mg/kg of retrorsine, a chemicallyrelated alkaloid which is found in other Seneciospecies and is available commercially, and threespiked with the same concentration of the N-oxideof retrorsine were subjected to dithionite reductionfor 1 h before extraction and analysis as describedabove. Comparable concentrations of retrorsine were

Dow

nloa

ded

by [

Uni

vers

ity o

f H

ong

Kon

g L

ibra

ries

] at

08:

45 0

2 O

ctob

er 2

013

424 C. Crews et al.

found in all samples, indicating that conversion ofthe oxide to retrorsine was quantitative. Othervalidation tests were based on samples spiked withPAs in the reduced form.

Instrument linearity. The instrumental response wascalibrated and detection limits estimated fromHPLC-MS analysis of solutions of the standardcompounds over the concentration range 0-0005-l-0ug/ml. All of the calibrations showed goodlinearity above 005ug/ml, which, with the methoddescribed, was equivalent to 0-001 mg/kg in honeyafter allowing for typical recoveries. Calibrationcoefficients are given in table 4 together with relativestandard deviations for replicate measures at concen-trations of 0-005 ng/ml and 0-1 (ig/ml (« = 4).

For seneciophylline and senecionine the calibrationwas extended to higher concentrations and wasacceptably linear up to 10 ug/ml but reached a plateauabove this.

In the case of jacobine, chromatograms (m/z 352) ofthe standard showed evidence of a substantial amountof impurity which made baseline positioning difficultand also must have caused a bias in the results for this

compound. Jacobine gave the lowest calibration slopewhich is also consistent with the concentration of thestandard being lower than expected.

Recovery. Preliminary experiments showed that allof the PAs apart from jaconine were recovered fromthe Extrelut™ column with reasonable efficiency.Since method development was associated with onlya single short term study, recovery was notcharacterized fully in advance but a spiked samplewas included in each batch of extractions. Theresults are given in table 5.

There is no entirely satisfactory explanation for theconsistently high apparent recovery of jacobine whichwas a factor of 2-7 greater than the mean recovery ofthe other compounds. Further investigation was notcarried out, but it should be noted that the impuritiesin the standard described above, which shared thesame molecular weight as jacobine, were not observedin extracts of samples spiked with the same standard.There is a possibility that these impurities were partlydue to related compounds and that the conditions ofreduction and extraction resulted in conversion to asingle species. The effect would be for survey results

Table 4. Calibration coefficients for HPLC-MS determination of pyrrolizidine alkaloids over the concentration range0-0005—1-0 [ig/ml and relative standard deviations for four analyses each at concentrations of 0-005 and 0-1 [ig/ml. Thecalibration was based on a total of 28 data points covering 13 concentrations.

Alkaloid

Jacoline

JacozineJacobineSeneciphyllineSenecionine

Slope

2-38

3-641-733-642-67

Intercept

- 0 0 0 3

-0014- 0 0 1 5- 0 0 2 0-0014

Correlationcoefficient

0-999

0-9990-9880-9980-999

Residualstandarddeviation

0034

0043007600590031

Relative standard

0-005 ng/ml 01

10-4

2-83-94-6

20-2

deviation

ug/ml

4-7

3-22-52-27-6

Table 5. Recovery of PAs from honey spiked at a concentration of 0-1 mg/kg.

Recovery (%)

Alkaloid

JacolineJacozineJacobineSeneciphyllineSenecionine

Batch A

n/an/an/an/an/a

Batch B

6352

1676571

Batch C

8168

1967782

Batch D

6857

1806073

Batch E

6649

1444255

Mean

7057

1726170

Dow

nloa

ded

by [

Uni

vers

ity o

f H

ong

Kon

g L

ibra

ries

] at

08:

45 0

2 O

ctob

er 2

013

Pyrrolizidine alkaloids in honey 425

to be overestimated and the results should be treatedwith appropriate caution. Even so, comparisonsbetween different samples can be made safely.

The chlorinated pyrrolizidine, jaconine, could not berecovered from the Extrelut™ column and has thusbeen excluded from analysis. However, it is not amajor component of Ragwort and comprised onlyabout 5% of the total alkaloids extracted from aflower sample and analysed by gas chromatography(see below).

It was found in preliminary tests that extractionrecoveries of monocrotaline were higher (>90%)than the Ragwort PAs and its use was thus restrictedto an injection standard added to the final extract toallow compensation for fluctuations in instrumentsensitivity.

Limit of detection. Based on comparison of thechromatographic peaks with the background noiseobserved in analysis of extracts of blank honey(Figure 3), and allowing for recoveries as low as50%, the lower limit of detection in honey wasestimated as being 0-002 mg/kg and this value hasbeen used as a reporting limit. This is equivalent(with a 50% recovery) to a final solution con-centration of 0-01 ug/ml which was well above thelower end of the linear working range.

Survey quality control. A duplicate analysis wasperformed on each sample in a separate batch andresults in reasonable agreement obtained. A total offive batches of analyses were performed eachcontaining a field blank and a spiked blank sample.The results for spiked samples are discussed above.PAs were not detectable in the blanks.

Pollen and moisture determinations. The results ofthe pollen and moisture analyses are included inTables 1-3.

The moisture levels of the series of samples from thetest hives (Table 1) were higher than the desired levelfor retail honey 17-8%, as water evaporation fromthe nectar had not been completed at the time ofsampling.

Pollen results are expressed as the number of Ragwortpollen grains/g of honey and also as the contributionof Ragwort pollen to the total pollen count. Thepollen levels may not reflect accurately the proportionof Ragwort nectar in the honey as Ragwort pollen isnot actively collected by honey bees (Harper and

Wood 1957), they may, however, indicate the relativefrequency of honey bee visits to the plant.

In the samples from the test hives (Table 1), Ragwortpollen was not found in samples from sites A and C,and this is consistent with reports from beekeepers oflittle or no Ragwort growth at these sites. At site Bthe finding of a low pollen count (2 grains/g) on 13July is apparently contradicted by a zero count for thefollowing week but taking into account the low countand sample size, this may not be statistically signifi-cant. The Ragwort pollen count increased rapidly toaround 20 grains/g in early August and remainedroughly constant for the remainder of the samplingperiod. The proportion of the total pollen count madeup of Ragwort pollen reached a maximum on 5August and then decreased, presumably reflectingan increased input of nectar from other plants.However, in the absence of information on theavailability of pollen and nectar from different plantsand the mass balance of the hive it is not possible tointerpret these figures in a useful manner. The factthat the bees were, in fact, foraging on Ragwort is,however, clearly established.

In the two samples coded JH1990 and JH1994 (Table2), which had a prominent taint which was attributedto Ragwort and made the honey unpalatable, theRagwort pollen counts and contributions to the totalcounts were comparable to those from site B dis-cussed above.

In the retail samples (Table 3) Ragwort pollen wasidentifiable in only one sample (RH4) at the low levelof 2 grains/g.

Pyrrolizidine alkaloid concentrations. A quantity offlowers from a test hive site were examined for thepresence of PAs by extraction, derivatization andfull-scan GC-MS. The presence of senecionine,seneciphylline and jacobine in approximately equalconcentrations was confirmed together with smalleramounts (about 5%) of jaconine and anothercompound that was not identified but appeared tobe a further PA.

Taking into account the beekeeper's assessment of thepresence of flowering Ragwort plants and the pollenanalysis results samples from the hive site B, the'Ragwort' honey, and retail honey RH4 were selectedfor PA analysis with retail honey RH8 being used as acontrol.

Results of the PA determinations on these samplesare included in Tables 1-3. In view of the low levels

Dow

nloa

ded

by [

Uni

vers

ity o

f H

ong

Kon

g L

ibra

ries

] at

08:

45 0

2 O

ctob

er 2

013

426 C. Crews et al.

found compared with the concentrations at whichrecoveries were assessed, and the uncertainty asso-ciated with the apparent recovery of jacobine, theresults given have not been corrected for recovery.Since the purity of the standards was unknown thepossibility of a considerable amount of bias in theseresults must be recognized.

PAs were detected in the four later samples from siteB. In each the largest contribution to the total wasfound to be made by jacobine but the possibilityof a positive bias as a result of an impure standardshould be borne in mind. Roughy similar contribu-tions to the total were made by jacozine, seneciophyl-line and senecionine, with jacoline making a muchsmaller contribution. The highest total concentration(0-056 mg/kg) was found in the earliest of thesesamples with sample 12, taken from the same hive afew weeks later, containing only 20% of this. Thisdoes not correlate with the pollen count data, butcould result from a number of factors which includedilution of the honey with nectar from other plants,instability of PAs in honey, and non-representativesampling.

In the samples coded JH1990 and JH1994 muchlarger total concentrations of PAs were found thanin samples from site B, although the Ragwort pollencounts were similar. It is noteworthy that in thesemore contaminated samples the total concentrationwas dominated by contributions from seneciphyllineand senecionine which together made up about 90-95% of the total, in marked contrast to the patternfound in the samples from site B. Only speculativereasons can be advanced, including the possibilitythat PA production was influenced by local climaticor other conditions. Borstel et al. (1989) have, how-ever, reported wide variations in the alkaloid compo-sition in different populations of two annual speciesof Senecio.

The identification of these alkaloids in sample JH1990was confirmed by recording full scan mass spectra.The total ion current chromatogram and recon-structed ion chromatograms used to facilitate loca-tion of appropriate peaks are given in Figure 4.Spectra from the main peaks (Figure 5) confirm theidentity and purity of the HPLC peaks with con-fidence.

No PAs were detected in the two retail samplesanalysed.

TIC

(a) jacozinem/z350

jacobine (b)

0'inrfr |i

m/z352

100-1seneciphylline

"m/2 334

A .

100nsenecionine

m/z336

10.00 15.00 20.00 2S.00 30.00 35.00

Figure 4. Total ion current and reconstructed ion currentchromatograms from full scan analysis of extract ofsample JH1990. Chromatograms were smoothed withone pass of a five-point Savitsky-Golay filter.

Conclusions

The method developed, based on SPE and HPLC-MS, offers advantages over the earlier publishedprocedures particularly in terms of speed, sensitivity,specificity and the ability to give confirmation ofidentity of suspected PA peaks. The extraction pro-cedure should be readily adaptable to the determina-tion of PAs and their metabolites in other aqueoussamples such as milk and body fluids. However, thechlorinated PA jaconine could not be recovered fromthe SPE column.

The results confirm that PAs were transferred fromRagwort into some honeys. PA contaminationreached levels of 1-48 and 0-42 mg/kg (not corrected

Dow

nloa

ded

by [

Uni

vers

ity o

f H

ong

Kon

g L

ibra

ries

] at

08:

45 0

2 O

ctob

er 2

013

Pyrrolizidine alkaloids in honey All

350

196

185

IMSpeakc

100-i 91

peakd

peakf

336

pealcg

100 150 200 250 300 350 400 450 500 550

Figure 5. Mass spectra of major peaks (labelled c, d, fand g in Figure 3) from full scan analysis of extract ofsample JH1990. Spectra across each peak were averagedbut no background subtraction was performed. The ion atm/z 424 which appears in the spectrum of Peak doriginated from a slightly earlier eluting compound.

for recovery) in two samples where the presence ofRagwort nectar was recognized by the beekeepers andthe honey was considered unpalatable. These levelsare comparable to those reported by Deinzer et al.(1977). Pollen counts and PA concentrations werepoorly correlated.

Acknowledgements

This work was funded by the Food Science Divisionof the UK Ministry of Agriculture, Fisheries andFood. We would like to thank Mr E. Fenner of the

British Beekeepers' Association for identifying testhives collecting and honey from them, and for con-ducting determinations of moisture content. Wewould like to thank Dr Johnson at Herbert J. Evansand Partners Public Analyst's Laboratory, CapelDewi, UK for conducting the pollen analyses. Wewould like to thank J. and M. A. Henson forsupplying honey containing Ragwort nectar.

References

ANON, 1959, The Weeds Act 1959 (London: HMSO).BICCHI, C., D 'AMATO, A., and CAPPELLETTI, E., 1985, Determination

of pyrrolizidine alkaloids in Senecio inaequidens D.C. bycapillary gas chromatography. Journal of Chromatography,349, 23-29.

BORSTEL, K. V., WITTE, L., and HARTMANN, T., 1989, Pyrrolizidine

alkaloid patterns in populations of Senecio vulgaris, S. vernalisand their hybrids. Phytochemistry, 28, 1635-1638.

BRADBURY, R. B., and CULVENOR, C. C. J., 1954, The alkaloids of

Senecio jacobaea L. Australian Journal of Chemistry, 1, 378—383.

BROWN, M. S., MOLYNEUX, R. J., and ROITMAN, J. N., 1994, A

general method for High Performance Liquid Chromatographyof pyrorolizidine alkaloid free bases and N-oxides Phytochem-ical Analysis, 5, 251-255.

COOPER, M. R., and JOHNSON, A. W., 1984, MAFF Reference Book161 (London: HMSO).

CULVENOR, C. C. J., EDGAR, J. A., and SMITH, L. W., 1981,

Pyrrolizidine alkaloids in honey from Echium plantagineum L.Journal of Agricultural and Food Chemistry, 29, 958-960.

DEINZER, M. L., THOMPSON, P. A., BURGETT, D. M., and ISSACSON,

D. L., 1977, Pyrrolizidine alkaloids: their occurrence in honeyfrom Tansy Ragwort (Senecio jacobaea L.). Science, 195, 497-499.

HARPER, J. L., and WOOD, W. A., 1957, Biological flora of the BritishIsles: Senecio jacobaea L. Journal of Ecology, 45, 617-637.

HARTMAN, T., and TOPPEL, G., 1987, Senecionine N-oxide, theprimary product of pyrrolizidine alkaloid biosynthesis in rootcultures of Senecio vulgaris. Phytochemistry, 26, 1639-1643.

HARTMAN, T., and ZIMMER, M., 1986, Organ-specific distribution andaccumulation of pyrrolizidine alkaloids during the life historyof two annual senecio species. Journal of Plant Physiology, 122,67-80.

HARTMANN, T., EHMKE, A., EILERT, U., VON BORSTEL, K., and

THEURING, C., 1989, Sites of synthesis, translocation andaccumulation of pyrrolizidine alkaloid N-oxides in Seneciovulgaris L. Planta, 177, 98-107.

HUXTABLE, R., STILLMAN, A., and CIARAMITARO, D., 1977, Char-

acterisation of alkaloids involved in human Senecio (pyrrolizi-dine) poisoning, Proceedings of the Western PharmacologicalSociety, 20, 445.

KEDZIERSKI, B., and BUHLER, D. R., 1986, Method for determinationof pyrrolizidine alkaloids and their metabolites by high perfor-mance liquid chromatography. Analytical Biochemistry, 152,59-65.

MOLYNEUX, R. J., and JAMES, L. F., 1990, Pyrrolizidine alkaloids inmilk: thresholds of intoxication. Veterinary and Human Tox-icology, 32, 94-103.

Dow

nloa

ded

by [

Uni

vers

ity o

f H

ong

Kon

g L

ibra

ries

] at

08:

45 0

2 O

ctob

er 2

013

428 Pyrrolizidine alkaloids in honey

PARKER, C. E., VERMAN, S., TOMER, K. B., REED, R. L., and BUHLER,D. R., 1990, Determination of Senecio alkaloids by thermo-spray liquid chromatography/mass spectrometry. Biomedicaland Environmental Mass Spectrometry, 19, 1-12.

RAMSDELL, H. S., and BUHLER, D. R., 1981, High performance liquidchromatographic analysis of pyrrolizidine alkaloids using areversed phase styrene-divinyl benzene resin column. Journalof Chromatography, 210, 154-158.

RIDKER, P. M., and MCDERMOTT, W. V., 1989, Comfrey herb tea andhepatic veno-occlusive disease. Lancet, March 25, 657—658.

RIDKER, P. M., OHKUMA, S., MCDERMOTT, W. V., TREY, C., andHUXTABLE, R., 1985, Hepatic venocclusive disease associated

with the consumption of pyrrolizidine containing dietarysupplements. Gastroenterology, 88, 1050-1054.

ROITMAN, J. N., 1983, Ingestion of pyrrolizidine alkaloids: a healthhazard of global proportions. ACS Symposium series 234(Xenobiotics in food and feeds), 345-378.

SAISHO, K., TOYODA, M., TAKAGI, K., SATAKE, M., TAKAHASHI, S.,YAMAMOTO, Y., KASAI, K., HASHIMOTO, S., and SAITO, Y.,1994, Identification of aconitine in raw honey that caused foodpoisoning. Journal of the Food Hygiene Society of Japan, 35(1),46-50.

SÜTLÜPINAR, N., MAT, A., and SATGANOGLU, Y., 1992, Poisoning bytoxic honey in Turkey. Archives of Toxicology, 67, 148-150.

Dow

nloa

ded

by [

Uni

vers

ity o

f H

ong

Kon

g L

ibra

ries

] at

08:

45 0

2 O

ctob

er 2

013