1RIKEN Cell Bank, The Institute of Physical and Chemical Research (RIKEN), 3-1-1
In October 1992, the Japanese Society of Alternatives to Animal Experiments (JSAAE) organized an inter-laboratory validation study on cytotoxicity assays for development of an alternative(s) to the in vivo Draize eye irritation test. The main purpose is to evaluate practicability of five proposed cytotoxicity assays with two cell lines each through a large scale inter-laboratory assessment. The five assays were colony formation (CF) assay, crystal-violet staining (CV) assay, lactate dehydrogenase release (LDH) assay, neutral red uptake (NR) assay, and MTT assay. They were selected because of their popularity in Japan. We chose six detergents as model chemicals for this first step validation. One of the six chemicals, which was revealed to be Tween 20 after breaking the code, was additionally included in the doubly-mask-coded chemical set for simultaneous evaluation of intra-laboratory variation. Technology transfer for all the assays was made cumulatively 80 times to less-experienced laboratories. Concomitantly, we performed the Draize test to confirm toxicity of the coded chemicals in vivo. A total of 3,810 final data files including preliminary test results were submitted from 42 laboratories. Of 1,535 raw data files with final definitive assay results, 292 files were rejected because of not only apparent misunderstanding of the protocols provided by the Working Group but also for violating pre-set acceptance rules. Acceptability of data files was also examined by a computer-assisted logistic analysis program (LAP-JSAAE) with a six step-wise check code to detect abnormality in the data file. After generating ED50 values through the program, 5 data file sets of 7 tested chemicals were judged unreliable since the large differences in ED50 values were found for the same but differently coded test chemical, Tween 20. This clearly indicates considerable intra-laboratory variation. After excluding these data files, analyses of inter-laboratory variation were made on 969 data files with the box- whisker plot analysis. The important results of our study are as follows. (1) CF, CV, MTT, and NR assays are recommendable from the view point of performance of these assays. Performance rate of each assay was calculated on the number of finally accepted assay data files divided by the expected number of data files. The highest rate was for the CF assay with BALB/3T3 A31-1-1 cells followed by the CV assay with two cell lines. Lower performance rates were observed in the sub-divided LDH assays. The performance rate was considered to reflect simplicity of the method and labor needed for the assay; (2) From the view point of the intra-laboratory variation of the same but differently coded chemical Tween 20, medians of the log(ED50) values of each assay were satisfactorily close; (3) After eliminating the sub-divided LDH assays which gave a small number of acceptable data files per assay and therefore resulted in unstable hinge-spread of log(ED50) values, the CV assay with CHL cells and the MTT assay with SQ-5 cells were found to have given the smallest mean hinge-spread of log(ED50) followed closely by the CF assay with HeLa S3 (SC) cells and the CV assay with HeLa S3 (SC) cells. These assays were therefore considered to give small inter- laboratory variation; (4) the CF assay with HeLa S3 (SC) cells resulted in the largest "power for distinction" of toxicities between the least and the most toxic chemicals defined as the ratio of difference in medians of log(ED50) values of two chemicals to the mean hinge-spread, followed by the CV assay with HeLa S3 (SC) cells. However, the CF assay is not necessarily advantageous as far as distinguishing moderately irritating chemicals from non-irritating chemicals; (5) Medians of log(ED50) values enabled us to classify tested chemicals into at least three categories, namely, non-, weakly-, and highly-cytotoxic chemicals. Non- and highly-cytotoxic chemicals corresponded to non- and severe-irritants in the in vivo Draize test. (6) Considering performance rate, inter-laboratory variation of data reflected on the mean hinge-spread, power for distinction of chemical cytotoxicity, time needed for an assay (i.e. the CF assay requires longer incubation time than others), and the data on the common cell line HeLa S3 (SC), we concluded that the CV assay is the most practical and recommendable as a part of alternatives to the in vivo Draize test. Many problems that were revealed during the present validation study including human factors were discussed. The assay results will be further described in the ensuing articles in this issue, i. e., the calculation of ED50 values by LAP-JSAAE, problems on the CF, CV, LDH, MTT, and NR assays. A fact-data base was constructed on the data files of this validation study which will be available on request.