**4. Discussion**

Here, we investigated the PA levels in 21 types of (herbal) teas that were derived from both PA- and non-PA producing plants, based on which we performed a risk assessment using the MOE approach. In addition, we applied the REP factors for each PA that were analyzed to correct the PA concentrations and further calculated EDI and MOE for the (herbal) teas based on a lifetime and shorter duration exposure. Overall, the PA-containing plants-derived teas present a much wider range of the measured total PA levels from 13.4 to 286,682.2 μg/kg d.m., as compared to the levels ranging from 15.0 to 845.1 μg/kg d.m. for the teas from non-PA producing plants. This is in line with the findings reported by Griffine et al. (2014) [1] and Mulder et al. (2015) [13]. When taking REP factors into consideration, most of the samples showed decreased PA levels and, correspondingly, elevated the EDI and MOE values. Nevertheless, the daily consumption of tephroseris, borage and lemon balm (melissa) tea during a lifetime may pose a potential risk to human health.

One of the advantages in the present study is that a comprehensive set of 68 individual 1,2-unsaturated PAs was included, because we intend to reduce the risk of missing relevant PAs as much as possible. We found that PAs were occurring in 19 out of 21 types of the (herbal) teas, albeit that 23 PAs were absent from all the tested samples. Multiple studies on risk assessment for (herbal) teas were based on a set of 17–28 PAs [2,3,13]. Several PAs that were enriched in the (herbal) teas analyzed in this study had not been included in those previous studies, such as atropine, petasitenine, neosenkirkine, and integerrimine, supinine, echinatine and *N*-oxide isomers of these three PAs. It is conceivable that due to the omission of the major PAs, the reported total PA levels in the above-mentioned studies would be lower than those measured in the same type of (herbal) teas based on our method. For instance, the total PA level for borage tea has been shown to be 29,694 μg/kg d.m. based on the set of 28 PAs [3], while this amounted to be 167,846.6 μg/kg d.m. when using the set of abundant PAs in this study (Table 2). Of note, our data showed that supinine plus supinine *N*-oxide, which were usually omitted in previous studies, accounted for about 41% of the total PA levels for borage tea. Neosenkirkine was also missing in previous studies but it turned out to be the major PA contributor in five types of teas in the present study. A lower number of analyzed PAs and a lower analytical sensitivity have been implicated with a greater difference in the derived MOE values based on lower bound exposure estimates in the tea samples [3]. Therefore, it may suggest that the set of sufficient types of PAs should be included to improve the accuracy of the evaluation on the total PA levels occurring in (herbal) teas and the associated health risks.

It Is well accepted that the potency to induce toxicity may be different from one type of PA to another due to distinct metabolisms and the toxic effects of PA metabolites. To approach the improved accuracy of the risk assessment due to the exposure, we applied the REP factors for each individual PA that were analyzed and obtained the REP-corrected PA levels. By doing so, we found that the risk assessment for several teas was significantly affected by REP correction. For example, the MOE results indicated that a daily intake of lungwort and the consumption of borage for two weeks a year during a lifetime may pose a health risk, while, by REP correction, the same regime of tea consumption would be of no concern (Figure 1). It should be noticed that the interim REP factors used in this study were derived from the genotoxicity data and did not take some physiological conditions into account, e.g., the tumor formation and in vivo toxicokinetics [12]. This may compromise the accuracy of the risk assessment to some extent when using these REP values. Preferably, the REP factors used for the combined exposure to PAs should be derived from in vivo carcinogenicity potencies, which actually are largely lacking so far. In support of this notion, the Joint FAO/WHO Expert Committee on Food Additives (JECFA) and ESFA considered that the existing data are not sufficient to identify REPs for different PAs [20,26]. Hence, more researches aiming for actual carcinogenicity data that are capable of deriving REP factors for a large set of PAs should be fully encouraged in the future.

To facilitate the risk management for PAs in botanicals and botanical preparations, several organizations have established the regulation to define a maximum daily use and TDI for PA levels. For example, the BfR in Germany and RIVM in the Netherlands proposed a maximum limit for intake of PAs with 0.1 μg/day for a long-term (over six weeks) exposure scenario [24,25]. In the present study, the intake of PAs resulted from the daily consumption of nine types of (herbal) teas that exceeded the maximum limit set by the BfR and RIVM, even after REP correction (Table 2). It should be pointed out that an intake of PAs at 1 μg/day for a lifetime by a 70 kg person would result in an MOE of 16,600, which provides a sufficient safety margin. With respect to the non-cancer effects of PAs, a NOAEL of 0.01 mg/kg bw/day for hepatocyte cytomegaly was derived based on a long-term rat study [18,27]. According to that study and considering the safety factors, the RIVM provided a TDI of 0.1 μg/kg bw/day to indicate an exposure level of PAs that may cause non-carcinogenic effects. From a view of mechanism of action, the occurrence of pyrrole-protein adducts was thought to be the primary cause for PA-induced liver damage, as manifested in both humans and rodents in vivo studies [28–33]. Based on the REP-corrected EDIs calculated in our study, consuming one cup of tephroseris tea daily could be likely to induce non-cancer toxicity to human health. Altogether, our data

suggest that the daily consumption of tephroseris tea may pose both carcinogenic and non-carcinogenic risks to human health, and thus a risk management may be needed.

It should be noted that multiple s"udie' have been done with regards to risk assessment for PAs in (herbal) tea infusions. In 2017, the EFSA Panel on Contaminants in the Food Chain (CONTAM Panel) assessed the risks related to the presence of PAs in tea infusions as well as the other PA-containing foodstuff [20]. To ensure the actuality and robustness of EDI and MOE values for different age groups of the population, the CONTAM Panel employed at least six dietary surveys and over 60 observations per age group, as highly suggested before [34]. With the well-defined consumption survey data, the Panel was able to derive mean and the 95th percentile values of EDI and MOE for different types of teas consumed by the adult (referring to adults, the elderly and very elderly) and young (infants, toddlers and other children) population. For example, based on the chronic mean exposure levels, MOE values were ranged from 4300 to above 1,000,000 and from 1000 to over 1,000,000 for the adult and young population, respectively [20]. In this study, we mainly focused on the individual MOEs for (herbal) teas, assuming that one cup of tea per day is representative of most common behavior and there is no additional exposure to PAs from any other sources. One could argue that our approach may tend to underestimate the intake amounts of PAs and the related risks, since individuals may consume more (herbal) teas and/or add honey, a foodstuff that generally contains abundant PAs [34–36]. Indeed, apart from four teas that may raise a concern, the daily consumption of two cups of chamomile tea and three cups of rooibos tea during a lifetime would result in an MOE value lower than 10,000 based on the measured PA levels in this study (data are not shown). We did not include the young population in the risk assessment for a life-time exposure of PAs from (herbal) tea intake, as we argue that there is an overestimation when considering the fact that their body weights are remarkably increasing before they reach adulthood. It is also worthwhile to mention that we adopted the BMDL10 of riddelliine for the MOE calculation, which agrees with the proposal by the CONTAM Panel, resulting in additionally increased MOE values by a factor of 3.4 compared to that that for lasiocarpine of 70 μg/kg bw/day in previous studies [14,15,20,37,38]. Since the MOE approach was proposed by the EFSA in 2005, this approach has been either employed to perform a risk assessment for (herbal) teas or for comparison with the other methods [15,20,39–42].

It is challenging to perform a risk assessment for the combined exposure to PAs due to the intake of different types of (herbal) teas and other PA sources. For instance, a risk assessment report has shown that there are a group of subjects with high PA exposure due to the consumption of PA-containing teas and honey [20]. In addition, herb medicines, which generally contain a large amount of PAs, are used in a group of patients or during a specific period [21,43,44]. The evaluation of more complex scenarios, such as a shorterthan-lifetime exposure, is also a challenging issue in the field of risk assessment. It should be acknowledged that a dedicated survey on the consumption habits of (herbal) teas by the average population and by the 95th percentile population (heavy consumers) should be conducted and will contribute to risk assessment for the exposure to PAs.
