Liquid Chromatography Tandem Mass Spectrometry ... - Springer Link

Food Anal. Methods (2015) 8:18–31 DOI 10.1007/s12161-014-9855-1

Liquid Chromatography Tandem Mass Spectrometry Detection of Targeted Pyrrolizidine Alkaloids in Honeys Purchased within Ireland Caroline T. Griffin & John O’Mahony & Martin Danaher & Ambrose Furey

Received: 2 January 2014 / Accepted: 26 March 2014 / Published online: 27 April 2014 # Springer Science+Business Media New York 2014

Abstract Pyrrolizidine alkaloids (PAs) are naturally occurring plant toxins which can contaminate foods such as honey. Due to this emerging health and food safety risk, there is an increasing need to establish high sample throughput detection methods which are sensitive, selective and robust for PAs. A sensitive and comprehensive method was developed to determine ten PAs and four PA-N-oxides (PANOs) in honey using liquid chromatography-electrospray ionisation tandem mass spectrometry. Target compounds were detected and quantified using mass spectrometry in dynamic multiple reaction monitoring mode (dMRM). The samples were pre-concentrated using a polymeric strong cation exchange (SCX) solid-phase extraction (SPE) which allowed simultaneous screening of PAs and PANOs in honey (87 % mean recovery ± 9 %). The method was validated for selectivity, taking into consideration the matrix effects, specificity, linearity, accuracy and precision. Good linear calibrations were obtained for all ten PAs and four PANOs in spiked honey samples (3.6–357.1 μg L−1; R2 ≥0.995). Acceptable inter-day repeatability was achieved for the target analytes in honey with % RSD values (n=17) less than 8 %. Limits of detection (LOD) and limits of quantitation (LOQ) were achieved with spiked PA and PANO samples, giving an average LOD of 1.0 μg kg−1 and LOQ of 3.4 μg kg−1. This method was successfully applied to 150 Electronic supplementary material The online version of this article (doi:10.1007/s12161-014-9855-1) contains supplementary material, which is available to authorized users. C. T. Griffin : A. Furey (*) Mass Spectrometry Research Centre (MSRC) and Team Elucidate Research Groups, Department of Chemistry, Cork Institute of Technology (CIT), Bishopstown Cork, Ireland e-mail: [email protected] J. O’Mahony : M. Danaher Food Safety Department, Teagasc Food Research Centre, Ashtown Dublin 15, Ireland

honeys purchased within Ireland between the years 2009 and 2011. The results show that PA contamination of honey is significant and comparative to other European studies, with 23 % of the samples testing positive for PA toxins in the range of 2.9 to 545.5 μg kg−1. Keywords Pyrrolizidinealkaloids(PAs) . Planttoxins . Honey . Food safety . Liquid chromatography-mass spectrometry (LC-MS/MS) . Solid-phase extraction (SPE)

Introduction Food analysis for the assessment of food safety and authenticity is a rapidly growing area due to the ever increasing global trade in food. Food safety receives the greatest attention due to an increasing number of food safety problems and growing consumer concerns. High profile food safety crises such as bovine spongiform encephalopathy (BSE), dioxins (or PCBs) in Belgian milk and Irish pork and melamine in Chinese milk products have harmed consumer confidence (Bánáti 2011). Most recently, concerns of misdescription, adulteration and authenticity of meat were raised during the ‘horse meat scandal’ (Premanandh 2013). As a result, the ‘from farm to fork’ methodology was adopted within Europe initiating a science-based approach to food safety testing and the establishment of key legislation (Malik et al. 2010). Liquid chromatography-mass spectrometry (LC-MS) plays a central role in the identification and quantitation of existing and emerging food contaminants (Fárre and Barceló 2013; Kinsella et al. 2010; Soler et al. 2007; Furey et al. 2002; Furey et al. 2001). The analytical technique of choice for food contaminants, particularly when dealing with complex matrices or multiple analytes, is triple quadrupole mass spectrometry due to its high selectivity and sensitivity when operated in selected reaction monitoring (SRM) or multiple reaction

Food Anal. Methods (2015) 8:18–31

19

monitoring (MRM) mode (Malik et al. 2010; Kantiani et al. 2010; Whelan et al. 2010; Núñez et al. 2012). As concerns in food safety grow, so too will the number of samples to be tested. With continuing advancements in LC-MS, the requirement of handling large volumes and varieties of samples in shorter analysis times is being satisfied. For example, the use of sub-2-μm particle analytical columns reduce total run/analyses times significantly while simultaneously improving target analyte chromatographic peak shape, resolution and sensitivity (Malik et al. 2010). Thus, the popularity of LC-MS has grown and is used extensively in food analysis (Núñez et al. 2012; Whelan et al. 2013; Power et al. 2012; Soler et al. 2006; Fernández Puente et al. 2004). Pyrrolizidine alkaloids (PAs) are naturally occurring plant toxins which are known to be hepatotoxic, teratogenic, genotoxic and carcinogenic to humans (Wiedenfeld 2011). PAs are gaining more attention due to the significant contamination risk they pose for food and feed. PAs occur in an estimated 3 % of all flowering plants, in particular the families Boraginaceae, Compositae and Leguminosae (Smith and Culvenor 1981). Representative structures are given in Fig. 1. Human exposure may result from ingesting herbal preparations containing PA plants (Wiedenfeld 2011; Rasenack et al. 2003), PA-contaminated grain (Kakar et al. 2011), honey produced from PA plants, carryover into milk or as PA-containing plant leaves in salads (Edgar et al. 2011; Hoogenboom et al. 2011). Wiedenfeld (2011) surmised that there have been a great number of reports of toxicity in humans due to PAs. However, where these reports are based on sub-chronic or chronic intoxications, the direct link cannot be proven due to the time lapse between the patient diagnosis

Fig. 1 Representative PA structures. Asterisk indicates those PAs detected in honey samples within this study

and a possible ingestion of PA-contaminated substances. Owing to this, the author only reports on intoxications where the source is unquestionably due to PAs. This amounts to 28 separate incidents, some fatal but all resulting in veno-occlusive disease (VOD). Within these cases, PA poisoning comes as a result of ingestion of herbal teas, wheat, millet and cereals and food or food supplements contaminated with PA-containing plants. Although the foremost source of PA intoxications is contaminated grains (Kakar et al. 2011; Asia 2008a; Asia 2008b; Azadbakht and Talavaki 2003; Altaee and Mahmood 1998; Chauvin et al. 1994), there is now a requirement for research to be undertaken in assessing all foods that have been linked to PA contamination. This will ensure that national regulatory food agencies can compile a comprehensive profile of the potential exposure to PAs in a given population. The European Food Safety Authority (EFSA) has recently conducted a safety assessment of PAs in food and feed. The Panel on Contaminants in the Food Chain (CONTAM Panel) concluded that there is a health concern for children and adults who are high consumers of honey. The panel assessed the risk of PAs to humans based on three representative age groups and the effects of both acute and chronic exposure levels. The importance of on-going analytical research into the PA content of food and feed was stressed, along with the need for more toxicological data concerning the PAs most commonly found in honey (EFSA - European Food Safety Agency 2011). This is a valid point as the honey results on which the EFSA report is based emerged from two research centres based in one EU member state only (Dübecke et al. 2011; Kempf et al. 2011a; Kempf et al. 2011b; Kempf et al. 2008). EFSA, the Committee

Family:

Boraginaceae

Genera:

All genera

Compositae

Particularly genera Senecioneae and Eupatorieae

Leguminosae

Crotalaria

Common PAs:

*Lycopsamine

*Echimidine

*Senecionine

*Seneciphylline

Crotaline

Trichodesmine

20

of Toxicity (COT) of Chemicals in Food (2008) and the Federal Institute of Risk Assessment (Bundesinstitut für Risikobewertung [BfR]; 2011) all recommend not to exceed an intake of 0.007 μg/kg bw/day of 1,2-unsaturated PAs. Doses at this level are unlikely to pose a risk of cancer. The aim of the work presented here was two-fold: Firstly, to assess the potential PA contamination in retail honeys sold within Ireland as a previous initial study conducted by our research group had shown their presence (Griffin et al. 2013); and secondly, to provide further data on the prevalence of PAs within honey and thus contribute to bridging the number of data gaps which exist in the area. In comparison with previously published LC-MS methods within this area, this validated method offers improvements in the method sensitivity and quantitation with each target analyte being matched to its own certified reference standard thus providing more accuracy. A significant reduction in time (for both sample extraction and analysis) and consumable costs was also achieved. The results show that PA contamination is significant, with 23 % of samples testing positive for PA toxins in the range of 2.9 to 545.5 μg kg−1. Our testing also provides an assessment of the stated origin of honeys being sold commercially within Ireland and data on the predominant PAs detected.

Materials and Methods Reagents and Materials Ultra-pure water (18.2 MΩ cm−1) was generated in-house using a Millipore (Cork, Ireland) water purification system. HPLC grade methanol and Chromasolv® grade acetonitrile, sulphuric acid (98 %) and ammonium hydroxide solution (33 % NH3 in H2O) were purchased from Sigma Aldrich (Wicklow, Ireland). Formic acid (extra pure, 98 %) used in mobile phase was obtained from Labscan (Gliwice, Poland). Formic acid used in sample preparation steps was supplied by Reagecon Ltd. (Co. Clare, Ireland). Strong cation exchange polymeric solid-phase extraction (SPE) cartridges, Strata-XC 33 μm (60 mg, 3 mL), were purchased from Phenomenex Inc. (Cheshire, UK) and 0.22 μm PTFE Chromacol syringe filters were from Lab Unlimited (Dublin, Ireland). All honey samples (n=150) were purchased from supermarkets (n=66; 44 %), health food stores (n=62; 41 %), market stalls (n=2; 1 %), greengrocers (n=15; 10 %) and local producers (n=5; 3 %) within Ireland (between the years 2009 and 2011) with no prior selection criteria. All brands of honey available at each location were sampled. Standards, Stock Solutions and Controls Reference pyrrolizidine alkaloids, including pyrrolizidine alkaloid N-oxides (PANOs), purchased and used in this study


were crotaline, retrorsine and senecionine (all analytical grade) [Sigma Aldrich, Ireland]; seneciphylline (>99 %) and senkirkine (>98 %) [Carl Roth, Germany]; echimidine, lycopsamine, otosenine, crotaline-N-oxide, retrorsine-N-oxide, seneciphylline-N-oxide and senecionine-N-oxide (all having purity > 98 %) [PhytoLab, Germany]; heliotrine (98 %) and trichodesmine (98 %) [Latoxan, France]. All pyrrolizidine alkaloid standards were prepared individually by dissolving 1 mg of PA or PANO in methanol (1 mL) to produce a primary stock solution of 1 g L−1. Working standards of 1,000 μg L−1 were prepared by serial dilution (1:1,000). Primary stock solutions were prepared every 3 months, dried under nitrogen and stored at −20 °C, and working standard solutions were prepared weekly and stored between 2 and 4 °C in solution (HPLC grade MeOH). All samples were quantified against the individual PA or PANO calibration curve produced from the reference standard. This is a more accurate and thorough quantitative methodology as the response of the free PA may not be the same as the PANO (Betteridge et al. 2005). Therefore, a detected PA/PANO should be quantified against its reference standard rather than using a free PA to determine its N-oxide or a retrorsine equivalent, as previously reported in literature (Dübecke et al. 2011). However, we are aware that in some cases, this was not possible due to a lack of commercially available reference PA/PANOs (Kempf et al. 2011a). A honey sample which did not contain detectable levels of PAs or PANOs was used as a negative control. This sample was extracted and subjected to SCX-SPE using the procedure outlined in the section “Sample Preparation”. The extract was spiked with PAs and PANOs at two known concentration levels, 3.6 and 71.4 μg L−1, to act as controls. A seven-point calibration curve was prepared in methanol ranging from 3.6 to 357.1 μg L−1 for all ten PAs and four PANOs. Sample Preparation All test honeys were stored at room temperature. The sample preparation was as per our prior publication (Griffin et al. 2013). This is based on previous methods reported by Boppré et al. (2005) and Betteridge et al. (2005) with modifications (mass of honey sample extracted was reduced, elimination of the need to maintain supernatant at 40 °C while SPE cartridge was loading, reduced supernatant load volume and reduced SPE cartridge bed size). These modifications were undertaken due to past experience with biological samples by our research group. Also, ion suppression is an important consideration and can be problematic for LC-MS quantitative analysis and the more effective and efficient means of addressing and reducing ion suppression is to analyse a more dilute sample (Furey et al. 2013). Prior to analysis, the honeys were warmed to 40 °C in a water bath and stirred vigorously in order to homogenise the sample. Samples (4 ± 0.1 g) were


weighed into centrifuge tubes (15 mL). Sulphuric acid (0.05 M; 6 mL) was added, and tubes were vortexed for 1 min and centrifuged at 6,000 rpm (3,421g) for 10 min. The resulting supernatant was transferred to a volumetric flask (10 mL), adjusted to the mark if necessary with sulphuric acid (0.05 M). This step ensures that all extracts are of equal volume before loading on SPE. A 2.5-mL aliquot (equivalent to 1 g honey) was applied to a preconditioned Strata-XC SPE cartridge. The preconditioning step consisted of methanol (2 mL) and deionised water (2 mL). After sample loading, the cartridge was washed with aqueous formic acid (0.1 %; 2 mL) followed by methanol (2 mL). PAs were eluted with ammoniated methanol (3 %; 9 mL) and dried in a turbovap at 45 °C under a nitrogen gas flow. Dried extracts were reconstituted in HPLC grade methanol (1 mL) and filtered using a syringe filter (0.22 μm) directly into a LC amber vial (2 mL). Liquid Chromatography-Mass Spectrometry Conditions The LC system employed was an Agilent Infinity 1290 with detection via an Agilent 6460 triple quadrupole mass spectrometer. The analytical column was a YMC Triart C18 (100 × 2.1 mm i.d., 1.9 μm particle size) equipped with a Waters Vanguard guard column (C18, 1.7 μm). All analytes were detected in the positive ionisation mode, with a drying gas temperature of 350 °C. Drying gas flow was set to 8 L min −1 . Sheath gas temperature was maintained at 400 °C, at a flow rate of 10 L min−1. Nebulizer pressure was maintained at 35 psi and capillary voltage set to 3,500 V. The column was maintained at a temperature of 30 °C and the flow rate was set to 0.2 mL min−1. The injection volume was 2 μL. The mobile phase consisted of (A) water with 0.05 % formic acid and (B) 100 % acetonitrile. A gradient flow was employed as follows: 0–2 min 10 % B, 2.1–6 min 10–30 % B, 6.1–9 min 50 % B, 9.1–12 min 100 % B, 12.1– 13 min 100–10 % B and 13.1–15 min 10 % B starting mobile phase conditions to re-equilibrate (this equates to 2.5 column void volumes which was experimentally deemed sufficient). Spectral data was acquired using N2 as the collision gas, in dynamic multiple reaction monitoring (dMRM) mode. The dMRM allows for optimised detection windows and cycle time based upon parameters imported from the optimiser software. A constant cycle time of 7 ms was used and a window of ±0.2 min from the relative retention time of each compound. Detection conditions for individual analytes, the two most abundant MRM transitions and the percentage target ratios between them are described in Table 1. The data was processed using Agilent MassHunter Workstation software (version B.040.01).

21

Method Quantitation Identification of ten free tertiary PAs and four PANOs was accomplished by comparing the retention time (±0.2 min) and the target ratio (±20 %) of the two selected precursor to product ion transitions for each PA and PANO reference standard. Calibration curves were constructed by plotting the peak area of the quantitative ion of each standard at concentrations of 3.6, 5.7, 7.1, 35.7, 57.1, 71.4, 142.9 and 357.1 μg L−1. An external calibration approach was used as there was no suitable internal standard or deuterated PA available. When a sample tested higher than the given calibration range, it was serially diluted to fall within the range. The limits of detection (LOD) and quantification (LOQ) were determined under chromatographic conditions using a signal-tonoise ratio (S/N) of 3 and 10, respectively (Table 2). Method precision was assessed using intra- and inter-day variations. For intra-day variability, the two spiked controls of known PA concentrations (3.6 and 71.4 μg L−1) were analysed and calculated (n=4) in 1 day. For inter-day variability, the two spiked controls were analysed in duplicate for a further six sequences over 6 days (n=17). Precision is stated as the relative standard deviation (% RSD) of the spiked honeys, as given in Table 3. Also reported in Table 3 is the recovery efficiency of the methods calculated by comparing the measured concentration to that of the spiked concentration.

Results and Discussion Sample Cleanup High sensitivity is essential for the analysis of PAs in food produce where concentration levels can be low or the food matrix is challenging, such as honey (Gajda et al. 2013). Honey is an amalgamation of many different classes of compounds. Therefore, the composition of honey can be problematic for sample cleanup especially when trace level detection and identification is required. The composition can also be affected by region, vegetation and season (Kaufmann et al. 2010). SPE provides the best means of reducing interferences from the honey matrix, improving sensitivity and allowing unambiguous identification of selected analytes, as found by Blasco et al. (2011) when analysing for insecticides within honey. SPE is also very effective at reducing the impact of ion suppression during LC-MS analysis (Furey et al. 2013). Mroczek et al. (2002) strongly recommended the use of polymeric strong cation exchange SPE for pyrrolizidine alkaloid analysis. This allows the PANOs to interact with the active sites of the sorbent rather than a cation exchange mechanism. Therefore, both PAs and PANOs can be isolated simultaneously via polar and cation exchange interactions and hence mixed mode.

22 Table 1 LC-MS/MS acquisition parameters employed for PA compounds under investigation


Compound

Rt (min)

Crotaline

2.59

Crotaline-N-oxide

3.64

Lycopsamine

3.76

Retrorsine

5.78

Heliotrine

5.95

Retrorsine-N-oxide

5.97

Trichodesmine

6.04

Otosenine

6.23

Seneciphylline

6.33

Seneciphylline-N-oxide

6.64

Senecionine

7.00

Senecionine-N-oxide

7.26

Echimidine

7.30

Senkirkine

7.42

Q quantifier, q qualifier

We have previously applied the protocol outlined here to honey samples (Griffin et al. 2013) with significant time and commercial modifications from those reported previously (Betteridge et al. 2005; Boppré et al. 2005). These were as follows: the sample size loaded on the SPE cartridge was reduced from 20 to 1 g, supernatant load volume was reduced from ~25 to 2.5 mL and the SPE cartridge bed size was reduced from 500 to 60 mg. Due to the reduction in the load volume, it was not necessary to maintain the supernatant at 40 °C prior to loading as per Betteridge et al. (2005), nor was there a need to divide the supernatant into equal aliquots and apply to separate SPE columns. Thus, reduction in sample load and SPE cartridge bed size provides a significant cost saving. It was necessary to heat the honey to 40 °C before removing the testing portion as this ensured a representative sample was obtained from a homogeneous mixture and that thorough mixing with the acidic solution for extraction was facilitated. It is noteworthy that no incidences of cartridge blockage were

Precursor ion (m/z)

Product ion (m/z)

Target ratio (%)

Fragmentation energy (V)

Collision energy (V)

Q q Q q Q q Q q Q q

326.2 326.2 342.2 342.2 300.2 300.2 352.2 352.2 314.2 314.2

194.1 138.1 137.1 119.1 138.1 94.1 120.1 94.1 156.1 138.1

52

132

40

150

62

114

144

132

25

114

28 32 28 32 16 24 28 40 24 16

Q q Q q Q q Q q Q q Q q Q q Q q Q q

368.2 368.2 354.2 354.2 382.2 382.2 334.2 334.2 350.2 350.2 336.2 336.2 352.2 352.2 398.2 398.2 366.2 366.2

120.1 94.1 222.1 120.0 168.1 122.0 120.1 94.1 119.7 94.1 138.1 120.1 119.7 94.1 220.1 120.1 168.1 150.1

69

150

80

132

74

114

85

114

51

120

73

132

40

150

17

104

35

132

40 52 28 44 24 32 28 36 36 48 28 28 36 52 12 20 28 24

experienced as previously reported by other authors. Also, the samples did not need to be filtered overnight, as performed by Dübecke et al. (2011), to remove any possible particulate matter. While all our samples were retail within this study and would not contain excess particulates, we did in a previous study (Griffin et al. 2013) process an unfiltered honey sample which was provided by a local beekeeper. Neither blockages nor obstructed flow was experienced with this sample, but this may be on account of reduced sample loading (supernatant was made up into a 10-mL volumetric flask and a fixed aliquot equivalent to 1 g honey was loaded onto the SPE cartridge). The dual ability of SPE in capturing both PAs and PANOs is not only beneficial from a time aspect in sample preparation but it also eliminates the use of hazardous chemicals such as dichloromethane, chloroform and LiAlH4 as used in liquid-liquid extraction (LLE) and reduction steps used in other reported sample preparations (Kempf et al. 2011a; Betteridge et al. 2005). In a recent publication, a QuEChERS sample extraction was used followed by

Food Anal. Methods (2015) 8:18–31 Table 2 Summary of LC-MS/ MS (0.2 mL min−1) calibration data

a

Units expressed as microgram per liter as prepared in MeOH

b

Units expressed as microgram per kilogram as prepared in honey extract

23

Compound

Calibration range (n=9, μg L−1)a

Linearity (R2)

LOD (n=17, μg kg−1)b

LOQ (n=17, μg kg−1)b

Crotaline Crotaline-N-oxide Lycopsamine Retrorsine Heliotrine Retrorsine-N-oxide Trichodesmine Otosenine Seneciphylline Seneciphylline-N-oxide

3.6–142.9 3.6–357.1 3.6–357.1 3.6–142.9 3.6–142.9 3.6–142.9 3.6–142.9 3.6–357.1 3.6–142.9 3.6–142.9

1.000 0.998 0.999 0.998 0.998 0.997 0.996 0.996 0.999 0.998

0.9 0.7 0.8 0.9 0.8 0.8 3.9 0.5 1.0 0.7

3.0 2.4 2.6 3.1 2.8 2.7 12.9 1.8 3.2 2.4

Senecionine Senecionine-N-oxide Echimidine Senkirkine

3.6–142.9 3.6–142.9 3.6–142.9 3.6–357.1

0.997 0.995 0.996 1.000

0.8 0.7 1.1 0.7

2.7 2.5 3.7 2.3

Table 3 Mean recoveries and relative standard deviations (% RSD) for control samples under LC-MS/MS (0.2 mL min−1) investigation Compound

Precursor ion (m/z)

Spiked conc. (μg kg−1)a

Mean recovery (%, n=4)

Intra-day (% RSD, n=4)

Mean recovery (%, n=17)

Inter-day (% RSD, n=17)

Crotaline

326.2

Crotaline-N-oxide

342.2

Lycopsamine

300.2

Retrorsine

352.2

Heliotrine

314.2

Retrorsine-N-oxide

368.2

3.6 71.4 3.6 71.4 3.6 71.4 3.6 71.4 3.6 71.4 3.6

70 80 75 80 85 104 73 99 72 108 76

6 1 5 0 6 2 5 3 4 2 4

79 79 76 79 75 103 72 96 80 106 75

8 3 7 3 7 2 7 3 6 2 6

Trichodesmine

354.2

Otosenine

382.2

Seneciphylline

334.2

Seneciphylline-N-oxide

350.2

Senecionine

336.2

Senecionine-N-oxide

352.2

Echimidine

398.2

Senkirkine

366.2

71.4 3.6 71.4 3.6 71.4 3.6 71.4 3.6 71.4 3.6 71.4 3.6 71.4 3.6 71.4 3.6 71.4

93 75 95 70 102 77 98 71 90 76 100 71 91 78 125 87 98

3 4 1 3 2 5 3 3 2 2 2 2 4 5 2 2 2

91 76 93 74 100 76 96 72 88 76 96 71 87 73 119 79 96

4 8 3 4 2 7 3 6 3 6 3 7 4 7 3 6 3

a

PA concentration spiked into blank honey extract

24

dispersive SPE (dSPE) cleanup (Martinello et al. 2014). The authors employed a 90-min agitation step with zinc to reduce the PANOs to PAs. However, for the protocol outlined in this study, the full sample preparation and cleanup can be performed in less than 90 minutes. Method Performance The linear regression for ten PA and four PANO reference compounds were determined using the optimised dynamic MRM (dMRM) method with data acquisition performed using Agilent MassHunter software. The procedure utilises a dMRM technique which allows transitions to be constructed from a retention window specified for each target analyte, typically ±0.2 min. This permits greater selectivity and sensitivity and improves quantitation in comparison to conventional time segments. In the MS/MS spectral analysis of the standard PAs, characteristic fragments were obtained. Toxic retronecine-type PAs produce fragments at m/z 94, 120 and 138, while otonecine-type PAs give fragments at m/z 150 and 168 (Table 1). These characteristic fragments are consistent with those previously reported (Colegate et al. 2005; Pedersen and Larsen 1970). Given in Fig. 2 is a total LC chromatogram for the method illustrating the order of elution for the PAs and PANOs included within this study. Figure 3 outlines the MRM ion trace for two transitions of a PA standard, lycopsamine (Fig. 3a), which are matched to a positive honey sample found to contain this PA (Fig. 3b). Quantitation was performed using constructed calibration curves with results as outlined in Table 2. We noted that the linearity for some PAs and Fig. 2 LC-ESI-MS/MS chromatogram illustrating the elution order of all ten free tertiary PAs and four PANOs (concentration 71.4 μg kg−1) included within this study


PANOs did not continue beyond the 357.1 μg L−1 concentration. However, the correlation coefficients (R2) of all calibration curves used were greater than 0.99. The LOD and LOQ for each PA and PANO were calculated (Table 2). The improvement in both sensitivity and in calibration linearity achieved with this method is approximately 100-fold greater than previous methods using an LCQ ion trap mass spectrometer (Betteridge et al. 2005). Two controls (3.67 and 71.4 μg L−1) were prepared by spiking into PA-negative honey (previously tested) to assess trueness and intra- and inter-day % RSD. The overall mean recoveries were between 75 and 97 % with intra- and inter-day variations (RSD) less than 6 and 8 %, respectively (Table 3). Recoveries and RSDs for the higher spiked control of 71.4 μg L−1 proved better. This was to be expected as the lower spiked concentration of 3.6 μg L−1 was closer to the LOQ for some PA analytes (Table 2). The basis for running the method under LC-ESI-MS/MS conditions was to transfer our previous method (Griffin et al. 2013) from LC-IT-MS and between analytical laboratories. An UHPLC-MS/MS method was also investigated in order to establish a high throughput analysis as a time difference of 5 min per sample would be substantial when running sequences of 100 samples or more. This UHPLC method would achieve a saving on instrument time of 33 %. All conditions were comparable except for an increased flow rate of 0.6 mL min−1. A statistical analysis was carried out using SPSS Statistics 20.0 package. The related or paired t test was conducted to examine whether there was a significant difference between the recoveries of the higher spiked concentration of 71.4 μg L −1 from the LC-MS/MS


(a) Lycopsamine (m/z 300.2) Calibration Standard 71.4 µg L-1 x10

5

94.1)

3.76 min.

1

Ratio = 62.5 (100.2 %) Counts

+ MRM (300.2 Counts

Fig. 3 a MRM ion trace for two transitions of PA standard (lycopsamine) matched to b a retail honey sample found to contain the PA lycopsamine

25

x10

5

1

0.8

0.8

0.6

0.6

0.4

0.4

0.2

0.2

0

0 3

3.5 4 4.5 Acquisition Time (min)

3

300.2

94.1

300.2

138.1


(b) Lycopsamine detected in retail honey (Sample #3; 392.6 µg kg-1) x10

94.1)

3.77 min.

2

Ratio = 68.5 (109.7 %) Counts

Counts

+ MRM (300.2 4

x10

4

2

1.5

1.5

1

1

0.5

0.5

0

0 3


(0.2 mL min −1 ) and UHPLC-MS/MS (0.6 mL min −1 ) methods. The level of significance selected was 5 % ( α = 0.05). The test revealed that there was no significant difference (P=0.099; P>0.05) between the two methods. Althou gh rec ove ries for the high er flow ra te (0.6 mL min−1) were comparable with the method detailed herein, ranging from 71 to 125 %, there was a decrease noted in LOD and LOQ (Supplementary Material). This decrease in sensitivity is possibly due to the increased flow rate as ESI can perform better under lower flow rates (Wilson et al. 2007; Guillarme et al. 2010). Application of Method and Results Determination of PAs is usually carried out by liquid chromatography (LC) or gas chromatography hyphenated with mass spectrometry (LC-MS or GC-MS). LC-MS provides the benefit of detecting both free tertiary PAs and PA-N-oxides simultaneously without having to perform a further reduction step. Thus, sample preparation for GC analysis may be overcomplicated and time consuming. Previously, LC has been coupled to ion trap MS (Griffin et al. 2013; Betteridge et al. 2005; Boppré et al. 2005), electrospray ionisation (ESI) triple quadrupole (QqQ) MS (Dübecke et al. 2011; Kempf et al. 2011a) and ion trap with atmospheric pressure chemical

3

300.2

94.1

300.2

138.1


ionisation (APCI) (Beales et al. 2004) for the detection and determination of PAs in honey. Although APCI affords good stability, it is inclined to have lower sensitivity in comparison to ESI. Therefore, ESI-QqQ-MS is the detection method of choice for PAs in food matrices such as honey. Several studies on the PA content of honey and pollen products have been performed. Those carried out and published up to 2008 are summarised in Kempf et al. (2010), and since then, a further seven have been published (Dübecke et al. 2011; Kempf et al. 2011a; Kempf et al. 2011b; Griffin et al. 2013). However, some of these studies are limited (Cao et al. 2013) and selected, targeting a particular honey (Kempf et al. 2011a; Betteridge et al. 2005), floral honeys (Kempf et al. 2008; Colegate et al. 2005; Orantes-Bermejo et al. 2013) or pollen source (Kempf et al. 2010). Table 4 gives a summary of previous literature published for the detection of PAs and PANOs in honeys. The targeted multi-reaction monitoring (MRM) approach is best but can be limited by the lack of reference standards. In the large-scale study conducted by Dübecke et al. (2011), the authors included 12 PAs and seven PANOs, although only ten of the PA/PANOs had reference standards while the other identifications came from comparisons with literature. Thus, the authors state that some results are regarded as approximates (Dübecke et al. 2011). This is not

LC-(+ESI)-QMS

70 retail honeys

SCX-SPE

SCX-SPE

1 retail honey 1 bulk/supplier honey 2 mead

696 retail honeys and 2,839 bulk honeys (from export drum)

C18 aQ

SCX-SPE

SCX-SPE

SCX-SPE

216 retail floral honeys

9 retail honeys (5 honeys of E. vulgare)

63 bulk honeys (floral honeys)

b

a

LC-APCI-IT-MS

LC-APCI-IT-MS

HRGC-MS

HRGC-MS

5 PAs

–

C8

9 PAsb

7 PAs

b

6 PAs

6 PAs

LOD 9–20 μg kg−1 LOQ 30–68 μg kg−1a

LOD–N.R. LOQ 1–3 μg kg−1

Denotes that PAs were isolated from plant sources and authenticated by NMR and MS

Individual plots and literature

Individual plots and equivalents Individual plots and literature

LOD 1.5 μg kg−1 LOQ–N.R.

LOD–N.R. LOQ 10 μg kg−1 LOD–N.R. LOQ–N.R.

Individual plots LOD 3 μg kg−1 and equivalents LOQ 10 μg kg−1

Individual plots LOD–N.R. and equivalents LOQ 10 μg kg−1

LOD–N.R. LOQ 1 μg kg−1 (all except monocrotaline)

Individual plots LOD–N.R. LOQ 1–3 and equivalents μg kg−1

3–2,200 μg kg−1

17–2,850 μg kg

−1

10–130 μg kg−1 (fennel honey) 10–484 μg kg−1 (mead) 10–40 μg kg−1 (candy) Average PA content of 5,170 μg kg−1 19–120 μg kg−1

780 μg kg−1 (bulk honey) 236–540 μg L−1 (mead) 1–267 μg kg−1 (retail honey) 1–1,087 μg kg−1 (bulk honey) 311–411 μg kg−1 (Echium honey) 1–625 μg kg−1 (retail honey)

Average PA content of 310 μg kg−1

1–237 μg kg−1

Pollen products contain a higher level of PAs. Quantitation as retronecine equivalents Floral honeys selected. Quantitation as retronecine equivalents Retail honeys purchased in New Zealand. Calibration linearity questionable after 5 μg mL−1 Bulk honeys sourced in Australia. Bias towards floral PA-containing sources, as advised by supplier

Persistence of PAs in processed foods that contain honey as an ingredient. Quantitation as retronecine equivalents

Retail honeys sourced in Germany. Predominant PAs in retail samples; echimidine, lycopsamine and senecionine. Retail honeys sourced in Germany. Predominant PAs in retail samples; echimidine and lycopsamine

Honeys purchased in Italy. Predominant PAs echimidine and lycopsamine Honeys sourced in Spain. Predominant PAs echimidine, lycopsamine and their N-oxides Honeys purchased in Ireland. Predominant PAs echimidine and lycopsamine Limited study presented as short communication. Echimidine and trace echiumine

Levels of PA/PANOs Comments detected

LOD 0.021–1.39 μg kg−1 1–169 μg kg−1 LOQ 0.081–4.35 μg kg−1

Quantitation

Use of equivalents LOD 50 μg kg−1 LOQ–N.R.

Individual plots

Individual plots

9 (no PANOs) Individual plots

8 PAs and 2 PANOs

Heliotrine

11 PAs (no PANOs)

9 PAs and 7 PANOs

–

C18 aQ

Calibration

9 (no PANOs) Individual plots

Denotes that the corresponding microgram per kilogram units were calculated

N.R. not reported

SCX-SPE

55 retail pollen products (food supplements)

16 retail honeys (8 QuEChERS and LC-(+ESI)-QqQ-MS C18 Echium, 8 random) online SPE (comparison to and bulk honey (from GC-MS reported) hives in location of J. vulgaris) Food produce with honey SCX-SPE HRGC-MS – ingredient 5–37 %

LC-(+ESI)-QqQ-MS C18

LC-(+ESI)-IT-MS

LC-(+ESI)-IT-MS

SCX-SPE

50 retail honeys

C18 aQ

LC-(+ESI)-QqQ-MS C18 aQ

C8

LC stationary PA reference phase standard(s)

103 floral honeys (bulk SCX-SPE prior to immediate sale and retail samples)

QuEChERS

Detection method

Number of samples tested Sample preparation

Table 4 Summary of previous literature for PA detection in honey

(Beales et al. 2004)

(Betteridge et al. 2005)

(Kempf et al. 2008)

(Kempf et al. 2010)

(Kempf et al. 2011b)

(Kempf et al. 2011a)

(Dübecke et al. 2011)

(Cao et al. 2013)

(Griffin et al. 2013)

(Orantes-Bermejo et al. 2013)

(Martinello et al. 2014)

Reference

26 Food Anal. Methods (2015) 8:18–31


27

uncommon as there is a lack of PA and PANOs commercially available (Kempf et al. 2011a). Within Europe, in excess of 200,000 tons of honey is consumed annually. Over half of this is imported from outside the European Union (EU) to satisfy demands. In 2010, EFSA reported that 148,000 tons of honey was imported between 27 EU member states. Some countries such as Belgium, Germany, Ireland and the Netherlands import in the range of 78 to 93 % of the honey they consume. EFSA suggests that these countries may experience a higher PA content in their retail honeys due to the blending of non-EU with EU honeys (EFSA, 2011). Thus, we deemed that a survey of the current (2009–2011) retail honey market in Ireland for probable PA contamination was pertinent. Our studies also served as a test for the stated origin of the retail honeys. EFSA reported that approximately 40 % of imported bulk honeys to Ireland originate from Central and South America, 13 % from Oceania and 25 % from domestic product. However, as Fig. 4 illustrates, we have found that the percentage from Oceania far outweighs that from Central and South America, and the domestic product is far less than the 25 % stated. Figure 4 displays the origin of the retail honey as a percentage for the 150 samples included in this study. The method described herein was applied to 150 honeys all purchased within Ireland from food retail outlets with no prior selection criteria. Of the 150 samples purchased, 34 samples tested positive for one or more PAs and/or PANOs included in the target analysis. Since there are no minimal risk levels (MRLs) set for PAs and/or PANOs, a positive within this study is one where the level detected is above the LOD of the method (Table 2). Table 5 details all 34 positive samples detected, listing the country of origin as per the retail label, the classification or type of honey, the supplier or retail outlet and the individual and total sum of PAs detected. In particular, the PAs lycopsamine and echimidine were most prevalent. This finding is in agreement with previous publications (Dübecke

Fig. 4 Country of origin (as stated on retail label) as a percentage of samples taken (n=150) between the years 2009 and 2011

et al. 2011; Kempf et al. 2011a; Martinello et al. 2014; Cao et al. 2013; Orantes-Bermejo et al. 2013) and is summarised in Table 4. Identification was confirmed by retention time, two characteristic MRM transitions and specific target ion ratios between the quantifier and qualifier ion (±20 % of actual ion abundances based on those produced by the certified reference standards). It is important to note that all PAs and PANOs used in the development of this method were by means of authentic reference standards (as outlined in the section “Standards, Stock Solutions and Controls”). Therefore, there was no requirement to convert the results obtained to PA equivalents, such as retrorsine equivalents, which could lead to under- or overestimations of PA quantitation and toxicity, as reported in previous literature (Kempf et al. 2008; Cao et al. 2013). Our results show that 75 % of the positive samples have a total sum of PA concentrations ranging from 2.9 to 106 μg kg−1, with a median value of 31 μg kg−1. This correlates with ranges for retail honey as previously reported (Dübecke et al. 2011; Martinello et al. 2014; Orantes-Bermejo et al. 2013) and outlined in Table 4. The relationship between the total concentration range of all PAs and/or PANOs detected within a sample and its country of origin as stated on the label is summarised by means of a pie chart in Fig. 5. This figure is based on a similar model used by Kempf et al. (2008). Thirty-three percent of the positive samples were classified as blends of EU and non-EU honeys. The total sum of PAs and PANOs within these samples ranged from 8 to 160 μg kg−1, consistent with those reported by Martinello et al. (2014) of 1–169 μg kg−1. The authors stated that the highest PA concentrations found in their study were from blends of EU and non-EU honeys. The higher PA concentrations in this study, using the total sum of PAs, were detected in samples originating from Australia and New Zealand (Oceania). These have a PA concentration range from 119 to 546 μg kg−1. Those honeys contributing to high levels of PAs originate from Oceania and blends of EU and non-EU

29% 28%

17%

7% 5%

5%

4%

3% 1%

Blend EU Oceania & Non-EU

EU

Non-EU

North Central & America South America

Ireland

Blend Ireland & Non-EU

Asia

28


Table 5 Retail honey samples which tested positive for PAs (n=34), listed in descending order of the concentration of PA(s) detected and classified by country of origin, honey composition and supplier Sample ID Country of origin (per label)

Classification Suppliera Lycopsamine Echimidine Heliotrine Seneciphylline Senecionine Total sum of (μg kg−1) (μg kg−1) (μg kg−1) (μg kg−1) PAs detected (μg kg−1) (μg kg−1)

1 2 3 4 5 6 7 8 9 10

New Zealand Australia New Zealand Australia New Zealand Blend of EU and non-EU Australia Spain New Zealand Blend of Australia and New Zealand Non-EU Mexico and Bulgaria Blend of EU and non-EU Australia Australia Blend of EU and non-EU Blend of EU and non-EU Mexico New Zealand Blend of EU and non-EU Australia Blend of EU and non-EU Blend of EU and non-EU Blend of EU and non-EU Blend of EU and non-EU

Manuka Clear Manuka Clear Forest Floral Organic Floral Manuka Manuka

HF SM HF SM SM SM SM MS HF SM

ND 347.1 392.6 142.5 157.7 130.6 146.7 ND 118.5 56.6

545.5 123.6 ND 86.3 5.3 28.7 5.0 134.3 ND 11.4

ND NQ ND ND ND ND ND ND ND ND

ND ND ND ND ND ND ND ND ND ND

ND ND ND ND ND ND ND ND ND ND

545.5 470.7 392.6 228.8 163.0 159.3 151.7 134.3 118.5 68.0

Clear Organic Forest Clear Clear Clear Floral Organic Floral Floral Clear Floral Floral Floral Clear

SM SM SM SM SM GG SM HF GG HF SM SM HF HF SM

35.1 43.1 11.4 37.5 15.6 34.0 30.9 29.9 16.0 ND 18.8 21.5 17.5 15.9 6.0

9.2 ND 23.8 ND 18.6 ND ND ND 12.0 20.0 6.2 ND ND ND ND

ND ND ND ND ND ND ND ND ND ND ND ND ND ND ND

ND ND 5.7 ND ND ND ND ND ND 5.3 ND ND ND ND ND

ND ND ND ND ND ND ND ND ND ND ND ND ND ND 8.4

44.3 43.1 40.9 37.5 34.2 34.0 30.9 29.9 28.0 25.3 25.0 21.5 17.5 15.9 14.4

Blend of EU and non-EU New Zealand EU Blend of EU and non-EU Blend of EU and non-EU Blend of EU and non-EU Mexico and Spain Blend of Australia and New Zealand South America

Clear Manuka Floral Forest Organic Set Floral Manuka

SM HF SM SM SM HF SM SM

14.2 11.7 ND ND 10.9 10.0 8.7 ND

ND ND 12.5 11.4 ND ND ND 2.9

ND ND ND ND ND ND ND ND

ND NQ ND ND ND ND ND ND

ND ND ND ND ND ND ND ND

14.2 11.7 12.5 11.4 10.9 10.0 8.7 2.9

Organic

HF

2.9

ND

ND

ND

ND

2.9

11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 a

Supplier codes: HF health food store, SM supermarket, MS market stall, GG greengrocer

ND not detected,