47 Appendix 4, Regulation (EU) 2019/1390

In forcesince 31 Jul 2019
1.

The hrERα competitive binding assay measures the binding of a single concentration of [3H]-17β-estradiol in the presence of increasing concentrations of a test chemical. The competitive binding curve is plotted as specific [3H]-17β- estradiol binding versus the concentration (log10 units) of the competitor. The concentration of the test chemical that inhibits 50 % of the maximum specific [3H]-17β-estradiol binding is the IC50.

2.

Data from the control runs are transformed (i.e. percent [3H]-17β-estradiol specific binding and the log concentration of the control chemical) for further analysis. Estimates of log(IC50) values for the positive controls (e.g. reference estrogen and weak binder) should be determined using an appropriate nonlinear curve fitting software to fit a four parameter Hill equation(e.g. BioSoft; GraphPad Prism) (2). The top, bottom, slope, and log(IC50) can typically be left unconstrained when fitting these curves. Robust regression should be used when determining the best fit unless justification is given. The method chosen for robust regression should be stated. Correction for ligand depletion was not needed for the FW or CERI hrER assays, but may be considered if needed. Following the initial analysis, each binding curve should be reviewed to ensure an appropriate fit to the model. The relative binding affinity (RBA) for the weak binder can be calculated as a percent of the log (IC50) for the weak binder relative to the log (IC50) for 17β-estradiol. Results for the positive controls and the non-binder control should be evaluated using measures of assay performance and acceptability criteria as described in this test method (paragraph 20), Appendix 2 (FW Assay, paragraphs 41-51) and Appendix 3 (CERI Assay, paragraphs 41-51). Examples of 3 runs for the reference estrogen and weak binder are shown in Figure 1.

3.

Data for all test chemicals should be analysed using a step-wise approach to ensure that data are appropriately analysed and that each competitive binding curve is properly classified. Each run for a test chemical should initially undergo a standardised data analysis that is identical to that used for the reference estrogen and weak binder controls. Once completed, a technical review of the curve fit parameters as well as a visual review of how well the data fit the generated competitive binding curve for each run should be conducted. During this technical review, the observations of a concentration dependent decrease in the percent [3H]-17β-estradiol specifically bound, low variability among the technical replicates at each chemical concentration, and consistency in fit parameters among the three runs are a good indication that the assay and data analyses were conducted appropriately. Professional judgment should be applied when reviewing results from each run for a test chemical, and the data used to classify each test chemical as a binder or non-binder should be scientifically defensible.

4.

Occasionally, there may be examples of data that require additional attention in order to appropriately analyse and interpret the hrER binding data. Previous studies had shown cases where the analysis and interpretation of competitive receptor binding data can be complicated by an upturn of the percent specific binding when testing chemicals at the highest concentrations (Figure 2). This is a well-known issue that has been encountered when using protocols for a number of competitive receptor binding assays (3). In these cases, a concentration dependent response is observed at lower concentrations, but as the concentration of the test chemical approaches the limit of solubility, the displacement of [3H]17β-estradiol no longer decreases. In these cases, data for the higher concentrations indicate that the biological limit of the assay has been reached. For example, this phenomenon is many times associated with chemical insolubility and precipitation at high concentrations, or may also be a reflection of exceeding the capacity of the dextran-coated charcoal to trap the unbound radiolabelled ligand during the separation procedure at the highest chemical concentrations. Leaving such data points in when fitting competitive binding data to a sigmoid curve can sometimes lead to a misclassification of the ER binding potential for a test chemical (Figure 2). To avoid this, the protocol for the FW and CERI hrER binding assays includes an option to exclude from the analyses data points where the mean of the replicates for the percent [3H]17β-estradiol specific bound is 10 % or more above the mean observed at a lower concentration (i.e. This is commonly referred to as the 10 % rule). This rule can only be used once for a given curve, and there must be data remaining for at least 6 concentrations such that the curve can be correctly classified.

5.

The appropriate use of the 10 % rule to correct these curves should be carefully considered and reserved for those cases where there is a strong indication of a hrER binder. During the conduct of experiments for the validation study of the FW hrER Binding Assay, it was observed that the 10 % rule sometimes had an unintended and unforeseen consequence. Chemicals that did not interact with the receptor (i.e. true non-binders) often showed variability around 100 % radioligand binding that was greater than 10 % across the range of concentrations tested. If the lowest value happened to be at a low concentration, the data from all higher concentrations could potentially be deleted from the analysis by using the 10 % rule, even though those concentrations could be useful in establishing that the chemical is a non-binder. Figure 3 show examples where the use of the 10 % rule is not appropriate.

(1)

OECD (2015). Integrated Summary Report: Validation of Two Binding Assays Using Human Recombinant Estrogen Receptor Alpha (hrERα), Health and Safety Publications, Series on Testing and Assessment (No 226), Organisation for Economic Cooperation and Development, Paris.

(2)

Motulsky H. and Christopoulos A. (2003). The law of mass action, In Fitting Models to Biological Data Using Linear and Non-linear Regression. GraphPad Software Inc., San Diego, CA, pp 187-191. Www.graphpad.com/manuals/Prism4/RegressionBook.pdf

(3)

Laws SC, Yavanhxay S, Cooper RL, Eldridge JC. (2006). Nature of the Binding Interaction for 50 Structurally Diverse Chemicals with Rat Estrogen Receptors. Toxicological Sci. 94(1):46-56.