Novel strategies for risk assessment of pyrrolizidine alkaloids

Scope: Botanicals and botanical preparations may contain natural constituents that are of concern for human health. One group of such natural toxic compounds that may raise a concern is the group of pyrrolizidine alkaloids (PAs). Especially 1,2-unsaturated PAs are hepatotoxic and may act as genotoxic carcinogens in humans. At the current state-of-the-art, risk assessment of botanicals and botanical preparations is generally not required before they can enter the market, thus may pose a potential risk to human health.

Objective: The present thesis aimed to perform the risk assessment for PAs derived from botanical products following daily life-time exposure and also more realistic exposure scenarios. Another aim of the present thesis was to investigate whether animal-free testing strategies could be of use in tackling data and knowledge gaps for PAs by predicting in vivo toxicity of different PAs and whether proofs-of-principle for applying such alternative testing approaches could be provided for two selected PA model compounds, riddelliine and lasiocarpine.

Material and methods: Herbal teas, herbal medicines and plant food supplements (PFS) were bought from different counties. LC-MS/MS was used to detect the PA levels in these botanical and botanical preparations. Because PAs are genotoxic carcinogens, Margin of exposure (MOE) approach was applied for risk assessment. For development of the physiologically based kinetic (PBK) modelling to predict in vivo liver toxicity for lasiocarpine and riddelliine in rat and human, the kinetic parameters were obtained from in vitro incubations assays. The microsomal incubation assays were performed with rat and human tissue fractions to determine Vmax and Km values for lasiocarpine and riddelliine clearance, using a substrate depletion approach. The MTT assay was used to detect in vitro liver toxicity for lasiocasrpine and riddelliine.  In a subsequent step, the same approach was used for prediction of genotoxicity in rat, another endpoint relevant for PA toxicity. To this end, the in vitro concentration-responses curves obtained from in vitro genotoxicity studies using the ƴH2AX assay were translated into in vivo dose-response curves using PBK modelling-facilitated reverse dosimetry.

Main results: When consumption of one cup of tea a day during a whole lifetime would result in MOE values lower than 10000 for several types of herbal teas, indicating a priority for risk management for these products, these products not only derived from PA-producing plants but also derived from non-PA-producing plants. Using MOE approach, combined with Haber’s rule was employed to analyse the risks of shorter-than-lifetime exposure for total of 39 herbal teas, 8 herbal medicines, and 19 PFS. This analysis revealed that shorter-than-lifetime use would result in MOE values lower than 10000 upon use for 40 up to 3450 weeks during a lifetime (depending on the preparation). Only for a limited number of herbal teas and medicines, use of two weeks a year (150 weeks during a 75 year lifetime) would still raise a concern. The PBK models were used for translation of in vitro concentration-response curves to in vivo dose-response curves in rat. From these in vivo dose-response curves, the predicted BMDL5-BMDU5 (lower/upper limit of the 90% confidence interval of the benchmark dose that gives a 5% response) of lasiocarpine and riddelline were obtained which were 23.0-34.4 and 4.9-8.4 mg/kg bw/day, respectively. The predicted BMDL5-BMDU5 of lasiocarpine falls well within the range of the point of departures (PoDs) derived from available in vivo toxicity data. The same Quantitative in vitro to in vivo extrapolation (QIVIVE) method was subsequently used to predict the inter-species and inter-ethnic human differences in liver toxicity of lasiocarpine and riddelliine using a human PBK model. The BMDL5-BMDU5 of lasiocarpine were found to amount to 14.7-41.2 mg/kg bw/day for Chinese and 7.4-23.7 mg/kg bw/day for Caucasian, indicating the Chinese to be less sensitive. The predicted BMDL5-BMDU5 of riddelliine were 1.0-5.9 mg/kg bw/day for Chinese and 0.2-1.2 mg/kg bw/day for Caucasian. These values were subsequently compared to those previously obtained in rat to evaluate inter-species differences. The inter-species differences amounted to 2.0-fold for lasiocarpine and 8.2-fold for riddelliine with humans being more sensitive than rats. When extending the developed PBK modelling to predicted in vivo genotoxicity for lasiocarpine and riddelliine in rat, the predicted BMD10 for lasiocarpine and riddelliine amounted to 8.82 and 3.41 mg/kg bw/day, respectively, and were in line with the experimental data on in vivo genotoxicity available in the literature for these two PAs.

Conclusions and implications: The risk assessment of herbal products revealed that daily life-time consumption of some of these products would be a priority for risk management. When considering realistic exposure scenarios, exposure to most of these herbal products would have MOE values higher than 10000 indicating a low priority for risk management. Moreover, this thesis demonstrated that the combined in vitro PBK modelling-based reverse dosimetry approach could adequately predict in vivo liver toxicity and genotoxicity for two model PAs, lasiocarpine and riddelliine. Such QIVIVE methods may prove to be of use in defining more realistic relative potency values for the different food/feed-related PAs. The results obtained reveal the feasibility of this combined quantitative in vitro-in silico approach to determine a PoD for a chemical without the use of experimental animals and to address the issue of how to use in vitro data for risk assessment.