Following a request from the European Commission, the Panel on Contaminants in the Food Chain (CONTAM Panel) was asked to deliver a scientific opinion on the risks to human and animal health related to the presence of pyrrolizidine alkaloids (PA) in food and feed. PAs are toxins exclusively biosynthesised by plants. They are typical plant secondary metabolites against herbivores. It has been estimated that approximately 6000 plant species world wide, representing 3 % of all flowering plants, may contain pyrrolizidine alkaloids. PAs are mainly found in the distantly related angiosperm families of the Boraginaceae (all genera), Asteraceae (tribes Senecioneae and Eupatorieae) and Fabaceae (genus Crotalaria). The PA-content of plant material depends on a large number of factors (species, plant organ, harvest, storage, extraction procedures). Reported contents vary from trace amounts up to 19 % based on dry weight. The name pyrrolizidine is the chemical description of two-fused 5-membered rings with a nitrogen atom at the bridgehead. This motif is the central structure of a variety of PAs. PAs generally consist of an amino alcohol which is referred to as necine or necine base and an acid part which is called a necic acid. Most of the known PAs are esters of hydroxylated 1 methylpyrrolizidine or otonecine-type necine bases. To date, an estimate of approximately 600 different PA structures is known. The rich diversity is derived through factors such as combination of a pool of necine bases with an even larger pool of necic acids. The variability is further increased by the possible formation of monoesters at different positions and open or cyclic diesters. In addition, many PAs frequently co-occur in two forms, their N-oxide (PANO) and as tertiary base PAs.
Currently, only methods with mass spectrometric (MS) detection provide the prerequisites to analyse PAs at trace levels in food and feed. Basically, two MS based approaches are applied, either in combination with gas chromatography (GC) or high performance liquid chromatography (HPLC) in tandem mass-spectrometry (MS/MS) mode.
Following a European Food Safety Authority (EFSA) call for data covering mycotoxins and phytotoxins, including pyrrolizidine alkaloids, no replies for PAs were received even by the extended deadline in January 2011. Therefore, the industry was contacted and two submissions were received from one Member State covering the presence of a range of PAs in bulk honey as well as in retail honey. Overall, results for 14,604 samples of honey were reported to EFSA of which 13,280 samples concerned bulk honey and 1324 samples covered retail honey that is mostly blended and ready for consumption. No occurrence data in food other than honey were received. A further submission of 351 results was received from the National Competent Authority in one Member State covering PAs in feed. All food and feed samples were reported to be analysed by liquid chromatography-tandem mass spectrometry (LC-MS/MS). The limit of detection (LOD) for food samples was reported as 0.5 µg/kg for most PAs and 4.5 µg/kg for all PAs in feed after correction for the dry weight matter (to 88 %). As the submissions for food and feed were each coming from one Member State, they cannot be regarded as representative for the occurrence of PAs across Europe.
The two submissions covering results of PAs found in honey included testing of different ranges of alkaloids. The submissions had eight PAs in common (echimidine, echimidine-N-oxide, heliotrine, lycopsamine, retrorsine, senecionine, seneciphylline and senkirkine), while one submission included a further six (heliotrine-N-oxide, lasiocarpine, lycopsamine-N-oxide, retrorsine-N-oxide, senecionine-N-oxide and seneciphylline-N-oxide) giving a total of fourteen, and the other a further nine (acetylechimidine, acetylechimidine-N-oxide, acetylechiumine-N-oxide, echiumine, echiumine-N-oxide, echiuplatine, echiuplatine-N-oxide, echivulgarine and echivulgarine-N-oxide) giving a total of seventeen PAs.
The substitution method was used for expressing left-censored results, i.e. results at or below the limit of detection (LOD) or the limit of quantification (LOQ) were assigned either a value of zero (lower bound approach) or the value of the LOD or LOQ (upper bound approach). For bulk honey, the lowest proportion of left-censored results was recorded for lycopsamine at 49 %, followed by echimidine and echiumine at 56 % and 71 %, respectively. For retail honey, the situation was slightly different with the lowest proportion of left-censored results recorded for echimidine at 16 %, followed by lycopsamine and echiumine at 36 % and 45 %, respectively. Apart from acetylechiumine-N-oxide (one positive result), there were a further eleven PAs with only left-censored results reported for the retail honey of which six also reported 99 % left-censored data for the bulk honey.
The average levels of the different PAs for bulk honey varied from 0-9.7 µg/kg for the lower bound (LB) and 0.1-10 µg/kg for the upper bound (UB), and for retail honey from 0-6.5 µg/kg for the lower bound and 1-6.7 µg/kg for the upper bound. The maximum reported levels of PAs for bulk honey were for echimidine-N-oxide at 2031 µg/kg, echimidine at 1522 µg/kg, lycopsamine at 1448 µg/kg and seneciphylline-N-oxide at 1441 µg/kg. Maximum reported levels for retail honey were much lower with echimidine at 150 µg/kg, lycopsamine at 126 µg/kg and echiuplatine at 115 µg/kg.
The number of samples with any PAs above the LOD or LOQ is higher and the overall average level is lower when comparing retail honey to bulk honey. The maximum levels found in retail honey are only 10 % or less of the levels found in bulk honey. The eight PAs in common for the two submissions comprised between 75 and 90 % of the total upper bound sum of PAs in the respective alkaloid grouping. The sum of PA levels for the eight PAs in the total number of samples and for the PA sum in the groups of 14 and 17 alkaloids tested were used for the exposure calculations.
The CONTAM Panel decided that both acute and chronic exposure to PAs should be assessed. Consumption data were taken from the EFSA Comprehensive European Food Consumption Database (Comprehensive Database). Three representative age groups were selected for the exposure analysis, toddlers from 1 to 3 years of age, other children from 3 to10 years and adults from 18 to 65 years. Since there was little difference in consumption patterns between adults on the one hand and adolescents, elderly and very elderly population groups on the other, there was no need to present consumption data separately for these latter groups.
The highest acute exposure to PAs through retail honey was calculated for toddlers with a daily intake providing from 0.80 to 48.6 ng/kg body weight (b.w.) and from 3.3 to 114 ng/kg b.w. when applying the country range of minimum and maximum mean and 95th percentile consumption, respectively, and mean lower and upper bound PA concentrations. Consuming retail honey at the 95th percentile concentration level could potentially increase acute exposure two to three times compared to the mean concentration, with the highest value of 254 ng/kg b.w. calculated for toddlers consuming 40 g of retail honey in one day.
For the chronic scenario, PA exposure for the mean consumption and concentration scenario for toddlers could reach a high of 37.4 ng/kg b.w. per day in “honey consumers only”. However, it is probably closer to the result of 5.10 ng/kg b.w. per day calculated for consumption distributed over all survey participants in the respective age group, based on the uncertainty associated with interpolating a few survey days to long-term consumption and a rather low number of honey consumers in the surveys.
The theoretical exposure calculated for consumption of unblended (bulk) honey was in general about 50-100 % higher than the results of the calculations for retail honey or sometimes slightly higher than that. However, this calculation is mainly based on occurrence results for honey imported from countries outside Europe and such honey would usually be blended before retail.
Levels of PAs in 351 feed materials, sampled between 2006 and 2010 were submitted by one Member State. The compounds were merged into four groups of structurally related PAs (senecionine-, lycopsamine-, heliotrine- and monocrotaline-type PAs). In 55 % of samples, concentrations of PAs were below the LOD (4.5 μg/kg).
Livestock and domestic animals may be exposed to PAs by the consumption of forage and roughage contaminated with plant (parts) of Senecioneae and Boraginaceae spp. In particular, lucerne (alfalfa; Medicago sativa) forage was occasionally found to be contaminated with substantial amounts of PAs, which is most likely due to contamination with Senecio vulgaris. Horses may be more exposed than other livestock due to their high consumption of lucerne. Herbal mixtures used as feed which are contaminated with PA-containing plants (or their parts) are another possible source of exposure of livestock to PAs, but these feeds generally represent only a small proportion of the diet. Overall, the data on PA occurrence in feed are too limited to undertake a reliable estimate of the animal exposure.
Based on the available literature and the occurrence data submitted on PAs in honey and feed, the CONTAM Panel identified the following PAs (including the tertiary amine as well as the corresponding N-oxide forms) of particular importance for food and feed:
- Senecionine-type PAs: acetylerucifoline, erucifoline, integerrimine, jacobine, jacoline, jaconine, jacozine, retrorsine, senecionine, seneciphylline. These PAs occur particularly in the Senecioneae (Asteraceae family), but are also found in Crotalaria spp. (Fabaceae family).
- Lycopsamine-type PAs: acetylechimidine and isomers, echimidine and isomers, echivulgarine, lycopsamine and isomers, vulgarine. These PAs occur in the Boraginaceae family and in the Eupatorieae (Asteraceae family).
- Heliotrine-type PAs: europine, heliotrine, lasiocarpine. These PAs occur in Heliotropium spp. (Boraginaceae family).
- Monocrotaline-type PAs: fulvine, monocrotaline, retusamine, trichodesmine. These PAs occur in Crotalaria spp. (Fabaceae family).
Data on several 1,2-unsaturated PAs show that they are readily absorbed from the gastrointestinal tract and undergo extensive metabolism in mammals. 1,2-unsaturated PAs undergo metabolic activation by hepatic cytochrome P450 from experimental and livestock animals and humans to reactive pyrrole metabolites.
The toxicity of the 1,2-unsaturated PAs in experimental animals is characterised by hepatotoxicity, developmental toxicity, genotoxicity and carcinogenicity. Some also exhibit pulmonary toxicity. The liver is the primary site for genotoxicity of 1,2-unsaturated PAs. The 1,2-unsaturated PAs from different structural classes (i.e., retronecine, heliotridine, and otonecine; di-esters and mono-esters) undergo metabolic activation to reactive pyrrolic intermediates and form a common set of DHP adducts at dG and dA sites in rat liver DNA. These findings suggest that a genotoxic carcinogenic mechanism is applicable for all 1,2-unsaturated-PA esters and their N-oxides, which can be metabolically converted into PAs.
The concomitant induction of mutations compatible with DHP adduct formation in liver cells in transgenic rats, and the formation of hemangiosarcomas and hepatomas in riddelliine (retronecine type PA)-treated male and female rats and mice, provide strong evidence for a genotoxic mechanism for hepatocarcinogenicity. In contrast to 1,2-unsaturated PAs, 1,2-saturated PAs do not undergo metabolic activation to reactive pyrrolic species responsible for hepatotoxicity and genotoxicity. Therefore the CONTAM Panel decided to base the risk characterisation on the 1,2-unsaturated PAs.
Human case reports of poisonings due to PA containing herbal medicines and teas and large outbreaks of human poisonings, including deaths associated with grain crops contaminated with PA containing weeds, have demonstrated the toxicity of 1,2-unsaturated PAs in humans, affecting predominantly liver and lung. Poisoning with 1,2-unsaturated PAs in humans is characterised by acute hepatic veno-occlusive disease (HVOD). The acute disease is associated with high mortality, and a sub-acute or chronic onset may lead to liver cirrhosis.
The lowest known doses associated with acute/short-term toxicity in humans are reported to be 3 mg PA/kg b.w. per day (exposure of a boy for a 4 day-period, lethal outcome) and 0.8 -1.7 mg PA/kg b.w. per day (exposure of a girl for a 2 week-period, HVOD). The lowest known dose associated with long-term toxicity (HVOD) in humans is reported to be 15 µg PA/kg b.w. per day (exposure for a period of 6 months). Substantial long-term follow-up data or epidemiological studies to assess whether exposure to 1,2-unsaturated PAs results in cancer in humans are not available.
Overall, based on the present knowledge of metabolism, activation, DNA adduct-formation, genotoxicity and carcinogenicity studies, the CONTAM Panel concluded that 1,2-unsaturated PAs may act as carcinogens in humans. Therefore, the data from experimental animals are relevant to humans and the carcinogenicity data provide the most suitable basis for the risk characterisation.
Because 1,2-unsaturated PAs are genotoxic and carcinogenic, the CONTAM Panel concluded that it was not appropriate to establish a Tolerable Daily Intake (TDI), and decided to apply the Margin of Exposure (MOE) approach. A BMDL10 for excess cancer risk of 70 µg/kg b.w. per day was calculated for induction of liver haemangiosarcomas by lasiocarpine in male rats and used as reference point for comparison with the estimated dietary exposure. Lasiocarpine is amongst the most toxic of the PAs that have been tested. In the data on PAs submitted to EFSA, lasiocarpine was below the LOD or LOQ in 99 % of the honey samples. Some PAs such as lycopsamine, which was one of the most frequently detected PA in honey, are more than an order of magnitude less toxic. Since the toxicity influences the carcinogenicity, the carcinogenic potency of most PAs present in honey is likely to be lower, therefore basing the risk characterisation on the BMDL10 for lasiocarpine is a conservative approach, which is likely to allow also for concomitant exposure to co-occurring PAs.
In relation to PAs in retail honey, the MOEs for adults are in the ranges of 57,000 - 3,500,000 and 7400 - >7,000,000, at the mean and 95th percentile of consumption (based on maximum UB and minimum LB across European countries). The EFSA Scientific Committee has concluded that a MOE of 10,000 or higher, based on a BMDL10 from an animal study, is of low concern from a public health point of view. Taking into account the influence of samples with non-quantifiable levels of PAs, and the conservative nature of using the BMDL10 for a potent PA as the reference point, these MOEs are not likely to represent a health concern.
For toddlers, the MOEs are in the ranges of 14,000 - 7,000,000 and 1200 - >7,000,000, respectively. For other children, MOEs are in the ranges of 25,000 - 1,800,000 and 3900 - >7,000,000 at the mean and 95th percentile population consumption of honey.
For individuals who regularly eat locally produced unblended honey, exposure to PAs could be up to twice that of people who consume retail honey.
The CONTAM Panel concluded that there is a possible health concern for those toddlers and children who are high consumers of honey.
Estimates of acute dietary exposure to PAs in honey are four orders of magnitude lower than the lowest known PA dose associated with acute/short term toxicity in humans, indicating that PAs in honey will not lead to acute toxicity.
In addition to honey, there are other possible sources of dietary exposure to PAs. Based on the few available data indicating limited carry over from animal feed, meat, milk and eggs are not likely to be major sources, but this requires confirmation. Exposure to PAs from herbal dietary supplements can potentially be very much higher than dietary exposure from honey and is known to have caused human illness. Data on PAs in herbal dietary supplements are generally not available. However, if such supplements are prepared from PA-containing plants, then they could present a risk of both acute and chronic effects in the consumer. Furthermore, borage oil and Echium oil marketed as dietary supplements, and salad crops contaminated with PA-plants such as Senecio vulgaris (common groundsel), could present a risk to the consumer, but data were not available for the CONTAM Panel to perform exposure assessments or risk characterisation for these sources.
Since the publication of the EFSA (2007) opinion on pyrrolizidine alkaloids as undesirable substance in animal feed, a number of reports have been published on the effects of PA intake by livestock and companion animals, although the clinical signs and pathological findings described in the 2007 opinion remain valid. Even though all animal species are susceptible to both acute and chronic PA intoxication, the risk of PA poisoning in the EU appears to be low. Most poisonings reported recently have been due to accidental exposure, but in the absence of integrated data on the incidence in the EU, it has not been possible to quantify this risk.
The CONTAM Panel recommended, inter alia, that ongoing efforts should be made to collect analytical data on occurrence of PAs and PANOs in relevant food and feed commodities. These should include milk, eggs and meat. The PAs monitored should include at least the compounds identified in this opinion as markers for the main PA-containing plant families. In order to improve the analytical methods, there is a need for a larger and more diverse set of certified reference standards, covering both PAs and PANOs identified as markers of the main PA containing plant families. As limited literature results on honey showed considerable differences in the PA concentrations depending on the country of origin, more data are needed to correlate the occurrence of various PAs with the geographical and botanical origin. Finally, there is a need for toxicological data relating to the PAs most commonly found in honey.