Development, validation and routine application of the in vitro REA and DR-CALUX reporter gene bioassays for the screening of estrogenic compounds and dioxins in food and feed

A dedicated cell-line was developed by the Department of Toxicology of Wageningen University in a joined project with the University of California in Davis and the RIKILT-WUR - Institute of Food Safety in Wageningen. This DR-CALUX ^® bioassay was tested, optimised and validated for its use to determine low elevated levels of dioxins in bovine milk around the existing limits. It was shown that this mammalian cell based test is very sensitive for 2,3,7,8-substituted dioxins and related PCBs, thereby reflecting the relative potencies (TEF) of these compounds as set by the World Health Organisation (WHO). These toxic equivalency factors (WHO-TEFs) express the toxicity of a compound in comparison to the most toxic compoundcongener 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD, TEF=1). The response obtained with a mixture of dioxins was additive, in accordance with the TEQ-principle. Milk fat was isolated by centrifugation followed by clean-up of the fat with n-pentane, removal of the fat on a 33% H₂SO₄silica column, and determination of Ah receptor agonist activity with the DR-CALUX ^® bioassay. To investigate the performance of this 33% H2SO₄silica method, milk fat was cleaned with activated carbon and spiked with a mixture of 17 different 2,3,7,8-substituted PCDD and PCDF congeners at 1, 3, 6, 9, 12 and 15 pg WHO-TEQ per g fat, as confirmed by GC/MS. In this concentration range, the method showed a recovery of TEQs around 67% (58-87%). The reproducibility showed a coefficient of variation (CV) varying between 4% and 54%, with the exception of the sample spiked at 1 pg WHO-TEQ/g (CV 97%). The repeatability determined with the sample spiked at 6 pg WHO-TEQ per g showed a CV of 10% These results clearly demonstrate that the reproducibility of the silica-CALUX procedure with samples containing more than 1 pg WHO-TEQ per g fat is relatively good, in particular regarding the fact that no internal standards could be used in the assay for correction of data for varying recoveries. The fact that the CV was much higher for the sample spiked at the lowest level, confirmed the calculated limit of quantification of 1 pg WHO-TEQ per g fat. The current tolerance limit for bovine milk in the EU is 3 pg WHO-TEQ per g fat, with an action limit of 2 pg WHO-TEQ per g fat. Therefore, the DR-CALUX ^® bioassay can be a useful pre-screening tool for selecting milk samples that may contain dioxin levels exceeding this tolerance limit. This was supported by the results obtained with 22 field samples, since all five samples exceeding the 2 pg WHO-TEQ per g fat concentration gave a higher response in the DR-CALUX ^® bioassay.The DR-CALUX ^® bioassay in combination with the 33% H2SO₄clean-up procedure results in a specific test for the determination of dioxins and dioxin-like PCBs, allowing the screening of relatively large sets of samples for the presence of unacceptable high levels of these compounds. This results in a reduction of costs involved in the analysis of food for the presence of these compounds, enabling more intense monitoring programs.Following the successful optimisation and validation of the test for milk fat, the bioassay was first used at RIKILT in the food and feed area during the 1998 Brazilian citrus pulp incident. The test procedure was subsequently optimised and validated for animal feed. During the German bakery waste incident in 2003, animal feed was contaminated with dioxins due to the use of waste wood for drying of the material. Besides Germany, the material was also shipped to the Netherlands. Levels up to 12 ng WHO-TEQ/kg were detected, being about 15 times over the current limit of 0.75 ng WHO-TEQ/kg. A combined strategy of screening with the DRCALUX ^® -bioassay and the HRGC/HRMS confirmatory method was used in the Netherlands to rapidly control the incident. Pigs were contaminated by the incident but only to a very limited extent. Despite the rather low limits for pig meat (1 pg WHO-TEQ per g fat), the DR-CALUX ^® bioassay, in combination with an extra acid pre-treatment of the fat samples, showed excellent performance, confirming once again the value of this bioassay. Shown during the recent incidents with kaolinic clay (2004) and the contaminated HCl used for gelatine production (2006), the assay is still the best availablescreening test for dioxins and dioxin-like PCBs.The second aim of the research in this thesis was to develop, validate and apply a new recombinant yeast screen to detect chemicals with an estrogenic mode of action in animal feed, urine and illegal preparations. A recombinant yeast cell that stably expresses the human estrogen receptor α (hERα) and yeast enhanced green fluorescent protein (yEGFP) as a reporter protein in response to estrogens was developed at the RIKILT. The EC50 revealed by the RIKILT yeast Estrogen bioAssay (REA) was 0.5 nM for 17b-estradiol and was comparable with reported EC50 values for yeast estrogen bioassays that contain β-galactosidase as a reporter. However, the yEGFP assay can be performed completely in 96 well plates within 4 hours and does not need require cell wall disruption, nor does it need the addition of a substrate. This makes the test sensitive, rapid and convenient with high reproducibility and small variation. The robustness and ease of the yeast cells in combination with the qualities ofyEGFP,ensure that the assay will be suited to be used as a high through put system.The properties of the RIKILT yeast Estrogen bioAssay expressing thehERα(REA) were further studied by testing a series of estrogenic compounds. In addition, a similar assay was developed based on the stable expression of human estrogen receptor β (hERβ). When exposed to 17b-estradiol, the maximum transcriptional activity of the hERb cytosensor was only about 40% of the activity observed with hERa, but the concentration where half-maximal activation is reached (EC50), was about 5 times lower. The relative estrogenic potencies (REP), defined as the ratio between the EC50 of 17b-estradiol and the EC50 of the compound, of the synthetic hormones dienestrol, hexestrol and especially mestranol were higher with ERa than with ERβ, while DES was slightly more potent with ERb. The gestagens progesterone and medroxyprogesterone-acetate showed no response, whereas the androgen testosterone showed a very weak response and only at high concentrations. The isoflavones genistein, genistin, daidzein and daidzin, the coumestran coumestrol and the flavonoid naringenin were relatively more potent with ERb than with ERa. Coumestrol and genistein were by far the most potent of these compounds with ERb. However, 8-prenylnaringenin, a phytoestrogen present in hops, was relatively more potent with ERa than with ERb and was actually the most potent phytoestrogen with ERa. The data demonstrate that the REA shows clear dose-response curves when exposed to estrogenic compounds. Since good dose-response curves can be obtained after only 4 h of exposure, the often questioned permeability of the yeast cell wall does not seem to be an obstacle in our yeast estrogen bioassays.The RIKILT yeast Estrogen bioAssay stably expressing human estrogen receptor α (REA) was validated as a qualitative screening method for the determination of estrogenic activity in calf urine and animal feed. These validations were performed according to EC Decision 2002/657, which prescribes the determination of the detection capability (CCb), the specificity and the stability. To determine these performance characteristics, twenty blank urine samples of 19 week old calves were collected and spiked with 17b-estradiol (E2b) at 1 ng/ml-1, diethylstilbestrol (DES) at 1 ng/ml-1, 17a-ethynylestradiol (EE2) at 1 ngml-1a-zearalanol at 50 ngml-1 or mestranol at 10 ng/ml-1. Following enzymatic deconjugation and solid phase extraction, 100 ml equivalents of these blank and spiked urine samples were screened for estrogenic activity in a 96 well plate using the REA. All of these blank and low estrogen spiked feed samples fulfilled the CCα and CCβ criterions, meaning that all 20 blank urine samples gave a signal below the determined decision limit CCα and were thus classified as compliant and at least 19 out of the 20 spiked samples gave a signal above this CCα (β=5%) and were thus classified as suspect. The specificity of the method was determined with blank urine samples spiked with a high dose of testosterone or progesterone (1000 ng/ml-1). No response to these substances was detected in the REA. There was also no interference of a high dose of testosterone or progesterone on the response of a low dose of the estrogens. Stability of urine samples was checked with spiked urine samples that were kept frozen for up to 90 days, showing that urine samples could be stored at -20°C for up to 60 days without changing the screening result of the assay. The assay was validated for animal feed in a simalar way, using twenty blank animal feed samples, including milk replacers and wet and dry feed samples.As all the performance characteristics met the criteria that were put forward in EC Decision 2002/657 for validation of a qualitative screening method, the described clean-up/yeast estrogen bioassay procedures were proven to be valid for the determination of estrogenic activity in calf urine and animal feed. The clean-up procedures for urine and feed samples are relatively simple and the yeast estrogen bioassay, using yEGFP as a reporter protein, is sensitive, rapid, convenient and reproducible. Due to the good sensitivity of the bioassay, only 2 ml of urine or 1 gram of feed were enough to be processed. Combined this resulted in a low cost bioassay that is suited to be used as a high through-put system for the screening of estrogenic activity in calf urine and animal feed. Like the DR CALUX ^® assay, Tthe method acquired an ISO 17025 accreditation status in the Netherlands for both of these matrices. The examples of the MPA-incident with wet pig feed and the fishfeed,described in Chapter 7 of the thesis, demonstrate the applicability of the bioassay method as an early warning system for pharmaceutical waste and hormone use respectively. This is the first successful example of a developed, validated and applied bioassay for the screening of hormonal substances in feed. At present this method has been in routine use at RIKILT for more than two years.Overall the work presented in this thesis shows that bioassays are valuable tools for rapid and high throughput screening of samples for both known and unknown compounds. As such they may contribute to an earlier detection of new emerging risks and prevent the use of illegal growth-promoting agents with thus far unknown identity.