Case study on the use of integrated approaches to testing and assessment for the prediction of a 90-day repeated dose toxicity study (OECD 408) for 2-Ethylbutyric acid using a read-across approach from other branched carboxylic acids.
Regulatory framework: In this read-across we assume, that 2-Ethylbutyric acid (2-EBA) has to be registered under REACH and is produced in Europe at tonnages of more than 100 t/a. The standard REACH information requirements ask for a 90 days study with oral exposure. We use a category approach to predict the outcome of a subchronic toxicity study. The category comprises nine branched carboxylic acids.
Synopsis: The structure of the target compound 2-EBA comprises a short chain, branched aliphatic carboxylic acid in position 2. Nine aliphatic carboxylic acids with different branched aliphatic side chains are regarded as most similar to the target compound. Beside high structural similarity the grouped compounds show a consistent trend for physicochemical (pc) parameters, e.g. logPow and MW increases slightly with side chain length, whereas water solubility and vapour pressure decreased. The pc-parameters do however not alert for a potential bioaccumulation in vivo. Two compounds have in vivo animal studies with repeated oral exposure. 2-Ethylhexanoic acid (2-EHA) has subchronic guideline studies, in which liver hypertrophy was observed together with an increase of the relative liver weight. Valproic acid (VPA) induced liver steatosis in shorter-term subacute studies. The read-across hypothesis is therefore, that 2-EBA is a liver toxicant with special concern for steatosis. In addition to the nine structural analogues, Pivalic acid (PVA) is tested as negative control compound. PVA has a third substituent in position 2 and did not induce any liver toxicity in a subacute study up to the highest tested dose. A negative compound is needed to judge on the accuracy of NAM data.
The selection of NAMS and testing scope is dependent on the read-across hypothesis as recently published (Escher &Kamp et al. 2019). In example, the grouped compounds might have in common i) an AOP (adverse outcome pathway), ii) a specific toxicological adverse effect or iii) an unspecific toxicological effect/ a generally low toxicity. In this read-across assessment, an AOP network for liver steatosis is known for the primary toxicological effect. From this AOP network, molecular initiation events (MIEs) and one key event (KE) are tested to see, in how far the grouped compounds might induce this adverse outcome. In addition, the perturbation of general biological pathways and cellular functions are tested together with cytotoxicity to discover potential major differences between the grouped compounds (section 4.).
NAM data showed a consistent trend with regard to toxikokinetics and toxikodynamics within the grouped compounds. The results are briefly summarised in the following:
Toxikokinetics: A rat physiology-based pharmacokinetic (PBPK) model was established, based on in vivo data, and used to calculate plasma and target organ concentrations, which guided the selection of a relevant concentration range for in vitro testing. Human PBPK models were established for all read-across compounds based on physiochemical properties and in vitro clearance data (e.g. plasma protein binding (ppb) and intrinsic hepatic clearance (CLint, Hep). Human in vivo pharmacokinetic data for VPA was identified and verified good predictive performance based on observed plasma concentration data in humans. Based on this proof of concept IVIVE-PBPK models were used for in vitro to in vivo extrapolations for all analogues.
Toxikodynamics: Several adverse outcome pathways are available describing the development of liver steatosis. About 50 published signalling pathways leading to steatosis ENV/JM/MONO(2020)20 ? 9 Unclassified were compiled from literature and summarised in an adverse outcome pathway (AOP) network. The AOP network guided the selection of in vitro assays, to determine MIE (molecular initiation event) and KE (key event) activation. Two high throughput models, the CALUX and GFP reporter assays, measured six MIEs being present in the AOP network. With increasing chain length, the number of activated MIEs related to steatosis increased. The target compound 2-EBA activated one MIE, PPAR-?. It can therefore not be excluded that a pathway towards lipid accumulation is activated by 2-EBA. In addition, three liver models measured intracellular triglyceride accumulation, a key event regarded as direct surrogate for liver steatosis. After single and/or repeated exposure, lipid accumulation was mainly observed for long chain analogues, whereas short-chain analogues remained inactive. The two compounds with in vivo data on liver steatosis induced lipid accumulation, whereas the in vivo negative compound was inactive up to the highest in vitro tested dose. 2-EBA was inactive in HepG2 and HepaRG cells (primary human hepatocytes were not measured) up to the highest in vitro tested dose. No difference was observed within the grouped analogues with regard to cytotoxicity (in liver and kidney cells), MIEs not present in the AOP network or other endpoints pointing towards general biological perturbations (e.g. glutathione depletion, mitochondrial membrane potential, mitochondrial superoxide formation etc.).
Data integration: The decision theory Dempster Shafer (DS) indicated that the absence of lipid accumulation for 2-EBA can be predicted with 100% certainty from the in vitro assays used in this case study. DS further showed that the results of lipid accumulation and cytotoxicity from HepG2 cells and the Calux reporter gene panel give already enough information for this conclusion.
Conclusion: We have shown in this dossier that the NAMs predict all three compounds with in vivo data correctly, either those that induce or do not induce liver steatosis. 2-EBA was in all assays less toxic then the two liver toxic analogues with in vivo animal data, 2-VPA and 2-EHA. The NAM data investigated in this case study indicate that 2-EBA will not induce liver steatosis up to the highest tested in vitro dose. Kidney models did not show any difference in cytotoxicity, with all compounds being of general low cytotoxicity. Also, endpoints measuring the perturbation of general biological processes like mitochondrial membrane potential or cytotoxicity support the trend of higher activity with longer side chain length. We used the 10th percentile of the most sensitive in vitro endpoint of 2-EBA to derive an oral equivalent dose. QIVIVE results in an oral equivalent dose of 730 to 948.6 mg/kg bw/d for rats, which might be used to fill the data gap of a subchronic toxicity study. Furthermore, QIVIVE was used to determine directly a corresponding human oral equivalent dose, which is estimated to be 243 to 245.7 mg/kg bw/d.